CN104851086A - Image detection method for cable rope surface defect - Google Patents

Abstract

An image detection method for cable surface defects, the main purpose of which is to detect whether the cable surface image has defects. The implementation steps are: the cable surface image is grayscaled first, and the grayscaled image uses improved local grayscale contrast The method carries out enhancement processing, and then uses the improved maximum correlation method to segment the enhanced image, and finally judges the defect according to the size of the segmented area of the segmented image. The invention can realize intelligent detection and judgment of cable surface defects, reduce manual visual inspection workload and fatigue, improve detection efficiency, and the method is safe and reliable.

Description

An Image Detection Method for Surface Defects of Cables

technical field

The invention relates to an image detection method for surface defects of cables, belonging to the technical field of image processing.

Background technique

As one of the main stress-bearing components of bridges—cables, their surfaces will be damaged to varying degrees due to long-term exposure in the air. In the past, the surface of the cable was mostly inspected by manual observation, but manual inspection is very easy to cause fatigue, resulting in missed inspections and other problems.

With the development of science and technology, image processing technology has become one of the research directions in many fields. With the advantages of fast speed, high precision, and never fatigue, it has achieved results in many industries. Using image processing technology to detect cable surface defects can not only improve the detection efficiency and reduce labor costs, but also lay the foundation for the objective standard of the cable quality evaluation system.

Cable surface images acquired by a cable-climbing robot usually contain noise components. Because it is unavoidable to be disturbed by noise during the collection and transmission process. Therefore, it is necessary to perform noise removal processing on the cable image. Image filtering and denoising techniques can be used for denoising; specific methods include image enhancement, image smoothing, and image sharpening. Image enhancement refers to a processing method that highlights some information in an image and reduces other information according to specific requirements. Image enhancement is not to enhance the information of the original image, but to enhance the discrimination of some information, and image enhancement will lose some information of the image. At present, there is no unified evaluation standard for image enhancement effect. Of course, there are many image enhancement methods, which can be divided into two categories according to the processing domain: spatial domain enhancement and frequency domain enhancement. Image enhancement techniques can also be roughly divided into direct enhancement and indirect enhancement. Indirect enhancement is to enhance the contrast of the image by improving the image histogram. Contrast-limited adaptive histogram equalization and global histogram equalization are commonly used methods in indirect enhancement techniques. The direct enhancement method is to increase the contrast of the image by applying a contrast measurement method. Image contrast refers to the measurement of the different levels of brightness between light and dark areas in an image. It plays a very important role in direct image enhancement.

Image segmentation is an essential step in defect detection of cable surface images. Image segmentation can extract the target area of interest to the user in the image. Image segmentation is generally based on the discontinuity and similarity of image gray to process the image. Generally, image segmentation is divided into threshold segmentation and edge detection. Edge detection is based on the discontinuity of image grayscale; threshold segmentation is based on the similarity of image grayscale. Threshold segmentation automatically obtains the optimal segmentation threshold according to certain rules, and separates the target from the background. Threshold segmentation is the basic technology in image segmentation technology. Threshold segmentation methods can be roughly divided into local segmentation methods and global segmentation methods. In the local segmentation method, the segmentation threshold of each local area is dynamically calculated by using the neighborhood of the pixel point. In the global segmentation method, the entire image region is used to obtain a fixed threshold. Due to the simplicity and effectiveness of the global segmentation method, it is applied in many fields.

Contents of the invention

Technical problem: The present invention aims at the accuracy and practicability of cable surface defect detection, and provides a method based on grayscale image, using local contrast enhancement method and maximum correlation threshold segmentation method, which realizes the detection of cable surface defect and improves the Accuracy and Practicability of Cable Surface Defect Detection Image Detection Method for Cable Surface Defects.

Technical solution: The image detection method for cable surface defects of the present invention includes the following steps:

Step 1: Perform grayscale processing on the surface image of the cable to be detected, and obtain the grayscale image as f(x,y):

f(x,y)＝0.212671×c _r (x,y)+0.71516×c _g (x,y)+0.072169×c _b (x,y)

Among them, the cable surface image to be detected is a color image c(x,y), which contains three components: red component c _r (x,y), green component c _g (x,y), blue component c _b ( x, y), and the height of the image is H pixels, the width is W pixels, (x, y) represents the two-dimensional coordinates of the image, and x∈[1,W], y∈[1,H];

Step 2: The grayscaled image f(x,y) is enhanced by the improved local grayscale contrast method to obtain the enhanced image z(x,y);

Step 3: Segment the enhanced image z(x, y) using the improved maximum correlation method to obtain the segmented image d(x, y).

In the preferred version of the present invention, in said step 2), the specific process is as follows:

Step 2.1: Divide the grayscaled image f(x,y) into several non-overlapping small window area sub-images Wherein, i represents the serial number of several small window images; the processing method is as follows: regard y as a constant, and divide the image f(x, y) whose height is H pixels and width is W pixels into subsections whose height is H pixels and width is 1 pixel image And i∈[1,W];

Step 2.2: Calculate each small window area sub-image The average value of the gray value of the pixel When the subimage is And when i∈[1,W]:

${\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; ;</span>$

Step 2.3: Calculate the local grayscale contrast value v(x,y) of each pixel in the sub-image of each small window area, when the sub-image is And when i∈(1,W):

$\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\frac{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}}& <span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ></span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; x</span>}}\end{array}\right.<span\; class="notranslate">\; ;</span>$

Step 2.4: Map the local gray contrast value to the enhanced image z(x,y) of 256 gray levels according to the following formula:

$\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; .</span>$

In the preferred version of the present invention, the specific process of said step 3) is as follows:

Step 3.1: Calculate the frequency p _j of the gray value j in the enhanced image z(x,y), where j∈[0,255]:

${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; h</span>}}$

Where f _j represents the frequency of gray value j appearing in image z(x,y);

Step 3.2: Calculate the probability distribution A:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cccc}\frac{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; ,</span>& \frac{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; ,</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; ,</span>& \frac{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}\end{array}\right\}$

Among them, T is the optimal segmentation threshold, and T∈[0,255], P ₀ and P ₁ mean the frequency of gray value 0 in the image and the frequency of gray value 1 in the image, and P _T is gray The total probability of the degree value j on the interval [0,T], namely:

${\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; ,</span>$

When P _T =0, define A={0,0,...,0};

Step 3.3: Calculate the correlation C _A (T) of the probability distribution A:

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; ln</span>}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; \{</span>\frac{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; \}</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

Step 3.4: Compute the improved correlation T _Cw (T):

T _Cw (T)＝(1-P _T ) ^α ·C _A (T)

Among them, α is the weight parameter;

Step 3.5: Traverse T∈[0,255], take T when T _Cw (T) is the largest, as the best threshold obtained, then obtain the segmented image d(x,y):

$\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; T</span>}\end{array}\right.<span\; class="notranslate">\; ;</span>$

In the preferred solution of the present invention, step 4 is also included: according to the divided image d(x, y), count the number N of d(x, y)=0, and determine whether c(x, y) has a defect:

Based on the gray image, the invention adopts an improved local contrast enhancement method and an improved maximum correlation threshold segmentation method to determine whether the image has defects according to the size of the segmented area. At present, the detection method of cable surface defects is manual visual inspection, or after recording the video of the cable surface with a climbing robot, manually observe the video to find defects. The present invention can liberate people to use computers to find defects. As for the image processing method, the present invention proposes for the first time a segmentation method using a maximum correlation threshold value after local gray level enhancement for cable surface defect detection. That is, the combination of the local gray enhancement method and the maximum correlation threshold segmentation method is proposed for the first time. In terms of cable surface defect detection, compared to using other enhancement methods (such as histogram equalization enhancement, contrast-limited adaptive histogram equalization enhancement, local normalization enhancement) and other segmentation methods (maximum between-class variance segmentation, The combination of maximum distance segmentation and maximum entropy segmentation) has high precision and recall.

Beneficial effect: compared with the prior art, the present invention has the following advantages:

With regard to the detection of cable surface defects, the method of the present invention refers to the use of image processing methods to detect cable surface defects, which is different from the use of manual observation in the past; different from the image processing methods proposed by Gao Chao et al. of Chongqing University, their method is to collect Effective information interception of the image, followed by fast median filter noise reduction, surface projection correction, sobel edge detection, dynamic threshold segmentation and mathematical morphology for target segmentation, and then defect identification and storage. However, in the method of the present invention, the collected image is grayscaled first, and then the local grayscale is enhanced, combined with the maximum correlation threshold segmentation, and then the defect is judged by the size of the segmented area. Compared with the manual observation method, the method of the present invention has low cost, high efficiency and good safety; compared with the method proposed by Chongqing University, the present invention has the advantages of fast operation speed, simple steps, high accuracy rate and high reliability .

With regard to the image processing method, the local grayscale enhancement proposed by the present invention in combination with the maximum correlation segmentation is proposed for the first time, and it is different from existing methods [enhancement methods (such as global histogram equalization, adaptive histogram equalization with limited contrast, local specification (e) in combination with segmentation methods (such as maximum inter-class variance, maximum distance, maximum entropy)], the method of the present invention is simple, fast, and highly accurate.

As for the specific steps of the present invention, the original image is processed by grayscale, and the data volume that originally needs to be processed in three channels is reduced to one channel, which can effectively improve the processing efficiency without reducing the accuracy rate.

The local contrast enhancement method is adopted, and according to the characteristics of the cable image, because the cable is cylindrical, there are obvious differences between the middle area of the collected image and the two sides, so considering the difference of the entire image, if the global enhancement method is selected, it will be very large. To reduce the enhancement effect to a certain extent, the present invention selects the local enhancement method, and considering the characteristics of the cable image: the gray value of the defect area is usually smaller than the gray value of other areas, the present invention selects the contrast enhancement method. And it can weaken the different effects of light and cable surface reflection characteristics, and effectively remove most of the noise.

The maximum correlation segmentation method is also based on the characteristics of the cable image: the gray histogram of the cable image shows a unimodal characteristic, and the maximum inter-class method is not selected, and the maximum entropy method is not selected in order to improve the efficiency of the method.

Moreover, the method is used to detect the surface defect of the cable, which has the characteristics of non-contact, fast speed, simple operation, good repeatability, etc.; and the processing speed is fast, the precision is high, and it can work stably for a long time and has high reliability. It solves the shortcomings of time-consuming, labor-intensive and low-efficiency manual detection methods.

Description of drawings

Fig. 1 is the flowchart of the whole process of the present invention

Fig. 2 is the simple flowchart of the whole process of the present invention, and the first row is the effect diagram after the second row's corresponding technical processing

Detailed ways

The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings. Under the windows operating system, VC++2010 is selected as the programming tool, and the defect detection is carried out on the cable image obtained by the equipment. This example uses a cable image acquired by the device as the detected object, and finally determines whether it has defects.

Fig. 1 is a flowchart of the whole process of the present invention.

Fig. 2 is a simple flowchart of the whole process of the present invention, and the first row of images corresponds to the result of the second row of technical processing. For example, the processing result of the "threshold segmentation" technique is the corresponding image above it, and the specific "threshold segmentation" technique is an improved maximum correlation threshold segmentation method adopted in the present invention.

Aiming at the accuracy and practicability of cable surface defect detection, the present invention first uses the grayscale technology to grayscale the collected color images, which can effectively reduce the amount of data processed by the algorithm. Then the improved local contrast method is used to remove most of the noise of the image, and then the improved maximum correlation segmentation method is used to segment the image with binary value, and finally it is determined whether the detected image has defects according to the size of the segmented area.

When the detected cable surface defect images are statistically analyzed, it is convenient to carry out unified quality management on the cables, establish a corresponding quality evaluation system, and have a wide range of applications and is safe and reliable.

The invention uses image processing technology to detect the surface defect of the cable, and has the characteristics of non-contact, fast speed, simple operation, and good repeatability; the system can use natural light as a light source, has fast processing speed, high precision, and can work stably for a long time. Fast speed and high reliability.

Specific steps are as follows:

Step 1: Perform grayscale processing on the surface image of the cable to be detected, and obtain the grayscale image as f(x,y):

f(x,y)＝0.212671×c _r (x,y)+0.71516×c _g (x,y)+0.072169×c _b (x,y)

Among them, the cable surface image to be detected is a color image c(x,y), which contains three components: red component c _r (x,y), green component c _g (x,y), blue component c _b ( x, y), and the height of the image is H pixels, the width is W pixels, (x, y) represents the two-dimensional coordinates of the image, and x∈[1,W], y∈[1,H];

The second picture in the first row of Figure 2 is the image after grayscale.

Specific grayscale methods include component method, maximum value method, and average value method. The method used in the present invention is the weighted average method. Since the human eye has the highest sensitivity to green and the lowest sensitivity to blue, a more reasonable grayscale image can be obtained by weighting the red, green, and blue components according to the above formula. . In the present invention, the main purpose of grayscale processing is to reduce the amount of processed data and ultimately improve efficiency.

Step 2: The grayscaled image f(x,y) is enhanced using the improved local grayscale contrast method to obtain the enhanced image z(x,y). The specific process is as follows:

Step 2.1: Divide the grayscaled image f(x,y) into several non-overlapping small window area sub-images Wherein, i represents the serial number of several small window images; the processing method is as follows: regard y as a constant, and divide the image f(x, y) whose height is H pixels and width is W pixels into subsections whose height is H pixels and width is 1 pixel image And i∈[1,W].

The non-overlapping small window areas in this step are in the shape of strips. In fact, small window areas of other shapes, such as squares, circles, and other shapes, can also be selected according to the specific characteristics of the image. The strip shape selected by the present invention is based on that the cable is cylindrical, and the acquisition camera collects the cable image along the length direction of the cable along with the cable-climbing robot. The optical characteristics of the cable surface are similar; other shapes such as squares and circles are not selected, considering that the surface optical characteristics along the circumferential direction of the cable vary greatly.

Step 2.2: Calculate each small window area sub-image The average value of the gray value of the pixel When the subimage is And when i∈[1,W]:

${\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; ;</span>$

Step 2.3: Calculate the local grayscale contrast value v(x,y) of each pixel in the sub-image of each small window area, when the sub-image is And when i∈[1,W]:

$\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\frac{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}}{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}}& <span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; i</span>}}\\ \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ></span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; x</span>}}\end{array}\right.<span\; class="notranslate">\; ;</span>$

According to the characteristics of the cable image, in the small window area, the gray value of the defect area is smaller than the gray value of the background area; in the small area, the external light can be considered the same, and the optical characteristics of the cable surface in the small area are similar; then the small window In , the area with a gray value smaller than the average value is more likely to be a defect; while the area with a gray value larger than the average value is more likely to be a noise; in the case of ensuring the detection accuracy, the gray value is greater than the average value The area of is set to 0, this kind of processing can not only reduce the amount of calculation, but also achieve the effect of eliminating part of the noise.

Step 2.4: Map the local gray contrast value to the enhanced image z(x,y) of 256 gray levels according to the following formula:

$\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; min</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; 255</span>}<span\; class="notranslate">\; ;</span>$

The third sub-picture in the first row of Fig. 2 is the image enhanced by the enhancement method of the present invention.

Step 3: Use the improved maximum correlation method to segment the enhanced image z(x,y) to obtain the segmented image d(x,y). The specific process is as follows:

Step 3.1: Calculate the frequency p _j of the gray value j in the enhanced image z(x,y), where j∈[0,255]:

${\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; W</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; h</span>}}$

Where f _j represents the frequency of gray value j appearing in image z(x,y),

Step 3.2: Calculate the probability distribution A:

$\mathrm{<span\; class="notranslate">\; A</span>}<span\; class="notranslate">\; =</span>\left\{\begin{array}{cccc}\frac{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; ,</span>& \frac{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; ,</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; ,</span>& \frac{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}\end{array}\right\}$

Among them, T is the optimal segmentation threshold, and T∈[0,255], P ₀ and P ₁ are the frequency of occurrence in the image when the gray value is 0 and 1, respectively. P _T is the total probability of the gray value j in the interval [0, T], that is:

${\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; ,</span>$

When P _T =0, define A={0,0,...,0};

Step 3.3: Calculate the correlation C _A (T) of the probability distribution A:

${\mathrm{<span\; class="notranslate">\; C</span>}}_{\mathrm{<span\; class="notranslate">\; A</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; ln</span>}\underset{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; \{</span>\frac{{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; T</span>}}}<span\; class="notranslate">\; \}</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

Step 3.4: Compute the improved correlation T _Cw (T):

T _Cw (T)＝(1-P _T ) ^α ·C _A (T)

Among them, α is the weight parameter;

The improved maximum correlation method is mainly based on the characteristics of the cable image: in the defect image, the defect area is small, and the optimal segmentation threshold is greatly affected by the correlation C _A (T).

In this step, the value of the weight parameter α is more important because it can change the selection of the optimal threshold. In this example, the value of the weight parameter α is 1. Research shows that the accuracy of detection increases with the increase of the value of the weight parameter α, while the recall rate of detection decreases with the increase of the value of the weight parameter α. This is because when the weight parameter α is larger, the optimal threshold obtained is smaller, so that the detection of defect areas is reduced, and the accuracy of detected defects is improved, and the accuracy is increased, and the defect areas will be missed, then The recall rate drops. In actual image processing, there are many defects that are missed, but there are few false defects among the detected defects.

Step 3.5: Traverse T∈[0,255], take T when T _Cw (T) is the largest, as the best threshold obtained, then obtain the segmented image d(x,y):

$\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ></span>\mathrm{<span\; class="notranslate">\; T</span>}\end{array}\right.<span\; class="notranslate">\; ;</span>$

The fourth sub-picture in the first row of Fig. 2 is the image segmented by the segmentation method of the present invention.

Step 4: According to the segmented image d(x, y), count the number N of d(x, y)=0, and determine whether c(x, y) has a defect:

It should be pointed out that there are many methods for defect judgment. The present invention considers that the above method is simple and practical, and under the condition of ensuring the detection accuracy, the judgment threshold can change. The judgment threshold selected in this step is based on the characteristics of the cable image: the defect area occupies a small area.

The foregoing embodiments are only preferred implementations of the present invention. It should be pointed out that for those of ordinary skill in the art, without departing from the principles of the present invention, several improvements and equivalent replacements can be made, which are important to the rights of the present invention. Technical solutions requiring improvement and equivalent replacement all fall within the protection scope of the present invention.

Claims (4)

1. for an image detecting method for cable surface imperfection, it is characterized in that, the method step is as follows:

Step 1: cable surface image to be detected is carried out gray processing process, obtaining the image after gray processing is f (x, y):

f(x,y)＝0.212671×c _r(x,y)+0.71516×c _g(x,y)+0.072169×c _b(x,y)

Wherein, cable surface image to be detected is coloured image c (x, y), and it comprises 3 components: red component c _r(x, y), green component c _g(x, y), blue component c _b(x, y), and picture altitude be H pixel, width is W pixel, (x, y) represents the two-dimensional coordinate of image, and x ∈ [1, W], y ∈ [1, H];

Step 2: adopt the local gray level contrast method improved to carry out enhancing process, image z (x, y) after being enhanced to the image f (x, y) after gray processing;

Step 3: adopt the maximal correlation method improved to carry out dividing processing to the image z (x, y) after strengthening, obtain splitting rear image d (x, y).

2. the image detecting method for cable surface imperfection according to claim 1, is characterized in that, described step 2), detailed process is as follows:

Step 2.1: the image f (x, y) after gray processing is divided into some nonoverlapping wicket regions subimage wherein i represents the numbering of some wicket images; Disposal route is: be constant depending on y, the subimage that the image f (x, y) that is highly W pixel for H pixel, width to be divided into be highly H pixel, width is 1 pixel and i ∈ [1, W];

Step 2.2: calculate each wicket region subimage the mean value of grey scale pixel value when subimage is and time i ∈ [1, W]:

Step 2.3: the local gray level correlative value v (x, y) calculating each pixel in the subimage of each wicket region, when subimage is and time i ∈ (1, W):

Step 2.4: described local gray level correlative value is mapped to image z (x, y) after the enhancing of 256 gray levels according to following formula:

3. the image detecting method for cable surface imperfection according to claim 1, is characterized in that, described step 3) detailed process as follows:

Step 3.1: calculate the frequency p that gray-scale value j occurs in image z (x, y) after enhancing _j, wherein j ∈ [0,255]:

Wherein f _jrepresent the frequency that gray-scale value j occurs in image z (x, y);

Step 3.2: calculating probability distribution A:

Wherein, T is optimal segmenting threshold, and T ∈ [0,255], P ₀, P ₁implication is respectively frequency that gray-scale value 0 occurs in the picture and the frequency that gray-scale value 1 occurs in the picture, P _tfor the general probability of gray-scale value j on interval [0, T], that is:

Work as P _twhen=0, definition A={0,0 ..., 0};

Step 3.3: the relevant C of calculating probability distribution A _a(T):

Step 3.4: the relevant T of computed improved _cw(T):

T _Cw(T)＝(1-P _T) ^α·C _A(T)

Wherein, α is weight parameter;

Step 3.5: traversal T ∈ [0,255], gets and makes T _cw(T) T time maximum, as the optimal threshold obtained, then obtains image d (x, y) after segmentation:

。

4. the image detecting method for cable surface imperfection according to claim 1,2 or 3, it is characterized in that, the method also comprises step 4: according to image d (x after described segmentation, y), statistics d (x, the number N of y)=0, judges whether c (x, y) has defect: