CN118657461A - Garbage transport vehicle unloading supervision system and method based on video monitoring big data - Google Patents

Abstract

The invention discloses a system and a method for supervising the unloading of a garbage transport vehicle based on video monitoring big data, in particular relates to the field of supervising the garbage unloading process, and aims to solve the problem of all-weather supervising of the garbage transport vehicle unloading process, and improve the supervising efficiency and transparency of the garbage transport process through an image processing technology and a data analysis means. Firstly, the key operation area of the garbage truck is identified and monitored through an image processing technology, so that the key area is ensured to be effectively covered. And (3) calculating the cumulative influence of all events in a specific time window by defining a weighted influence value of each positive return of the camera and combining binomial distribution and linear regression analysis, and evaluating the long-term stability of the combined part of the camera and the vehicle. And finally, identifying and recording the unloading process in real time, and marking the potential abnormal behavior in the unloading process. Thereby improving the compliance and environmental friendliness of the garbage unloading process, greatly supporting the monitoring and decision of the urban management department and promoting the sustainable development of urban environment.

Description

Garbage transport vehicle unloading supervision system and method based on video monitoring big data

Technical Field

The present invention relates to the field of garbage transportation supervision, and more specifically, to a garbage transportation vehicle unloading supervision system and method based on video monitoring big data.

Background Art

With the acceleration of urbanization, the amount of urban waste generation has increased dramatically, and waste disposal has become a serious challenge in urban management. In the traditional waste transportation and treatment process, common problems include insufficient optimization of transportation routes, low transportation efficiency, and insufficient transparency of the unloading process. Especially in the unloading process, due to the lack of effective monitoring methods, incomplete unloading or illegal dumping of garbage often occurs, which not only damages the urban environment but also may threaten public health. At present, the supervision of garbage transportation vehicles mainly relies on basic GPS tracking and manual monitoring. Although these methods can track vehicles, they have obvious limitations in ensuring the transparency and efficiency of the unloading process. For example, the GPS system cannot provide a detailed view of the unloading behavior, while manual monitoring is labor-intensive and inefficient. In addition, existing systems generally lack real-time data processing and analysis capabilities, resulting in delayed regulatory responses and difficulty in effectively preventing and solving problems.

In order to solve the above problems, a technical solution is now provided.

Summary of the invention

In order to overcome the above-mentioned defects of the prior art, the embodiments of the present invention provide a garbage transport vehicle unloading supervision system and method based on video monitoring big data, and improve the supervision efficiency and transparency of the garbage transportation process through image processing technology and data analysis. First, the key operating areas of the garbage transport vehicle are identified by image processing technology, and the pitch angle and yaw angle of the camera installed on the vehicle are dynamically adjusted to ensure that these key areas are effectively covered. The camera transmits video data to the central server in real time to ensure real-time updating and storage of monitoring data. On the central server, by defining the weighted impact value of each camera return and calculating the cumulative impact of all events within a specific time window, binomial distribution and linear regression analysis are further used to monitor and evaluate the long-term stability of the camera and vehicle combination. In addition, the vehicle's GPS data, weight sensor data and video monitoring data are integrated and synchronized to more accurately capture and analyze unloading behavior. By applying data analysis technology, potential abnormal behaviors in the unloading process, such as illegal dumping, incomplete unloading, etc., are identified and marked in real time, and detailed reports are automatically generated and archived for these abnormal behaviors. This not only improves the compliance and environmental friendliness of the garbage transportation and unloading process, but also helps urban management departments to monitor and support decisions more effectively, and ultimately promotes the sustainable development of the urban environment to solve the problems raised in the above background technology.

To achieve the above object, the present invention provides the following technical solution: Step S1, using image processing technology to identify and mark key operating areas of a garbage transport vehicle, dynamically covering these areas with cameras at designated locations on the vehicle, and transmitting video data to a central server in real time by adjusting the pitch angle and yaw angle of the camera in real time in combination with the network;

Step S2: On the central server, a weighted impact value of each camera return is defined, the cumulative impact of all events within the specified time window is calculated, and the long-term trend of the events is analyzed by binomial distribution and linear regression to monitor the stability of the combination of camera and vehicle;

Step S3, after confirming that the camera and the vehicle are integrated without any obstacles, integrating and synchronizing the vehicle's GPS data, weight sensor data, and video surveillance data, and applying data analysis technology to identify and mark potential abnormal behaviors;

Step S4: For any abnormal behavior identified, a report is automatically generated and archived.

In a preferred embodiment, step S1 includes the following contents:

S1-1, marking key operating areas on garbage trucks and using image processing technology to identify specific area features;

set up A collection of key operating areas for the vehicle;

,in It is a regional characteristic;

S1-2, install the motor-driven camera to a position that can dynamically cover all key operating areas of the vehicle;

Use the geometric model to determine the initial camera mounting positions so that the initial pitch and yaw angle settings cover all critical operating areas of the vehicle.

S1-3, real-time dynamic adjustment of the camera to respond and maximize coverage of key areas;

For each , calculate its position in the camera's current field of view ;

Set adjustment goals as the center ;

;

;

in, and is the adjustment sensitivity coefficient;

S1-4, configure the network to transmit the camera video to the central server in real time.

In a preferred embodiment, step S2 includes the following contents:

S2-1, based on the amplitude of the correction and the time interval from the last correction, a weighted impact value is defined for each correction;

No. The comprehensive amplitude of the return to positive: ;

The time interval between two corrections: ;

The weighted impact of each reversal: ;

in is the time decay coefficient, which is used to adjust the effect of time interval;

S2-2, calculate the cumulative impact of all return events within a given time window to assess the degree of abnormality;

Defining time windows ;

Cumulative impact of all return times within the time window: ;

S2-3, normalize the abnormal index;

S2-4, using the standardized abnormality index to compare with the abnormality threshold to determine whether there is a significant binding abnormality;

If the normalized abnormality index is greater than the abnormality threshold, it is determined that there is a significant binding abnormality, and a binding abnormality signal is generated.

In a preferred embodiment, S2-5, collecting standardized abnormality indexes in multiple consecutive time windows, setting a series of time windows , and calculate the corresponding standardized anomaly index for each window ;

S2-6, count the frequency of the standardized anomaly index greater than the anomaly threshold in all observation windows, and mark it as the anomaly frequency;

S2-7, use binomial distribution test to determine whether the frequency of abnormal occurrence is significantly higher than the random level, and judge the consistency and regularity of abnormal occurrence;

The probability of an event occurring in each trial is denoted by ;

Binomial Distribution: ;

in, represents the number of trials, is a random variable;

;in is the actual number of windows greater than the anomaly threshold;

S2-8, linear regression analysis of the change trend of standardized abnormal index over time: ;

in, It represents the slope coefficient, which is used to describe the linear relationship between the independent variable and the dependent variable;

S2-9, if the abnormal frequency is greater than the random level and the slope coefficient is greater than or equal to 0, an overall abnormal signal is generated; otherwise, an overall stable signal is generated.

In a preferred embodiment, step S3 includes the following contents:

S3-1, after confirming that the overall stable signal is obtained, synchronize the video data of the camera, the GPS location data of the vehicle and the weight sensor data in real time, and align the timestamps of the data received from different sources;

S3-2, extract key features from the synchronized data, including the unloading action in the video, the position change and weight change data recorded by GPS;

S3-3, establish a standard model of unloading behavior based on historical data;

S3-4, apply real-time data analysis algorithms to compare current behavior with standard patterns and mark all events exceeding corresponding thresholds as abnormal;

S3-5, record and respond to identified abnormal behaviors.

In a preferred embodiment, S3-2-1, using a convolutional neural network to automatically detect the start and end of the unloading behavior from the video;

S3-2-2, extracting the time and location of vehicle arrival and departure at the unloading point from GPS data;

S3-2-3, based on the weight sensor data, records the weight change before and after unloading.

In a preferred embodiment, S3-3-1, collecting historical unloading behavior data to determine the average duration, common locations, and weight variation range of unloading;

S3-3-2, use clustering to classify the patterns of unloading behavior and use the identified patterns as historical standard patterns.

The garbage transport vehicle unloading supervision system based on video surveillance big data includes a video acquisition optimization module, a stability trend monitoring module, a data synchronization integration module, an abnormal behavior analysis module, and an automatic report generation module;

The video acquisition optimization module uses image processing technology to identify and mark the key operating areas of the garbage truck, installs and adjusts the camera to dynamically cover these areas, and transmits the video data to the central server in real time;

The stability trend monitoring module defines and calculates the weighted impact value of each camera return on the central server, calculates the cumulative impact of all events in the specified time window, and analyzes the long-term trend of these events through binomial distribution and linear regression, monitors the stability of the combination of camera and vehicle, and sends the monitoring results to the data synchronization integration module;

The data synchronization and integration module calls the data from the central server, synchronizes and integrates the GPS data, weight sensor data and video surveillance data from the vehicle, aligns the timestamps of all data points, and transmits the processed data to the abnormal behavior analysis module;

The abnormal behavior analysis module uses data analysis technology to monitor and analyze the current unloading behavior pattern in real time, and compares the real-time data with the stored historical standard pattern to identify and mark potential abnormal behaviors, and sends the identified abnormal behaviors to the automatic report generation module;

The report automatic generation module generates a report for any abnormal behavior identified and automatically archives the report. The technical effects and advantages of the garbage transport vehicle unloading supervision system and method based on video surveillance big data of the present invention are as follows:

1. The present invention evaluates the stability of the combination of the camera and the vehicle by monitoring and analyzing the automatic self-correction events of the camera, thereby ensuring the reliability and effectiveness of the video surveillance system. The process includes calculating the weighted impact value of each self-correction, evaluating the cumulative degree of abnormality within a specified time window, and standardizing the abnormality index for comparison. These data are further compared with the set abnormality threshold to determine the significance of the combined abnormality, and the frequency and trend of the abnormality are determined by statistical tests (such as binomial distribution and linear regression analysis). These steps enable the operation and maintenance team to identify and resolve frequent or severe adjustment needs caused by uneven vehicle operation, unstable camera installation or technical defects. In this way, problems that may lead to the loss of key monitoring data or the decline of monitoring quality can be discovered and corrected in a timely manner, ensuring that the camera system is always in the best monitoring state, improving vehicle safety and monitoring effects, and optimizing the maintenance and operation efficiency of the monitoring system.

2. The present invention realizes real-time monitoring and analysis of unloading behavior by comprehensively utilizing video surveillance data, GPS data, and weight sensor data from the vehicle, ensuring that each unloading behavior is compared with the historical standard pattern to identify and mark behaviors that deviate from the norm. In this process, the start and end of the unloading action are automatically detected, and key indicators such as unloading duration, frequency, and location are compared in real time. Abnormal behaviors are identified through set deviation thresholds, such as unloading time that is too long or too short, and the location deviates from the common unloading point. This real-time data analysis and monitoring can effectively prevent garbage trucks from unloading at non-designated locations and enhance the compliance of the transportation process. In addition, the automatic recording and alarm of abnormal behaviors can promptly notify the operation team to intervene, ensure operational safety and efficiency, and thus greatly improve the response speed and accuracy of operation management.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a schematic diagram of a flow chart of a method for supervising the unloading of garbage transport vehicles based on video surveillance big data according to the present invention;

FIG2 is a schematic diagram of the structure of the garbage transport vehicle unloading supervision system based on video surveillance big data of the present invention.

DETAILED DESCRIPTION

The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

Example 1

FIG1 shows a method for supervising the unloading of garbage transport vehicles based on video surveillance big data according to the present invention, comprising:

Step S1, using image processing technology to identify and mark key operating areas of the garbage truck, dynamically covering these areas with cameras at designated locations on the vehicle, and transmitting video data to a central server in real time through real-time adjustment of the pitch and yaw angles of the cameras in conjunction with the network;

Step S2: On the central server, a weighted impact value of each camera return is defined, the cumulative impact of all events within the specified time window is calculated, and the long-term trend of the events is analyzed by binomial distribution and linear regression to monitor the stability of the combination of camera and vehicle;

Step S3, after confirming that the camera is integrated with the vehicle without any obstacles, integrate and synchronize the vehicle's GPS data, weight sensor data and video monitoring data, and apply data analysis technology to identify and mark potential abnormal behaviors.

Step S4: For any abnormal behavior identified, a report is automatically generated and archived.

Step S1 includes the following contents:

S1-1, marking key operating areas on the garbage truck, such as the discharge port and the interior of the truck;

Use image processing techniques to identify specific regional features, such as shape, color, or specific markings;

set up A collection of key operating areas for the vehicle.

,in It is a regional feature, such as a specific color area or a shape feature.

S1-2, install the motor-driven camera in a position that can dynamically cover all key operating areas of the vehicle.

The geometric model is used to determine the initial camera mounting positions so that the initial settings for pitch and yaw angles cover all critical operating areas of the vehicle.

S1-3, real-time dynamic adjustment of the camera to respond to changes in the vehicle's internal environment and maximize coverage of key areas;

For each , calculate its position in the camera's current field of view ;

Set adjustment goals as the center ;

;

;

in, and is the adjustment sensitivity coefficient.

S1-4, configure the network to transmit the camera video to the central server in real time.

The main reasons for setting key operating areas and enabling the camera to have an automatic self-centering function are as follows:

During the operation of the vehicle, especially when driving on uneven roads, vibrations are inevitable. These vibrations may cause the fixed camera to have a slight position shift or angle change, thus affecting the quality and accuracy of video surveillance. In addition, similar problems may also occur when the operator accidentally touches the camera. By setting the camera to automatically return to the preset key operating area, the system can automatically adjust the position and focal length of the camera to ensure that all key operating areas are always in the best monitoring angle. Such a setting not only increases the reliability of the system, but also ensures the continuity and integrity of the monitoring data, which is crucial for subsequent data analysis, security monitoring and compliance audits.

Analyzing the obstacles between the camera and the vehicle is critical to ensuring the stability and effectiveness of the video surveillance system. By recording and evaluating the camera's automatic self-centering timestamp and self-centering amplitude, video surveillance quality issues caused by abnormal vehicle operation or unstable camera fixation can be identified. Frequent or large-scale automatic self-centering may indicate that the vehicle is experiencing severe vibrations on certain sections of the road, or that there are defects in the camera installation, all of which may lead to the loss of critical monitoring data and affect the accurate recording and analysis of accidents and events. Therefore, regular performance of this analysis can help detect problems in a timely manner, optimize the camera's installation position and stability, and ensure that the monitoring system can provide continuous and clear video streams under various operating conditions, thereby improving overall vehicle safety and monitoring effects.

Step S2 includes the following contents:

S2-1, based on the correction amplitude and the time interval from the last correction, a weighted impact value is defined for each correction.

No. The comprehensive amplitude of the return to positive: ;

The time interval between two corrections: ;

The weighted impact of each reversal: ;

in is the time decay coefficient, which is used to adjust the effect of time interval.

A weighted impact value is defined for each re-alignment, based on the magnitude of the re-alignment and the time interval since the last re-alignment. The purpose of this is to quantify the specific impact of each camera re-alignment event on the system stability, and to consider the time factor to measure the persistence and urgency of the impact. This approach allows for more accurate assessment of stability issues at the interface between the camera and the vehicle, identifying frequent or severe adjustments due to uneven vehicle operation or installation issues. In addition, the application of time decay emphasizes the importance of events in the near term, allowing the system to dynamically adjust its response to emerging issues, thereby ensuring the continuity and reliability of video surveillance.

S2-2, calculate the cumulative impact of all positive return events within a given time window to assess the degree of abnormality.

Defining time windows ;

Cumulative impact of all return times within the time window: .

The purpose of calculating the cumulative impact of all return events within a given time window to assess the degree of anomaly is to comprehensively consider the overall impact of camera stability within a specific time range, thereby providing a quantitative indicator to determine whether the system is operating normally or if there are potential problems. By accumulating the impact of all events within the time window, frequent adjustment needs caused by vehicle vibration, equipment failure or improper installation can be captured, which helps identify weak links in the system or abnormal patterns in operation.

S2-3, the abnormal index is standardized to make it meaningful for comparison.

The normalized anomaly index per unit time provides a time-normalized anomaly metric: .

S2-4, use the standardized abnormality index to compare with the abnormality threshold to determine whether there is a significant binding abnormality.

If the normalized anomaly index is greater than the anomaly threshold, it is determined that there is a significant combination anomaly, which indicates that there is a major problem with the stability of the connection between the camera system and the vehicle. This situation may be due to unstable camera installation, excessive vibration of the vehicle, or frequent adjustments caused by technical problems with the camera itself. An anomaly index exceeding the threshold means that the number and amplitude of adjustments required for the camera in a given period of time far exceeds the normal range, indicating that the performance of the monitoring system has been seriously affected, which may affect the integrity and reliability of the monitoring data. Such a judgment prompts the operation and maintenance team to conduct further diagnosis and maintenance to find out the specific cause of the problem and make necessary repairs or adjustments to ensure that the monitoring system can resume normal operation, maintain its monitoring effect and safety, and generate a combination anomaly signal.

S2-5, collect standardized anomaly index in multiple consecutive time windows, set a series of time windows , and calculate the corresponding standardized anomaly index for each window ;

S2-6, count the frequency of the standardized anomaly index greater than the anomaly threshold in all observation windows, and mark it as the anomaly frequency;

S2-7, use binomial distribution test to determine whether the frequency of abnormal occurrence is significantly higher than the random level, and judge the consistency and regularity of abnormal occurrence;

The probability of an event occurring in each trial is denoted by ;

(Indicates that the random variable follows a binomial distribution)

in, represents the number of trials, is a random variable;

(It means that under the above binomial distribution conditions, the random variable Value greater than or equal to probability of );

in is the actual number of windows that are greater than the anomaly threshold.

S2-8, linear regression analysis of the change trend of standardized abnormal index over time: ;

in, Represents the slope coefficient, which is used to describe the linear relationship between the independent variable and the dependent variable.

S2-9, if the abnormal frequency is greater than the random level and the slope coefficient is greater than or equal to 0, it means that the abnormal problem of the combination of the camera and the vehicle is continuous and occurs frequently, generating an overall abnormal signal; otherwise, it means that during the monitoring period, the abnormal events at the combination of the camera and the vehicle do not occur frequently, and there is no trend of continuous increase, generating an overall stable signal.

The random level is an expected baseline or benchmark to compare the actual observed data to see if they are statistically significant. This baseline is usually based on random probability in hypothesis testing, that is, the expected results in the natural state without any specific influencing factors (such as systemic problems or external variables).

In statistical analysis, the random level is usually related to the assumed "null hypothesis". For binomial distribution tests (such as tests of abnormal frequency), if the assumption is that the system has no abnormalities, then the occurrence of abnormalities should conform to a theoretical random distribution. The random level here may be a normal abnormality rate estimated based on past data or industry standards.

In actual operation, a threshold can be set based on this baseline random level. For example, if historical data shows that under normal operating conditions, about 5 out of every 100 detections are abnormal (i.e., a 5% random level), then a corresponding threshold (such as 6% or higher) can be set. Only when this threshold is exceeded is the frequency of abnormal occurrence considered significant.

When performing statistical tests, such as calculating -value to determine whether the actual observed abnormal frequency is higher than the assumed random level (such as the 5% mentioned above). -The value is very small (such as less than 0.05), which means that the probability of observing such or more extreme situations when there is no abnormality is very low. Therefore, the null hypothesis can be rejected and the occurrence of the abnormality is considered to be statistically significant.

The present invention evaluates the stability of the combination of the camera and the vehicle by monitoring and analyzing the automatic self-correction events of the camera, thereby ensuring the reliability and effectiveness of the video surveillance system. The process includes calculating the weighted impact value of each self-correction, evaluating the cumulative degree of abnormality within the specified time window, and standardizing the abnormality index for comparison. These data are further compared with the set abnormality threshold to determine the significance of the combined abnormality, and the frequency and trend of the abnormality are determined by statistical tests (such as binomial distribution and linear regression analysis). These steps enable the operation and maintenance team to identify and resolve frequent or severe adjustment needs caused by uneven vehicle operation, unstable camera installation or technical defects. In this way, problems that may cause the loss of key monitoring data or the decline of monitoring quality can be discovered and corrected in a timely manner, ensuring that the camera system is always in the best monitoring state, improving vehicle safety and monitoring effects, and optimizing the maintenance and operation efficiency of the monitoring system.

Step S3 includes the following contents:

S3-1, after confirming that the overall stable signal is obtained, synchronize the camera's video data, the vehicle's GPS location data and weight sensor data in real time, and align the timestamps of the data received from different sources.

S3-2, extracts key features from the synchronized data, including the unloading action in the video, the position change and weight change data recorded by GPS.

S3-2-1, Automatically detect the start and end of unloading behavior from videos using convolutional neural networks;

Use deep learning models, such as convolutional neural networks (CNNs), to process video streams.

The model is trained to recognize specific image features of unloading actions, such as the movement of unloading machinery or the visual patterns of trash being removed from a carriage.

The model was set up to analyze the video stream in real time and automatically mark the frames where unloading begins and ends.

First, a large amount of video data is collected, which includes various scenes of the unloading process, such as different types of garbage, different unloading mechanical movements, and unloading activities under different environmental conditions.

Collect frames with the start and end of unloading in advance, and use these annotations as the "real" labels in the training data. Ensure that each unloading action is accurately labeled, including the starting point of the unloading machine and the moment when the garbage completely leaves the carriage.

The video is pre-processed, including cropping to include only key areas (such as the discharge port), adjusting the resolution and frame rate to reduce processing time and increase the processing speed of the model, and applying image enhancement technology to improve low-light and high-contrast situations.

Choose or design a model architecture based on a temporal convolutional network. The advantage of TCN is that it can effectively handle long-term dependencies of sequence data and is usually easier to train than recurrent neural networks (RNNs).

Configure the key parameters of TCN, such as the number of layers, kernel size, dilation coefficient, etc. These parameters determine the receptive field (i.e. the range of historical data that can be seen) and complexity of the model.

Use GPU for model training to speed up training.

Apply the cross entropy loss function and the Adam optimizer.

Perform multiple rounds of training until the model's performance on the validation set stabilizes or starts to overfit.

The model performance is evaluated on an independent test set. The focus includes the precision, recall and F1 score of unloading action recognition.

Adjust the parameters of TCN according to the test results, such as increasing or decreasing the number of layers, adjusting the expansion coefficient, optimizing the kernel size, etc., to improve the generalization ability of the model.

Verify the adjusted model again to ensure performance improvement.

Deploy the trained model to the actual monitoring system. The model will receive the video stream from the on-board camera in real time.

The model processes the video stream in real time, automatically detecting and marking the start and end of the unloading action.

The system outputs the detection results in real time, for example, displaying the detected unloading events through a visual interface, or triggering related automated processes.

The following is an example of a specific implementation, detailing the entire process:

Model selection:

We chose the Temporal Convolutional Network (TCN), which has several key features suitable for processing video sequence data:

Causal convolution: Ensures that only the current frame and previous frames are used when predicting the current frame, preventing information leakage.

Dilated convolution: Expands the receptive field of the model, allowing the model to learn dependencies over a longer range.

Residual connection: helps train deeper network models and prevents gradient disappearance during training.

Architecture construction:

Number of layers: A 3-layer TCN is set based on the complexity of the data, with each layer containing 128 hidden units.

Kernel size: A kernel of length 3 is chosen, which is suitable for capturing short-term dynamics.

Dilation factor: Set the dilation factor to [1,2,4], increasing with the depth of the layer to cover dynamics at different time scales.

Activation function: Use ReLU activation function to increase nonlinear processing capabilities.

Optimizer and loss function: Adam optimizer and cross entropy loss function are used for training.

Training preparation:

Video Collection: Collect hundreds of hours of unloading videos from actual waste transport vehicles in operation.

Annotation: Professionals will annotate the start and end frames of the unloading action in the video.

Cutting and normalization: Cut the video into small clips containing the complete unloading process and unify the resolution and frame rate of the video.

Model training and tuning:

Training was performed on 70% of the data and 30% was used as a validation set to monitor overfitting and tune model parameters.

After each epoch, the precision and recall of the model are calculated on the validation set to evaluate the performance.

Assume that the initially configured temporal convolutional network (TCN) model parameters are as follows:

Number of layers: 3

Core size: 3

Expansion factor: [1,2,4]

Learning rate: 0.001

Batch size: 32

With this configuration, start training the model and monitor its performance on the validation set. Assume that the main performance metrics of interest are accuracy and loss.

During the training process, the accuracy and loss on the validation set at the end of each epoch are recorded. The following is the simulated data:

Epoch Accuracy(%) loss 1 76.5 0.65 2 78.2 0.60 3 79.0 0.58 4 79.2 0.57 5 79.3 0.57

Based on the preliminary training results, it was observed that the performance improvement of the model began to slow down, probably because the capacity of the model was insufficient to capture more complex temporal dependencies. Therefore, it was decided to make the following adjustments:

Increase the number of layers: from 3 to 4 to increase the learning ability of the model.

Increase the kernel size from 3 to 5 to cover longer time series information.

Adjust the dilation factor: change to [1,2,4,8] to increase the receptive field.

Increase the learning rate from 0.001 to 0.0005 to reduce oscillations in training and improve convergence speed.

Reduce batch size: from 32 to 16 to improve the generalization ability of the model during training.

After adjusting the model parameters, continue to monitor the performance of the model on the validation set. The following is the simulated data after adjustment:

Epoch Accuracy(%) loss 6 80.1 0.55 7 81.4 0.53 8 82.2 0.51 9 83.0 0.49 10 83.8 0.47

The trained model is deployed to the vehicle monitoring system and runs on edge devices to process video data in real time.

The model continuously monitors the video stream and identifies unloading actions in real time.

Detected discharge events are automatically marked and recorded.

Through step S3-2-1, the temporal convolutional network can effectively identify and process the time series information in the video, improving the accuracy and efficiency of automatic detection of unloading actions. This not only improves the automation level of the monitoring system, but also provides timely and accurate data support for the operation and maintenance team, optimizing the operation and monitoring process of the vehicle.

S3-2-2, extracting the time and location of vehicle arrival and departure at the unloading point from GPS data;

Based on the GPS device on the vehicle, real-time recording of location information, including longitude, latitude and timestamp. These data are usually updated in seconds or milliseconds to ensure sufficient accuracy and data continuity.

Preliminary processing of GPS data to remove possible errors or outliers, such as jump points or GPS drift.

Ensure the temporal consistency of GPS data and video surveillance data, and align timestamps to facilitate subsequent cross-data source analysis.

Define geo-fencing of dump points, which is a virtual geographic boundary used to determine when a vehicle enters or leaves a specific area.

Arrival detection: When the vehicle’s GPS coordinates enter the geo-fence for the first time, record that time point and location as “arrival at unloading point”.

Departure detection: When the vehicle’s GPS coordinates last leave the geofence, that time point and location are recorded as the “departure discharge point”.

Analyze the speed changes of the vehicle near the unloading point. Usually the vehicle will slow down to a stop when unloading, which can serve as another verification point to confirm the unloading activity.

Check the duration that the vehicle remains at the dump point to rule out false positive events, such as vehicles simply passing slowly without actually stopping to dump material.

All identified arrival and departure events are stored in a database along with corresponding time and location stamps.

Through step 3-2-2, it is ensured that the key action time and position of the vehicle at the unloading point are accurately extracted from the GPS data, providing important information support for transportation and logistics operations and enhancing the functionality and practicality of the entire monitoring system.

S3-2-3, based on the weight sensor data, records the weight change before and after unloading.

High-precision weight sensors are installed at key locations on garbage transport vehicles, such as the bottom of the vehicle compartment. These sensors can continuously measure and record the vehicle's load weight.

Ensure that the data collection of the weight sensor is synchronized with the vehicle's GPS and video monitoring system to facilitate cross-system data integration and analysis.

The collected weight data is pre-processed to remove any obvious errors or abnormal readings, such as sudden jumps caused by equipment failure.

Weight record before unloading: Before the GPS data shows that the vehicle arrives at the unloading point, the weight at that moment is recorded as the weight before unloading.

Weight record after unloading: After the GPS data shows that the vehicle has left the unloading point, the weight at that moment is recorded as the weight after unloading.

By calculating the weight difference before and after unloading, the weight difference before and after unloading is obtained. This difference represents the net weight of the material during the unloading process.

The weight change data is checked for consistency to ensure that it matches the unloading action recorded in the video surveillance and the location information of the GPS data. The specific process is as follows:

First, a timestamp alignment check is performed to ensure that the key timestamps of the weight change are aligned with the start and end timestamps of the unloading action in the video and the location change timestamps in the GPS data within the allowable error range (for example, ±2 minutes). Next, a location consistency check is performed to verify whether the geographical location of the vehicle arriving and leaving the unloading point recorded in the GPS data matches the preset unloading point geo-fence. The location error should be controlled within the range allowed by GPS accuracy, such as within 10 meters. Then compare the weight change data with the unloading volume visible in the video surveillance, evaluate the visual volume or estimated amount of garbage during the unloading process in the video, and compare this with the weight change data. If the weight change is roughly consistent with the visual estimate of the unloading (for example, the error of the change is within 5%), it is considered consistent; if the difference is large, further investigation is required for possible recording errors or equipment failures. Finally, an automated script is implemented to analyze data consistency regularly or in real time, and automatically generate an exception report when any inconsistency is detected, notifying the operations team for further inspection, so as to ensure the consistency and accuracy of the data and effectively support the monitoring and management of the unloading process.

Through step S3-2-3, the consistency check of weight change data is ensured to be both accurate and comprehensive, which effectively supports the monitoring and management of the unloading process and improves the credibility of data and the transparency of operations. S3-3, a standard model of unloading behavior is established based on historical data.

S3-3-1, collect historical unloading behavior data to determine the average unloading time, common locations and weight variation range;

Collect data on historical unloading events, including GPS records from onboard systems, video surveillance data, and weight sensor readings.

Ensure that the collected data covers a sufficient time range and a variety of different conditions (such as different seasons, different time periods, different drivers, and different routes) to improve the representativeness and accuracy of the data.

Clean the data to remove any obvious errors or anomalies, such as data from GPS drift or sensor failure.

Format the data to ensure that all data points have a unified timestamp format and geographic coordinate system for easy subsequent processing and analysis.

Average duration calculation: Analyze the duration of all historical unloading events and calculate the average unloading duration, which involves calculating the time interval from the start to the end of each unloading event.

Identification of common unloading locations: Use GPS data to analyze where unloading occurs, and identify common unloading locations through statistical methods such as cluster analysis.

Weight variation range determination: Analyze the weight sensor data before and after discharge, calculate the average weight variation range and its standard deviation to determine the typical discharge weight variation.

This statistical analysis is updated regularly to include new data and reflect any changes in trends.

By executing step S3-3-1, not only can the typical characteristics of unloading activities be accurately understood, but daily operations can also be optimized based on data-driven insights to improve overall transportation efficiency and cost-effectiveness. The implementation of this approach will help ensure the efficiency and sustainability of waste transportation services while enhancing control and supervision capabilities over key operational links.

S3-3-2, use clustering to classify the patterns of unloading behavior and use the identified patterns as historical standard patterns.

Gather historical data containing various unloading-related characteristics, such as the time, location, duration, weight change of unloading, and relevant environmental factors (weather, time of day, etc.).

Clean the data, handle missing values and outliers, and ensure data quality. Perform necessary data transformations, such as converting timestamps to time of day, to meet the needs of cluster analysis.

Select the features that have a significant impact on the discharge behavior, such as the geographical location of the discharge point, the duration of the discharge, and the weight variation of the discharged amount.

Create new features (such as discharge rate, which is the ratio of discharge volume to time) and bin continuous variables to improve the effect of cluster analysis.

The key features that have an impact on clustering are selected, such as the location of the discharge point (latitude and longitude), discharge duration, and discharge amount (weight change).

Convert time to time period of the day, place to relative distance, etc.

Use the elbow method to determine the optimal number of clusters (k value): run the K-means algorithm multiple times, each time using a different k value, then evaluate the compactness and separation of the results, and choose the point where the sum of squared errors suddenly decreases as the optimal k value.

Initialize the cluster center, randomly select k data points as the initial center, and then iteratively update the center point until the clustering no longer changes significantly or reaches the preset number of iterations.

Each cluster may represent a discharge pattern, and the identified pattern is used as a historical standard pattern, for example:

Cluster 1: Rapid, small-volume unloading, which may occur at urban fast-consumption points.

Cluster 2: Slow, high-volume discharges, which may occur in industrial areas or processing centers.

According to step S3-3-2, using cluster analysis to classify patterns of unloading behaviors not only helps the operations team better understand different unloading behaviors, but also optimizes unloading operations based on these insights, improving the efficiency and adaptability of transportation and logistics activities.

S3-4, apply real-time data analysis algorithms to compare current behavior with standard patterns and mark all events exceeding corresponding thresholds as abnormal;

Receive data from vehicle sensors in real time, including GPS data, weight sensor data, and possible video surveillance data.

Use real-time data to update the system, capturing data every time a discharge event begins and ends.

Compare discharge duration, location and weight changes captured in real time to historical standard patterns.

Calculate deviations between real-time data and historical patterns, for example, comparing real-time discharge duration to historical average duration.

Set deviation thresholds, such as a duration deviation of more than 10%, a location deviation from a common unloading point of more than a certain distance (e.g. 50 meters), or a weight change that differs from expectations by more than 5%.

If the unloading behavior in the real-time data deviates from the historical pattern beyond the set threshold on any key indicator, it is automatically marked as abnormal behavior.

S3-5, record and respond to identified abnormal behaviors.

Automatic alerts are triggered to notify the operations team for further inspection and handling of possible abnormal behavior.

All abnormal events are recorded in the background for future analysis and improvement.

The present invention realizes real-time monitoring and analysis of unloading behavior by comprehensively utilizing video surveillance data, GPS data, and weight sensor data from vehicles, ensuring that each unloading behavior is compared with historical standard patterns to identify and mark behaviors that deviate from the norm. In this process, the start and end of the unloading action are automatically detected, and key indicators such as unloading duration, frequency, and location are compared in real time. Abnormal behaviors are identified through set deviation thresholds, such as unloading time that is too long or too short, and the location deviates from common unloading points. This real-time data analysis and monitoring can effectively prevent garbage trucks from unloading at non-designated locations and enhance the compliance of the transportation process. In addition, the automatic recording and alarm of abnormal behaviors can promptly notify the operation team to intervene, ensuring operational safety and efficiency, thereby greatly improving the response speed and accuracy of operation management.

Step S4 includes the following contents:

Report Generation:

Collect all data related to the identified abnormal behavior, including timestamps, locations, video footage, GPS data, weight change data, etc.

The report includes a detailed description of abnormal events, such as abnormal unloading time, position deviation, inconsistent weight changes, etc. At the same time, relevant data evidence, such as video screenshots, data charts, etc., is attached.

Archive and access:

Automatically save the generated reports to a database or dedicated archiving system to ensure data security and long-term accessibility.

Archived reports are indexed to make them easy to retrieve. Keyword search, time range search and other functions are provided to facilitate quick finding of reports for specific events.

Example 2

FIG2 shows a garbage transport vehicle unloading supervision system based on video monitoring big data of the present invention, including: a video acquisition optimization module, a stability trend monitoring module, a data synchronization integration module, an abnormal behavior analysis module and a report automatic generation module;

The video acquisition optimization module uses image processing technology to identify and mark the key operating areas of the garbage truck, installs and adjusts the camera to dynamically cover these areas, and transmits the video data to the central server in real time;

The stability trend monitoring module defines and calculates the weighted impact value of each camera return on the central server, calculates the cumulative impact of all events in the specified time window, and analyzes the long-term trend of these events through binomial distribution and linear regression, monitors the stability of the combination of camera and vehicle, and sends the monitoring results to the data synchronization integration module;

The data synchronization and integration module calls the data from the central server, synchronizes and integrates the GPS data, weight sensor data and video surveillance data from the vehicle, aligns the timestamps of all data points, and transmits the processed data to the abnormal behavior analysis module;

The abnormal behavior analysis module uses data analysis technology to monitor and analyze the current unloading behavior pattern in real time, and compares the real-time data with the stored historical standard pattern to identify and mark potential abnormal behaviors, and sends the identified abnormal behaviors to the automatic report generation module;

The automatic report generation module generates reports for any abnormal behavior identified and automatically archives the reports.

The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or may be transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium may be a solid-state hard disk.

Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

In the several embodiments provided in the present application, it should be understood that the disclosed systems and devices can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program codes.

The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims (8)

1. A method for monitoring the unloading of garbage transport vehicles based on video surveillance big data, characterized in that: Step S1, using image processing technology to identify and mark key operating areas of the garbage truck, dynamically covering these areas with cameras at designated locations on the vehicle, and transmitting video data to a central server in real time through real-time adjustment of the pitch and yaw angles of the cameras in conjunction with the network; Step S2: On the central server, a weighted impact value of each camera return is defined, the cumulative impact of all events within the specified time window is calculated, and the long-term trend of the events is analyzed by binomial distribution and linear regression to monitor the stability of the combination of camera and vehicle; Step S3, after confirming that the camera and the vehicle are integrated without any obstacles, integrating and synchronizing the vehicle's GPS data, weight sensor data, and video surveillance data, and applying data analysis technology to identify and mark potential abnormal behaviors; Step S4: For any abnormal behavior identified, a report is automatically generated and archived. 2. The method for monitoring the unloading of garbage transport vehicles based on video surveillance big data according to claim 1 is characterized by: Step S1 includes the following contents: S1-1, marking key operating areas on garbage trucks and using image processing technology to identify specific area features; S1-2, install the motor-driven camera to a position that can dynamically cover all key operating areas of the vehicle; Use the geometric model to determine the initial camera mounting positions so that the initial pitch and yaw angle settings cover all critical operating areas of the vehicle. S1-3, real-time dynamic adjustment of the camera to respond and maximize coverage of key areas; S1-4, configure the network to transmit the camera video to the central server in real time. 3. The method for monitoring the unloading of garbage transport vehicles based on video surveillance big data according to claim 2 is characterized in that: Step S2 includes the following contents: S2-1, based on the amplitude of the correction and the time interval from the last correction, a weighted impact value is defined for each correction; No. The comprehensive amplitude of the return to positive: ; The time interval between two corrections: ; The weighted impact of each reversal: ; in is the time decay coefficient, which is used to adjust the effect of time interval; S2-2, calculate the cumulative impact of all return events within a given time window to assess the degree of abnormality; Defining time windows ; Cumulative impact of all return times within the time window: ; S2-3, normalize the abnormal index; S2-4, using the standardized abnormality index to compare with the abnormality threshold to determine whether there is a significant binding abnormality; If the normalized abnormality index is greater than the abnormality threshold, it is determined that there is a significant binding abnormality, and a binding abnormality signal is generated. 4. The method for monitoring the unloading of garbage transport vehicles based on video surveillance big data according to claim 3 is characterized in that: S2-5, collecting standardized anomaly indices in multiple consecutive time windows, setting a series of time windows, and calculating the corresponding standardized anomaly index for each window; S2-6, count the frequency of the standardized anomaly index greater than the anomaly threshold in all observation windows, and mark it as the anomaly frequency; S2-7, use binomial distribution test to determine whether the frequency of abnormal occurrence is significantly higher than the random level, and judge the consistency and regularity of abnormal occurrence; S2-8, linear regression analysis of the change trend of standardized abnormal index over time: ; in, It represents the slope coefficient, which is used to describe the linear relationship between the independent variable and the dependent variable; S2-9, if the abnormal frequency is greater than the random level and the slope coefficient is greater than or equal to 0, an overall abnormal signal is generated; otherwise, an overall stable signal is generated. 5. The method for monitoring the unloading of garbage transport vehicles based on video surveillance big data according to claim 4 is characterized in that: Step S3 includes the following contents: S3-1, after confirming that the overall stable signal is obtained, synchronize the video data of the camera, the GPS location data of the vehicle and the weight sensor data in real time, and align the timestamps of the data received from different sources; S3-2, extract key features from the synchronized data, including the unloading action in the video, the position change and weight change data recorded by GPS; S3-3, establish a standard model of unloading behavior based on historical data; S3-4, apply real-time data analysis algorithms to compare current behavior with standard patterns and mark all events exceeding corresponding thresholds as abnormal; S3-5, record and respond to identified abnormal behaviors. 6. The method for monitoring the unloading of garbage transport vehicles based on video surveillance big data according to claim 5 is characterized in that: S3-2-1, Automatically detect the start and end of unloading behavior from videos using convolutional neural networks; S3-2-2, extracting the time and location of vehicle arrival and departure at the unloading point from GPS data; S3-2-3, based on the weight sensor data, records the weight change before and after unloading. 7. The method for monitoring the unloading of garbage transport vehicles based on video surveillance big data according to claim 5 is characterized in that: S3-3-1, collect historical unloading behavior data to determine the average unloading time, common locations and weight variation range; S3-3-2, use clustering to classify the patterns of unloading behavior and use the identified patterns as historical standard patterns. 8. A garbage transport vehicle unloading supervision system based on video surveillance big data, used to implement the supervision method described in any one of claims 1 to 7, comprising a video acquisition optimization module, a stability trend monitoring module, a data synchronization integration module, an abnormal behavior analysis module and an automatic report generation module; The video acquisition optimization module uses image processing technology to identify and mark the key operating areas of the garbage truck, installs and adjusts the camera to dynamically cover these areas, and transmits the video data to the central server in real time; The stability trend monitoring module defines and calculates the weighted impact value of each camera return on the central server, calculates the cumulative impact of all events in the specified time window, and analyzes the long-term trend of these events through binomial distribution and linear regression, monitors the stability of the combination of camera and vehicle, and sends the monitoring results to the data synchronization integration module; The data synchronization and integration module calls the data from the central server, synchronizes and integrates the GPS data, weight sensor data and video surveillance data from the vehicle, aligns the timestamps of all data points, and transmits the processed data to the abnormal behavior analysis module; The abnormal behavior analysis module uses data analysis technology to monitor and analyze the current unloading behavior pattern in real time, and compares the real-time data with the stored historical standard pattern to identify and mark potential abnormal behaviors, and sends the identified abnormal behaviors to the automatic report generation module; The automatic report generation module generates reports for any abnormal behavior identified and automatically archives the reports.