CN105005788A - Target perception method based on emulation of human low level vision - Google Patents

Abstract

The present invention discloses a target perception method based on emulation of human low level vision. The method comprises the following steps of: 1) performing saliency detection on a target image by using a frequency domain method, to obtain a corresponding pixel saliency map; 2) sorting saliency points in the pixel saliency map according to the saliency degree; 3) selecting the first N saliency points as watch points, wherein a minimum rectangular scope that comprises the watch points serves as a watch area; 4) performing random sampling on pixels within the watch area, and performing random pixel sampling of a same quantity outside watch area; and 5) obtaining a two-category SVM model through training by using a support vector machine training strategy; and classifying all pixels of the target image by using the model, and using a pixel area classified as a positive sample as a first watch target area. According to the present invention, human vision is emulated according to the human visual fixation process and by sorting watch points and using a neural network model, thereby realizing quick and effective watch and perception towards a target scene.

Description

Object perception method for simulating human low-level vision

technical field

The invention relates to the technical field of human vision simulation, in particular to a target perception method for simulating human low-level vision.

Background technique

With the development of information technology, computer vision has been widely used in low-level feature detection and description, pattern recognition, artificial intelligence reasoning and machine learning algorithms and other fields. However, computer vision is a task-driven method, that is, it needs to limit many conditions and design corresponding algorithms according to actual tasks. It lacks general algorithms, so it often encounters high-dimensional nonlinear feature spaces and large amounts of data to solve problems. And real-time processing and other issues make its research and application face great challenges.

For the human visual system to work efficiently and reliably in different environments, it has the following advantages: having attention mechanism, saliency detection and related selectivity and purpose in visual processing; being able to learn from low-level visual processing The use of prior knowledge in visual processing enables data-driven bottom-up processing and top-down knowledge guidance to coordinate with each other in visual processing; contextual information plays an important role in all levels of visual processing, and can comprehensively utilize the environment Information in various modalities. However, when the mechanism of human visual perception is not yet fully understood, there are still great difficulties in how to construct machine vision with human visual characteristics. and other applications have a significant impact.

Contents of the invention

In view of this, the technical problem to be solved by the present invention is to provide a target perception method for simulating human low-level vision that can simulate human vision and realize fast and effective fixation on target scenes.

Technical solution of the present invention is, the object perception method of the simulation human low-level vision of following steps is provided, comprises the following steps:

1) performing saliency detection on the target image by a frequency domain method to obtain a corresponding pixel saliency map, the pixel saliency map is consistent with the pixel position information of the target image;

2) sorting the salient points in the pixel saliency map according to the saliency;

3) Select the first N salient points as fixation points, and the minimum rectangular range including these fixation points as the fixation area;

4) random sampling is carried out to the internal pixels of the fixation area, and random sampling is carried out to an equal amount of pixels outside the fixation area; the internal pixels of the fixation area obtained by sampling are used as positive samples, and the external pixels of the fixation area are used as negative samples;

5) Utilize support vector machine training strategy, train the SVM model that obtains a binary classification, classify all pixels of described target image by this model, will be divided into the pixel area of positive sample as the first gaze target area.

Using the method of the present invention, compared with the prior art, the present invention has the following advantages: the saliency detection by the frequency domain method can quickly form a pixel saliency map, which is consistent with the pixel position information of the target image, and according to the saliency Sorting the minimum rectangular range formed by the selected fixation points as the fixation area for sampling, entering the neural network together with external samples, taking the area with high salience as the first fixation target area, and the first fixation target area can be established On the basis of , expand the fixation range again to form a corresponding fixation target area, and compare it with the first fixation target area to judge whether the result of the first fixation target area is stable. According to the process of human visual gazing, the present invention simulates human vision through gazing point sorting and a neural network model, thereby realizing fast and effective gazing and perception of target scenes.

As an improvement, select the first N+M salient points as fixation points, form a fixation area according to step 3), and then obtain the corresponding second fixation target area through steps 4) and 5); compare the first fixation target area and the second fixation target area The degree of overlap of the target area, a large degree of overlap indicates that the visual perception intensity of the target is large; a small degree of overlap indicates that the visual perception intensity of the target has not been formed, and the above process is continued until sufficient visual perception intensity is achieved. The fixation target area of is the superposition of all fixation target areas in the above process. This design can speed up the generation and output of visual perception targets, and obtain a more stable fixation target area, and the fixation result is more reliable.

As an improvement, after the fixation target area is obtained, this area is cleared in the target image and the pixel saliency map, and the salient points in the updated pixel saliency map are reordered according to the saliency, and steps 3) and 4) are repeated and 5), to obtain a new fixation target area, and sequentially obtain multiple target areas in the image. In this way, the gaze recognition and reading of the effective information of the entire image can be completed, and the accuracy and completeness of gaze can be improved.

As an improvement, the frequency-domain method refers to using the hypercomplex Fourier transform to take the three components of red, green and blue in the color image as the three imaginary parts of hypercomplexity to participate in the Fourier transform, only retaining the phase spectrum information, and performing the Fourier transform The inverse transformation obtains a pixel saliency map. The design is used to solve the problem that the prior art can only deal with black and white image recognition, and effectively improves the specific steps of the frequency domain method for color images.

Description of drawings

FIG. 1 is a flow chart of the object perception method for simulating human low-level vision in the present invention.

Detailed ways

The present invention will be further described below with regard to specific examples, but the present invention is not limited only to these examples.

The present invention covers any alternatives, modifications, equivalent methods and schemes made on the spirit and scope of the present invention. In order to provide the public with a thorough understanding of the present invention, specific details are set forth in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, for the sake of illustration, the drawings of the present invention are not completely drawn according to the actual scale, and are described here.

As shown in Figure 1, the object perception method of the simulation human low-level vision of the present invention comprises the following steps:

1) performing saliency detection on the target image by a frequency domain method to obtain a corresponding pixel saliency map, the pixel saliency map is consistent with the pixel position information of the target image;

2) sorting the salient points in the pixel saliency map according to the saliency;

3) Select the first N salient points as fixation points, and the smallest rectangular range containing these fixation points as the fixation area; selecting the smallest rectangular range can not only ensure the accuracy of sampling, but also improve the accuracy and stability of the fixation target area;

4) random sampling is carried out to the internal pixels of the fixation area, and random sampling is carried out to an equal amount of pixels outside the fixation area; the internal pixels of the fixation area obtained by sampling are used as positive samples, and the external pixels of the fixation area are used as negative samples;

5) Utilize support vector machine training strategy, train the SVM model that obtains a binary classification, classify all pixels of described target image by this model, will be divided into the pixel area of positive sample as the first gaze target area.

For the perception of simulating human vision, the image is equivalent to the scene that human vision is looking at. Regardless of the size of the scene, the range of imaging on the retina remains unchanged. Therefore, the same is true of images for machines and machine vision.

The saliency detection of the target image by the frequency domain method can be implemented by the following steps: perform two-dimensional discrete Fourier transform F[I(i, j)] on the target image I(i, j), transform the image from the spatial domain Convert to the frequency domain to get the phase P(u, v) information:

In the formula, F represents the two-dimensional discrete Fourier transform, Indicates phase operation. After the phase information is inversely transformed by Fourier, the saliency image Sa_Map can be obtained in the space domain.

Sa_Map(i, j)=|F ^-1 [exp{jP(u, v)}]| ² (2)

In Figure 1, training data, classification models, results, etc. are all implemented using support vector machine (SVM) training strategies. The specific implementation process is as follows:

Suppose the training set contains l samples is the input vector, and y _k ∈ {-1, +1} is the positive and negative category identification. SVM first uses the training set to learn modeling, the purpose is to find the optimal classification hyperplane in the feature space, and classify the test data as correctly as possible. Considering the general situation, when the training set is nonlinearly separable, first choose a Gaussian radial basis kernel function

K(x, x _i )＝exp{-q||xx _i || ² } (3)

Map the training set data _xi to a high-dimensional linear feature space to construct the optimal classification hyperplane. Where q is the parameter of the radial basis kernel function, and the discriminant function of the classifier is

$\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; sign</span>}<span\; class="notranslate">\; [</span>\underset{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; \&Element;</span>\mathrm{<span\; class="notranslate">\; Si</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}{{\mathrm{<span\; class="notranslate">\; \&alpha;</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}^{<span\; class="notranslate">\; *</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\mathrm{<span\; class="notranslate">\; K</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; b</span>}}^{<span\; class="notranslate">\; *</span>}<span\; class="notranslate">\; ]</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

The training process is known and q, etc., use the quadratic programming method to obtain the b ^* , α _i ^* and support vector (SV) in the formula (4) as the SVM model obtained by training; the test process is to use the SVM model to convert the unknown Data x is substituted into formula (4) to obtain its predicted category.

SVM uses the technique of kernel function to avoid the curse of dimensionality problem faced by traditional learning algorithms. Based on the principle of structural risk minimization, its classification performance is only determined by a small number of support vectors (SV), and it has good generalization performance. In practical problems, it is beneficial to use prior knowledge to select a small number of samples and construct a classifier through SVM learning. It overcomes the defect that the traditional learning algorithm is based on the principle of empirical risk minimization, and the performance can only be guaranteed theoretically when the number of samples tends to infinity; by solving the quadratic programming problem, it can avoid the empirical and easy construction of the traditional neural network algorithm Falling into local minimum solutions and other shortcomings; it is suitable for image targets that are complex to segment and difficult to describe quantitatively.

In order to optimize the present invention, it is necessary to judge whether the obtained first fixation target area is stable, which is reflected in the block diagram to judge whether there is a stable output. Therefore further target areas need to be formed:

Select the first N+M salient points as fixation points, form a fixation area according to step 3), and then obtain the corresponding second fixation target area through steps 4) and 5); compare the first fixation target area and the second fixation target area The degree of overlap, a large degree of overlap indicates that the visual perception intensity of the target is large; a small degree of overlap indicates that the visual perception intensity of the target has not been formed, and the above process continues to be repeated until sufficient visual perception intensity is achieved, and the final gaze target Region is the superposition of all fixation target regions in the above process.

After the fixation target area is obtained, the area is cleared in the target image and the pixel saliency map, and the salient points in the updated pixel saliency map are sorted again according to the saliency, and steps 3), 4) and 5) are repeated , to obtain a new fixation target area, and to sequentially obtain multiple target areas in the image. In this way, the information of all effective gaze areas can be segmented from the image, and a machine vision that simulates human vision can be constructed.

The frequency domain method refers to the use of the hypercomplex Fourier transform to use the red, green and blue components of the color image as the three imaginary parts of the hypercomplex to participate in the Fourier transform, only retaining the phase spectrum information, which is obtained by inverse Fourier transform Pixel saliency map. The design is used to solve the problem that the prior art can only deal with black and white image recognition, and effectively improves the specific steps of the frequency domain method for color images. A hypercomplex number consists of four parts, denoted as

q＝a+bi+cj+dk (5)

Where a, b, c, d are all real numbers, i, j, k are all imaginary units, and have the following properties: i ² =j ² =k ² =ijk=-1, ij=-ji=k, ki= -ik=j, jk=-kj=i.

The RGB model of a color image can be described as a pure hypercomplex number with no real part:

f＝R(m,n)i+G(m,n)j+B(m,n)k (6)

Among them, R(m, n), G(m, n), and B(m, n) represent the three components of image red, green and blue, respectively, and m and n are pixel coordinates. If q=f, then a=0, b=R(m,n), c=G(m,n), d=B(m,n). The hypercomplex Fourier transform can be performed on the color vector according to formula (6):

F ^R (v,u)=(real(fft2(a))+μ·imag(fft2(a)))+

i(real(fft2(b))+μ imag(fft2(b)))+ (7)

j(real(fft2(c))+μ·imag(fft2(c)))+

k(real(fft(d))+μ·imag(fft2(d)))

Among them, fft2() represents the traditional two-dimensional Fourier transform, real() represents the real part, and imag() represents the imaginary part.

is a unit virtual vector. Here, just take the phase spectrum P(f) of FR ⁽ v, u):

Order: A＝e ^jP(f) (9)

Using the traditional two-dimensional inverse fast Fourier transform (ifft2) combination can obtain the hypercomplex inverse Fourier transform, as shown in formula (10):

F ^-R (v,u)=(real(ifft2(A))+μ·imag(ifft2(A)))+

i(real(ifft2(B))+μ imag(ifft2(B)))+ (10)

j(real(ifft2(C))+μ·imag(ifft2(C)))+

k(real(ifft2(D))+μ·imag(ifft2(D)))

Wherein, B=fft2(b), C=fft2(c), D=fft2(d).

Sa_Map(m,n)=real(F- ^R (v,u)) (11)

is the obtained saliency map. Since the integrity of color pixels before and after data processing is maintained, color distortion caused by transformation or exchange of vector components is avoided.

The above is only an illustration of the preferred embodiments of the present invention, but should not be construed as a limitation on the claims. The present invention is not limited to the above embodiments, and its specific structure is allowed to vary. In a word, all kinds of changes made within the protection scope of the independent claims of the present invention are within the protection scope of the present invention.

Claims (4)

1. A target perception method for simulating human low-level vision, characterized in that: comprise the following steps: 1) performing saliency detection on the target image by a frequency domain method to obtain a corresponding pixel saliency map, the pixel saliency map is consistent with the pixel position information of the target image; 2) sorting the salient points in the pixel saliency map according to the saliency; 3) Select the first N salient points as fixation points, and the minimum rectangular range including these fixation points as the fixation area; 4) random sampling is carried out to the internal pixels of the fixation area, and random sampling is carried out to the pixels of the same amount outside the fixation area; the internal pixels of the fixation area obtained by sampling are used as positive samples, and the external pixels of the fixation area are used as negative samples; 5) Utilize support vector machine training strategy, train the SVM model that obtains a binary classification, classify all pixels of described target image by this model, will be divided into the pixel area of positive sample as the first gaze target area. 2. the object perception method of simulating human's low-level vision according to claim 1, is characterized in that: select first N+M salient points as fixation point, form fixation area according to step 3), then through steps 4) and 5) Obtain the corresponding second fixation target area; Compare the degree of overlap between the first fixation target area and the second fixation target area. A larger degree of overlap indicates a greater intensity of visual perception of the target; a smaller degree of overlap indicates that the intensity of visual perception of the target has not yet been formed. Continue to repeat the above process , until sufficient visual perception intensity is achieved, the final fixation target area is the superposition of all fixation target areas in the above process. 3. the target perception method of simulating human's low-level vision according to claim 1, is characterized in that: after obtaining fixation target zone, this zone is cleared in target image and pixel saliency figure, to the updated pixel saliency The salient points in the figure are reordered according to the degree of salience, and steps 3), 4) and 5) are repeated to obtain a new fixation target area; thus multiple target areas in the image can be sequentially obtained. 4. according to claim 1 and 2 described target perception methods of simulating human's low-level vision, it is characterized in that: described frequency domain method refers to by super-complex Fourier transform, red, green, blue three in color image The components participate in Fourier transform as three imaginary parts of hypercomplex numbers, and only the phase spectrum information is retained, and the pixel saliency map is obtained through inverse Fourier transform.