CN103760168A - Smoothness-characteristic-quantity-based surface gray scale defect detection method for rotary-cut wooden product - Google Patents

Abstract

The invention relates to a smoothness-characteristic-quantity-based surface gray scale defect detection method for a rotary-cut wooden product. The method comprises the following steps: acquiring an image of the rotary-cut wooden product by using CCD (Charge Coupled Device) cameras placed at the middle upper positions of two LED (Light-Emitting Diode) strip light sources which are arranged symmetrically and adopt high-angle illuminating ways serving as illumination light sources, converting the acquired image into a digital signal through an image acquiring card, and transmitting the digital signal to a computer system; performing target area determination, smoothness characteristic quantity calculation and gray scale defect judgment on the input image, wherein the judgment threshold value of each gray scale defect is determined by using a self-adaptive maximum interclass variance method. By adopting the detection method, automatic online detection on the surface gray scale defects of the rotary-cut wooden product is realized, the production cost is reduced, the detection efficiency is increased, the detection rate of the gray scale defects is increased greatly, and the phenomena of detection missing and error detection are well overcome.

Description

Rotary-cut class wood product surface gray scale defect inspection method based on smoothness characteristic quantity

Technical field

The present invention relates to machine vision, product quality detection field, refer in particular to a kind of rotary-cut class wood product surface gray scale defect inspection method based on smoothness characteristic quantity.

Background technology

Due to natural timber raw material and production technology aspect, cause its surface of disposable rotary-cut class woodwork of producing to contain polytype gray scale defect, for example, knot, variable color, rotten, channel, rascal and pollution etc.; And, this type of rotary-cut class woodwork that includes gray scale defect can not continue on for subsequent production and use again, therefore needs rotary-cut class wood product surface gray scale defect to detect, and rejects unacceptable product, or suitably adjust processing parameter, to improve yield rate and the product quality of woodwork.

The main artificial visual detection method that relies on detects the surperficial gray scale defect of rotary-cut class woodwork at present, and the human resources that this traditional visual detection method expends are huge, to relevant enterprise, brought huge economic loss; Next, it is larger that testing result and detection quality are affected by subjective factor; In addition, detection efficiency is very low, can not meet the automated production demand of modern industry.

Along with theoretical growing with Computer Applied Technology of machine vision, vision detection technology is being deep into industrial every field gradually, for example now be widely used in (patent of invention, notification number: CN102095731A in the surface defects detection of each type objects such as paper, Copper Foil, bar cigarette external packing tracing paper, steel ball, Plastic medicinal bottle or product; Utility model patent, notification number: CN202928963U; Patent of invention, notification number: CN102020036A; Utility model patent, notification number: CN201555812U; Patent of invention, notification number: CN102053091A).

Body surface gray scale defects detection algorithm based on computer vision technique is mainly realized by the color characteristic that extracts image, wherein common with histogram feature; And the histogram feature amount of often using comprises, average, variance, skewness, kurtosis, energy and entropy etc. (computer knowledge and technology, 2009,5 (32): 9032-9034).

Nowadays, utilize computer vision technique to carry out the relevant report of rotary-cut class wood product surface gray scale defects detection little, and concrete application technology is still immature; Meanwhile, because rotary-cut class wood product surface gray scale defect type is numerous, and directly apply the above-mentioned characteristic quantities such as variance, skewness, can only pick out wherein a part of gray scale defect, thereby easily cause the phenomenons such as undetected and false retrieval.

Summary of the invention

In order to overcome, the existing detection efficiency of existing rotary-cut class wood product surface gray scale defect inspection method is low, automaticity is low, can not use noncontact visible detection method, and existing based on problems such as existing undetected, the false retrievals of the statistical characteristic values such as average, variance, skewness, the invention provides a kind of rotary-cut class wood product surface gray scale defect inspection method based on smoothness characteristic quantity.

Technical solution of the present invention is: the rotary-cut class wood product surface gray scale defect inspection method based on smoothness characteristic quantity, comprise two LED strip sources that are arranged symmetrically with and adopt high angle lighting system, be placed in the ccd video camera of the position, upper middle of two light sources, image pick-up card and computer system; Described ccd video camera gathers rotary-cut class woodwork image, by described image pick-up card, be translated into digital picture and be input in described computing machine, then utilizing following three steps it to be carried out to image processing and judge whether this rotary-cut class woodwork exists gray scale defect:

Step 1, image target area is determined, is utilized Threshold sementation to extract the target area of rotary-cut class woodwork;

Step 2, smoothness characteristic quantity calculates.

Calculate the normalization histogram p (h) of target area, the wherein gray level of each pixel in h presentation video, for 8 bit images, the scope of gray level h is 0 to 255, p to represent the probability that gray level occurs; Utilize formula (1) and (2) to calculate average m and variances sigma ², further according to formula (3), obtain smoothness characteristic quantity R:

$m=\underset{i=0}{\overset{255}{\mathrm{\Σ}}}{\mathrm{ip}}_i---\left(1\right)$

${\mathrm{\σ}}^2=\underset{i=0}{\overset{255}{\mathrm{\Σ}}}{(i-m)}^2{p}_i---\left(2\right)$

$R=\frac{1}{1+{\mathrm{\σ}}^2}---\left(3\right)$

Step 3, the judgement of gray scale defect.

When smoothness characteristic quantity R is more than or equal to discrimination threshold T, represent rotary-cut class wood product surface normal, without gray scale defect; On the contrary, when R is less than discrimination threshold T, represent that rotary-cut class wood product surface contains gray scale defect.

In described detection method, the discrimination threshold T in described step 3 utilizes self-adaptation maximum variance between clusters to determine, concrete steps are:

1) choose the rotary-cut class woodwork image that n width surface is normal and all kinds of gray scale defects are contained on n width surface;

2) respectively to each width in 2n width image, according to the above step 1, carry out target area and determine, according to the above step 2, calculate the smoothness characteristic quantity R of target area;

3) by ascending sequence of smoothness characteristic quantity R of 2n width image; Within the scope of minimum and maximum R, successively etc. step-length is using each data as default classification thresholds T, and the corresponding image of R value that is more than or equal to T is divided into the 1st class, and the corresponding image of R value that is less than T is divided into the 2nd class; If the picture number that each class comprises is w _i, wherein i=1 or 2, in each class, the mean value of R value is M _i, according to formula (4), calculate the corresponding inter-class variance of each default classification thresholds T

σ ²：

${\mathrm{\σ}}^2=\frac{w_1{w}_2{(M_1-M_2)}^2}{{(w_1+w_2)}^2}---\left(4\right)$

4) find out the maximal value in all inter-class variance data, the corresponding default classification thresholds of this maximal value is the discrimination threshold T using in testing process.

The present invention compared with prior art has following technical advantage:

1, the present invention adopts two symmetrical blue led strip sources of installing and applies high angle lighting system, can at utmost improve the homogeneity of irradiating light intensity on bar shaped rotary-cut class wood product surface, thus the actual reflectance component of reaction wood product surface farthest; Meanwhile, blue illumination can make on rotary-cut class wood product surface to be difficult for to be enhanced by the rascal of human eye detection and the natural gray scale defect such as rotten, has improved the contrast of image, thereby has improved the recall rate of rotary-cut class wood product surface gray scale defect.

2, compared with Traditional Man visual detection method, the present invention is based on vision detection technology and realized robotization detection, improve detection efficiency, reduced production cost, improved economic benefit and the social benefit of relevant enterprise.

3, with respect to other statistical characteristic value as the detection method of average, variance, skewness, kurtosis, energy and entropy, the present invention adopts smoothness characteristic quantity to carry out gray scale defect estimation, to imaging surface local gray level, sudden change has higher sensitivity, dissimilar gray scale defect is had to the stronger ability of birdsing of the same feather flock together, thereby greatly improved the recall rate of the rotary-cut class woodwork that contains gray scale defect, overcome well undetected and false retrieval phenomenon.

4, in the present invention, discrimination threshold adopts self-adaptation maximum variance between clusters to determine, has reduced the request for utilization of detection system, and has further improved automaticity.

Accompanying drawing explanation

Fig. 1 is detection system structural representation;

Fig. 2 is the rotary-cut class woodwork of surface without gray scale defect;

Fig. 3 is the rotten defect of rotary-cut class wood product surface;

Fig. 4 is rotary-cut class wood product surface rascal defect;

Fig. 5 is rotary-cut class wood product surface knot defect;

Fig. 6 is that rotary-cut class wood product surface pollutes defect;

Fig. 7 is the smoothness characteristic quantity result of calculation of the rotary-cut class woodwork image of surperficial normal and surperficial gray scale defect;

In figure, 1.CCD video camera, 2. blue led strip source, 3. image pick-up card, 4. computer system, 5. detecting step.

Embodiment

The specific embodiment of the present invention is described in conjunction with the accompanying drawings and embodiments.According to Fig. 1, set up measuring system, comprise 1, two blue led strip source 2 of ccd video camera, image pick-up card 3 and computing machine 4; Described two light sources 2 are arranged symmetrically with and adopt high angle lighting system, and described ccd video camera 1 be placed in two light sources 2 between position on the upper side; Described image pick-up card 3 is connected with described computing machine 4 with described ccd video camera 1 respectively.

Carrying out in actual detection, in the time of under rotary-cut class woodwork is sent to ccd video camera 1 successively, ccd video camera 1 carries out image acquisition, then image is converted into digital picture and is inputted computing machine 4 by image pick-up card 3.Choose the rotary-cut class woodworks that 100 width surfaces are normal and all kinds of gray scale defects are contained on 100 width surfaces, respectively it is carried out that image acquisition, target area are determined, smoothness characteristic quantity calculates, according to maximum variance between clusters, determines discrimination threshold T; The discrimination threshold T=0.0089 determining in this example.

Detection system is formally started working, and rotary-cut class woodwork to be detected is carried out to image acquisition, then utilizes following visible detection method it to be carried out to image processing and judge whether this rotary-cut class woodwork exists gray scale defect:

Step 1, image target area is determined, is utilized Threshold sementation to extract the target area of rotary-cut class woodwork;

Step 2, calculates the normalization histogram of target area, then according to formula (3), calculates smoothness characteristic quantity R;

Step 3, the judgement of gray scale defect, when R is more than or equal to T, presentation surface is normal; On the contrary, if when R is less than T, presentation surface contains gray scale defect.

For the rotary-cut class woodwork of having classified in advance, randomly draw normal and 1000 samples that all kinds of gray scale defects are contained on surface in 1000 surfaces, surperficial normal and surperficial gray scale defect image example is as shown in Fig. 2 to 6; Utilize the present invention to carry out surperficial gray scale defects detection to 2000 rotary-cut class woodworks, testing result is: in 1000 normal woodworks in surface, have 3 to be detected defectiveness; 1000 have 991 to be detected defectiveness containing in the woodwork of gray scale defect, and defect detection rate is 99.1%, and it is 0.3% that mistake is selected rate, and it is 0.9% that rate is selected in leakage.Testing result shows, the present invention can effectively detect rotary-cut class wood product surface gray scale defect, and has high defect detection rate and low mistake selects rate and leakage to select rate.

Claims (3)

1. the rotary-cut class wood product surface gray scale defect inspection method based on smoothness characteristic quantity, comprise two LED strip sources that are arranged symmetrically with and adopt high angle lighting system, be placed in the ccd video camera of the position, upper middle of two light sources, image pick-up card and computer system; Described ccd video camera collects rotary-cut class woodwork image, by described image pick-up card, be translated into digital picture and be input to described computer system, then utilize gray scale defect inspection method further to judge whether this rotary-cut class woodwork exists gray scale defect.

2. gray scale defect inspection method according to claim 1, is characterized in that, described detection method comprises three steps below:

Step 1, image target area is determined, utilizes Threshold sementation to extract rotary-cut class woodwork target area;

Step 2, calculates normalization histogram and the smoothness characteristic quantity R of target area;

Step 3, the judgement of gray scale defect, when R is more than or equal to discrimination threshold T, represents that rotary-cut class wood product surface is without gray scale defect; On the contrary, when R is less than T, represent that rotary-cut class wood product surface contains gray scale defect.

3. detection method according to claim 2, is characterized in that, definite use self-adaptation maximum variance between clusters of discrimination threshold T in described step 3, and concrete steps are:

1) choose the rotary-cut class woodwork image that n width surface is normal and all kinds of gray scale defects are contained on n width surface;

2) respectively the every piece image in 2n width image is carried out to target area according to step 1 described in described visible detection method and determine, then according to described step 2, calculate the smoothness characteristic quantity R of target area;

3), by ascending sequence of R value of 2n width image, within the scope of minimum and maximum R value, successively etc. step-length is using each data as default classification thresholds T; The corresponding image of R value that is more than or equal to T is divided into the 1st class, and the corresponding image of R value that is less than T is divided into the 2nd class; If the picture number comprising in each class is w _i, wherein i=1 or 2, in each class, the mean value of R value is M _i, according to formula

calculate the corresponding inter-class variance of each default classification thresholds T;

4) find out the maximal value in all inter-class variance data, the corresponding default classification thresholds of this maximal value is the discrimination threshold T using in testing process.

Images