A method and system for visual inspection of bicycle saddle machining quality

Abstract

The present application relates to image processing and machine vision technical field, especially to a kind of visual inspection method and system of bicycle saddle machining quality.The present application includes the following steps:S1: obtain the color image of bicycle saddle, with the decorative small hole of bicycle saddle as center, separate the red, green and blue three color channels in color image;S2: evaluate the continuity and progression in three color channels in color image, based on the local area divided by each decorative small hole as center, calculate the mixing degree of local area by continuity and progression;The continuity in each color channel refers to the smoothness and similarity of pixel value in the same color channel.The present application ensures stable and reliable detection by high-resolution image processing and illumination normalization;Evaluate the continuity and progression of color channel, accurately identify defects;Compare regional color difference, combined with time series analysis, accurately locate and evaluate defects.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing and machine vision, and particularly relates to a visual detection method and system for bicycle saddle processing quality. BACKGROUND

[0002] The quality of a bicycle saddle directly affects the comfort and safety of a rider. A high-quality saddle should have good durability, shock-absorbing performance, and aesthetics. However, in actual production, defects such as cracks, scratches, and uneven color may appear on the surface of the saddle, which not only affects the appearance of the product, shortens its service life, but also brings potential safety hazards. Therefore, it is crucial to accurately and efficiently detect the processing quality of bicycle saddles.

[0003] Traditional manual visual inspection methods are inefficient and easily affected by subjective factors, making it difficult to meet the needs of mass production. With the development of computer vision technology, automatic detection methods based on image processing have gradually become the mainstream. However, existing methods still have limitations in dealing with complex lighting conditions, varying color and texture characteristics: images obtained under different lighting conditions can significantly affect the detection results. The color and texture of the saddle surface are diverse, increasing the difficulty of defect recognition. The presence of decorative small holes makes it difficult for traditional image contrast methods to accurately assess quality. SUMMARY

[0004] To overcome the low efficiency and accuracy of saddle quality detection, the present application provides a visual detection method and system for bicycle saddle processing quality.

[0005] The technical solution of the present application is: a visual detection method for bicycle saddle processing quality, comprising the following steps:

[0006] S1: Obtain a color image of a bicycle saddle, and separate the red, green, and blue color channels in the color image with the decorative small hole of the bicycle saddle as the center;

[0007] S2: Evaluate the continuity and progression in the three color channels of the color image, and calculate the mixing degree of the local area based on the continuity and progression, with each decorative small hole as the center; the continuity in the three color channels of the color image refers to the smoothness and similarity of the pixel values in the same color channel, the similarity refers to the color value difference between adjacent pixels, and the progression refers to the consistency degree of the color value change direction of adjacent pixels and the overall color value change direction in the region in the same color channel;

[0008] S3: dividing the bicycle saddle into two regions, the left region being a detection region and the right region being a verification region, and comparing the difference of the corresponding color channel values in the two regions to evaluate the difference degree of the color channels in the corresponding regions;

[0009] S4: determining a detection focus based on the mixing degree and the difference degree.

[0010] Preferably, the color image of the bicycle saddle is acquired, the decorative small holes of the bicycle saddle are taken as centers, and red, green and blue color channels in the color image are separated, comprising:

[0011] Acquiring the illumination intensity value of the decorative small hole region of the bicycle saddle;

[0012] Taking each decorative small hole as a center, red, green and blue color channel intensity values in a range with a radius of 1 / 2 of the distance from the center of the decorative small hole to the nearest adjacent small hole are extracted;

[0013] Respectively dividing the red, green and blue color channel intensity values by the illumination intensity value to obtain first defect judgment ratios of the decorative small hole region;

[0014] Under standard illumination conditions, a non-defective saddle sample is collected, the first defect judgment ratios of each color channel are calculated, and the maximum value is taken as a preset threshold value;

[0015] According to the preset threshold value, a suspected defect region of the bicycle saddle region is analyzed in the illumination intensity variation process.

[0016] Preferably, the analysis of the suspected defect region of the bicycle saddle region in the illumination intensity variation process according to the preset threshold value comprises:

[0017] In the illumination intensity variation process, the first defect judgment ratios of the entire bicycle saddle region are calculated;

[0018] If the three first defect judgment ratios of a certain region are all greater than the preset threshold value, the region is marked as a suspected defect region;

[0019] If the first defect judgment ratio of any one channel of a certain region is less than the preset threshold value, the region is not marked as a suspected defect region.

[0020] Preferably, the continuity in each color channel refers to the smoothness and similarity of pixel values in the same color channel, comprising:

[0021] Based on the suspected defect region, the variation values of the red, green and blue channel values in the suspected defect region are calculated and defined as first variation values;

[0022] Meanwhile, a variation value of the red, green and blue channel values of the adjacent region of the suspected defect region is calculated and defined as a second variation value;

[0023] If the first variation value and the second variation value are not equal, a variation value of the red, green and blue channel values of the adjacent region is further calculated and defined as a third variation value;

[0024] Based on the first variation value, the second variation value and the third variation value, the continuity degree of the suspected defect region is determined.

[0025] Preferably, the progression degree refers to the consistency degree of the variation direction of the color value of the adjacent pixel and the variation direction of the overall color value in the region in the same color channel, including:

[0026] Based on the suspected defect region, the red, green and blue channel values of all suspected defect regions are calculated;

[0027] Based on the red, green and blue channel values, the progression degree of the color channel value between each suspected defect region and the nearest neighbor suspected defect region is calculated;

[0028] The progression degree of the suspected defect region from the rear end of the saddle to the front end of the saddle is obtained with the rear end of the saddle as the starting point.

[0029] Preferably, the evaluation of the continuity degree and the progression degree in the three color channels of the color image is based on the local region divided by each decorative small hole as the center, the mixing degree of the local region is calculated by the continuity degree and the progression degree, including:

[0030] The union region of the continuity degree and the progression degree of the decorative small hole region is obtained; the mixing degree is calculated by the mixing degree formula, and the mixing degree formula is as follows,

[0031]

[0032] Wherein, is the mixing degree, is the continuity degree area of the i-th decorative small hole region; is the progression degree area of the i-th decorative small hole region; is the continuity degree of the i-th decorative small hole region; is the progression degree of the i-th decorative small hole region; is the number of the union region of the decorative small hole; is the minimum value.

[0033] ​​​​Preferably, the bicycle saddle is divided into two regions, the left side is a detection region, and the right side is a verification region. The difference between the corresponding color channel values in the two regions is compared to evaluate the degree of difference of the color channel in the corresponding region, including:

[0034] The difference between the three channel values of the decorative hole region of the detection region is calculated, specifically the difference between the maximum and minimum values of the three channel values. The difference value is used as the degree of difference of each region;

[0035] Based on the calculated degree of difference of each region, a preset threshold is used to screen out regions that exceed the preset threshold, and then compared with the verification region on the right side;

[0036] Based on the comparison result, the overall degree of difference of the bicycle saddle is determined.

[0037] Preferably, the detection focus is determined based on the mixing degree and the degree of difference, including:

[0038] Obtain the light intensity variation curve of the bicycle saddle at time;

[0039] Draw the mixing degree curve and the degree of difference curve of the decorative hole region of the bicycle saddle at time;

[0040] Draw the mean curve of the mixing degree curve and the degree of difference curve;

[0041] Determine the final defect region by a verification formula for the suspected defect region and the adjacent region.

[0042] Preferably, the final defect region is determined by a verification formula for the suspected defect region and the adjacent region, including:

[0043] Based on the light intensity variation curve, obtain the light intensity variation value;

[0044] Based on the mean curve, obtain the mean variation value;

[0045] Determine the final defect region by a verification formula, the verification formula is as follows,

[0046]

[0047] Wherein, is the defect degree, is the light intensity variation value, is the mean variation value.

[0048] A visual detection system for the processing quality of a bicycle saddle, comprising:

[0049] Image acquisition and preprocessing module: use high-resolution camera to acquire color image of bicycle saddle, separate red, green and blue color channels centered on decorative small hole; normalize each channel intensity by light intensity value, identify suspected defect area;

[0050] Continuity and progression evaluation module: based on suspected defect area, calculate the change of red, green and blue channel values to determine the continuity and progression;

[0051] Difference evaluation module: divide the saddle into detection area and verification area, compare the difference of color channel values in the two areas, evaluate the difference, filter out the area exceeding the preset threshold, and perform comparative analysis;

[0052] Detection focus determination module: combine the mixing degree and difference degree, draw the time curve of light intensity, mixing degree and difference degree, calculate the mean curve, and determine the final defect area through the verification formula.

[0053] Beneficial effects: by introducing high-resolution color image processing and light intensity normalization technology, the invention first solves the problem of inaccurate detection caused by environmental light changes and color diversity in traditional detection methods, ensuring the stability and reliability of the detection results. On this basis, further evaluation of the continuity and progression of each color channel is realized to accurately analyze the smoothness of the color transition on the saddle surface, enhancing the meticulousness of defect identification. Then, by comparing the color channel value differences in different areas of the saddle, potential processing defects are effectively screened out, improving the comprehensiveness of the detection. Finally, combined with time series analysis and comprehensive scoring mechanism, not only the defect location can be accurately positioned, but also the severity can be evaluated, providing a scientific basis for quality control in the production process. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 Flowchart of the bicycle saddle processing quality visual detection method of the invention;

[0055] Figure 2 System diagram of the bicycle saddle processing quality visual detection of the invention. DETAILED DESCRIPTION

[0056] The technical solutions in the embodiments of the invention will be described clearly and completely below with reference to the drawings in the embodiments of the invention. Obviously, the described embodiments are only part of the embodiments of the invention, not all. Based on the embodiments in the invention, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the invention.

[0057] Example 1: a bicycle saddle processing quality visual detection method, as Figure 1As shown, comprising the following steps:

[0058] S1: Obtain a color image of the bicycle saddle, take the decorative small holes of the bicycle saddle as the center, and separate the red, green, and blue color channels in the color image;

[0059] S2: Evaluate the continuity and progression in the three color channels in the color image, based on the local area divided by each decorative small hole as the center, calculate the mixing degree of the local area by the continuity and progression; the continuity in the three color channels in the color image refers to the smoothness and similarity of the pixel value in the same color channel, the similarity refers to the color value difference between adjacent pixels, and the progression refers to the consistency degree of the color value change direction of the adjacent pixels and the overall color value change direction in the region in the same color channel;

[0060] The adjacent pixels refer to the directly adjacent pixels (such as up, down, left, right, or diagonal), emphasizing the immediate proximity of the spatial position; the adjacent pixels refer to the pixels within a certain range (not limited to direct adjacency), focusing on the proximity within the region.

[0061] S3: Divide the bicycle saddle into two regions, the left side is the detection region, the right side is the verification region, and compare the difference of the corresponding color channel values in the two regions to evaluate the difference degree of the color channel in the corresponding region;

[0062] S4: Determine the detection focus based on the mixing degree and the difference degree.

[0063] Obtain a color image of the bicycle saddle, take the decorative small holes of the bicycle saddle as the center, and separate the red, green, and blue color channels in the color image, comprising:

[0064] Obtain the light intensity value of the decorative small hole region of the bicycle saddle;

[0065] Take each decorative small hole as the center, and extract the red, green, and blue color channel intensity values of the small hole region within a range of 1 / 2 of the distance from the center of the decorative small hole to the nearest adjacent small hole;

[0066] Divide the red, green, and blue color channel intensity values by the light intensity value, respectively, as the first defect judgment ratio of the decorative small hole region;

[0067] Under standard lighting conditions, collect a defect-free saddle sample, calculate the first defect judgment ratio of each color channel, and take the maximum value as the preset threshold;

[0068] According to the preset threshold, analyze the suspected defect region of the bicycle saddle region during the light intensity change process.

[0069] Further, the light intensity value at the center of each decorative hole is measured using a high-precision light meter ; then, the intensity values of the red, green, and blue color channels are extracted in a circular region centered on each decorative hole and having a radius of half the average distance between adjacent holes . Subsequently, these intensity values are divided by the light intensity value by normalization processing , , .

[0070] According to the preset threshold value, during the change in the light intensity, the suspected defect region of the bicycle saddle region is analyzed, including:

[0071] During the change in the light intensity, the first defect judgment ratio of the entire bicycle saddle region is calculated;

[0072] If the three first defect judgment ratios of a certain region are all greater than the preset threshold value, the region is marked as a suspected defect region;

[0073] If the first defect judgment ratio of any one channel of a certain region is less than the preset threshold value, the region is not marked as a suspected defect region.

[0074] Further, if all the first defect judgment ratios deviate from the preset threshold value, it indicates that the color or reflection characteristics of the region have abnormally changed under the current light condition, due to cracks, scratches, or other surface defects in the processing of the bicycle saddle. If all the first defect judgment ratios still exceed the preset threshold value under different light conditions, it indicates the presence of an actual defect. If at least one first defect judgment ratio is within the preset threshold value and does not significantly deviate from the expected color pattern. In this case, the color change of the region is considered to be normal, caused by the texture or slight color difference of the material itself, rather than a processing defect.

[0075] The continuity in each color channel refers to the smoothness and similarity of pixel values in the same color channel, including:

[0076] Based on the suspected defect region, the change value of the red, green, and blue channel values in the suspected defect region is calculated and defined as the first change value;

[0077] At the same time, the change value of the red, green, and blue channel values of the adjacent region of the suspected defect region is calculated and defined as the second change value;

[0078] If the first change value and the second change value are not equal, the change value of the red, green, and blue channel values of the adjacent region is further calculated and defined as the third change value;

[0079] Based on the first change value, the second change value and the third change value, the continuity degree of the suspected defect region is determined.

[0080] Further, if the first change value and the second change value are different, it indicates that the saddle has a defect, and the defect degree has a progressive relationship. In order to further confirm the defect severity, the third change value is introduced to verify the initial judgment. By calculating the color channel value change of the adjacent region, additional data support can be provided, so as to help more accurately evaluate the existence and severity of the defect, thereby ensuring the reliability of the detection result. The specific calculation method is that for each suspected defect region, the average values of the red, green and blue three color channels are extracted, and the color difference (RGB Euclidean distance) of all adjacent pixel pairs in the suspected defect region is calculated, and the average value is taken, reflecting the mutation degree of the color inside the region, defined as the first change value ,

[0081]

[0082] is the number of adjacent pixel pairs in the region, the color channel mean difference between the suspected defect region and the directly adjacent region is calculated to judge whether the defect spreads outward, defined as the second change value ,

[0083]

[0084] If the first change value and the second change value are different (i.e. ), wherein is a set threshold, the difference between the adjacent region (second layer adjacent region) of the adjacent region and the current adjacent region is further calculated to verify the progression of the defect,

[0085]

[0086] The third change value represents the color channel value difference between the adjacent region (i.e. the second layer adjacent region) of the adjacent region and the current adjacent region, which is used to further verify the progression of the defect. Among them, is the color channel mean value of the current adjacent region, , , is the color channel mean value of the second layer adjacent region.

[0087] The progression degree refers to the consistency degree of the color value change direction of the adjacent pixels in the same color channel and the change direction of the overall color value in the region, including:

[0088] Based on the suspected defect area, the red, green, blue three channel values of all suspected defect areas are calculated;

[0089] Based on the red, green, blue three channel values, the progression degree of color channel value between each suspected defect area and the nearest neighbor suspected defect area is calculated;

[0090] With the rear end of the bicycle saddle as the starting point, the progression degree of the suspected defect area from the rear end of the saddle to the front end of the saddle is obtained.

[0091] Further explanation is that, when the progression degree exists between the suspected defect areas with the closest distance between each other, it means that the color change of the intermediate area is blocked or covered, resulting in misjudgment. In order to ensure the accuracy of the detection result, it is necessary to judge the progression degree of these areas. If it is confirmed that it is not a misjudgment caused by color change, but a non-progressive change caused by other factors, these defects exist independently and need to be further analyzed and confirmed. By judging the progression degree between the suspected defect areas with the closest distance between each other, the true continuous defects and the misjudgments caused by color change can be effectively distinguished. The specific calculation method is that, for each suspected defect area, the average values of its red, green and blue three color channels are extracted; then, the regions with the closest distance between each other are selected, and the Euclidean distance formula is used,

[0092]

[0093] The color difference between them is calculated, and the comprehensive difference of two adjacent suspected defect areas in RGB color space is quantified by Euclidean distance. Then, starting from the tail end of the bicycle saddle, the progression degree of adjacent regions is calculated and accumulated pair by pair, and the total progression degree is obtained,

[0094]

[0095] Along the direction from the rear end to the front end of the saddle, the color difference values of adjacent regions are accumulated pair by pair to form a global progression index.

[0096] The continuity and progression degree in the three color channels in the color image are evaluated, and the mixing degree of the local area centered on each decorative hole is calculated based on the local area divided by the continuity and progression degree, including:

[0097] The union area of the continuity and progression degree of the decorative hole area is obtained; the mixing degree is calculated by the mixing degree formula, and the mixing degree formula is as follows,

[0098]

[0099] Among them, The mixing degree is, The first is the first color channel value, The degree of continuity of the area of ​​each decorative perforation; For the first The progressive area of ​​each decorative perforated region; For the first The degree of continuity of the decorative perforation area; For the first The degree of progression of the decorative perforated areas; The number of decorative pinhole union regions; It is a local minimum.

[0100] To further illustrate, suppose the total area of ​​a decorative perforation region is 10 square units, of which the area of ​​continuity accounts for 6 square units (60%) and the area of ​​progression accounts for 4 square units (40%). In this case, the numerator is relatively large (i.e., in the formula for the degree of mixing, the value of the numerator is significantly higher than the denominator). This manifests in the following two scenarios: High coverage of both continuity and progression regions: When a large area of ​​continuity-related regions (such as a color abrupt change region caused by a crack) and progression-related regions (such as a region where color differences diffuse outwards) exist simultaneously within the decorative perforation region, and the union of these two regions almost covers the entire perforation, the numerator value will increase significantly. For example, if the area of ​​continuity of a perforation... =6 cm 2 Progressive degree area =4 cm 2 The overlap area between the two is 2cm. 2 The numerator is 6 + 4 - 2 = 8 cm. 2 The denominator is relatively small: if the continuity score is... And the degree of progression rating Lower (e.g.) and If the value is close to the lower limit of the judgment threshold, then the ∑ in the denominator and ∑ It will remain at a smaller value. For example, when =0.2、 When the denominator is 0.3, the denominator is 0.2 + e0.3 ≈ 0.2 + 1.35 = 1.55. If the numerator is 8 cm2... 2 ,but ≈5.16 indicates that the molecule is significantly dominant. Practical significance: A relatively large molecular portion indicates the presence of a wide range of abnormal color changes (such as cracks or scratches) in the detection area, and these abnormal areas exhibit both abrupt changes (high continuity) locally and diffusion to the surrounding areas (high progression). At this point, the degree of mixing... An increase in the value indicates that the area requires priority for manual re-inspection or process adjustment to avoid potential quality defects. The denominator (the continuity value and the exponentially adjusted progression value) is 2 and 3 respectively. This leads to... The value is large, about 1.96. The large M value indicates that there is a significant color variation or potential defect in the region, which is caused by cracks or scratches during processing, suggesting that further detailed inspection is needed to confirm the quality problem. Looking at another decorative hole region, the total area is also 10 square units, but the continuous degree area is only 2 square units (20%), and the progressive degree area is 8 square units (80%). At this time, the numerator is relatively small, and the denominator (the continuous degree value and the progressive degree value adjusted by the index) is 8 and 7, respectively. This leads to The value is small, about 0.66. The small M value indicates that the color variation is relatively smooth and there is no obvious defect, suggesting that the region is of good quality and does not need further inspection. In calculating the continuous degree, according to the first variation value , the second variation value , and the third variation value defined earlier, a continuous degree judgment criterion is set. For example, the continuous degree related region judgment: when both the continuous degree index and the progressive degree index exceed the preset threshold (such as >α, >β), and the difference between the two exceeds the allowed range (i.e. ), the third variation value is further calculated to verify the progression of the defect. If also exceeds the verification threshold (such as >γ), the region is marked as a continuous degree related region. The pixel area of such regions in the decorative hole is counted by image analysis software, and converted into actual physical area based on the pre-marked resolution parameter (such as 0.1 mm² per pixel), as the continuous degree area . Progressive degree related region judgment: along the texture direction of the saddle surface (such as from the front end to the rear end), the color difference between adjacent suspected defect regions is calculated in turn (through the Euclidean distance formula). If the color difference between consecutive regions shows a monotonic increasing or decreasing trend (such as the absolute value of the linear regression slope k > δ), the region is marked as a progressive degree related region. The pixel area of such regions is counted using image analysis tools, and converted into actual area according to the same resolution, as the progressive degree area .

[0101] The bicycle saddle is divided into two regions, the left side is the detection region and the right side is the verification region, and the difference between the corresponding color channel values in the two regions is compared to evaluate the degree of difference of the color channel in the corresponding region, including:

[0102] ​The difference between the three channel values ​​of the decorative pinhole region in the detection area is calculated. Specifically, the difference between the maximum and minimum values ​​of the three channel values ​​is calculated, and the difference result is used as the degree of difference of each region.

[0103] Based on the calculated degree of difference in each region, regions exceeding the preset threshold are filtered out and then compared with the verification region on the right.

[0104] Based on the comparison results, the overall degree of difference in bicycle saddles was determined.

[0105] To further clarify, if the color changes in the detection area and the verification area are similar at the same location, it indicates that these changes are caused by actual defects; if they are inconsistent, it is a misjudgment due to ambient light, material texture, or other factors. By comparing, potential problems can be further confirmed or eliminated, ensuring the reliability of the test results. If the color changes in the detection area and the verification area are consistent and exceed a preset threshold, it is considered that there is an overall difference in the saddle, indicating a quality problem; if the color changes in most areas are within the preset threshold, the saddle is considered to be of good quality and no further inspection is required. Through comparison, the overall degree of difference in the saddle is comprehensively evaluated, providing more reliable test conclusions.

[0106] Determining the detection focus based on the degree of mixing and the degree of difference includes:

[0107] Get the bicycle saddle The curve showing the change in light intensity over time;

[0108] Draw the decorative perforation area on the bicycle saddle. The mixing degree curve and the difference degree curve at different times;

[0109] Plot the mean curves of the mixing degree curve and the difference degree curve;

[0110] The suspected defect area and adjacent areas are used to determine the final defect area through a verification formula.

[0111] The suspected defect area and adjacent areas are used to determine the final defect area using a verification formula, including:

[0112] Based on the light intensity change curve, the light intensity change value is obtained;

[0113] Based on the mean curve, the mean change value is obtained;

[0114] The determination of the central defect area is performed using a verification formula, as follows.

[0115]

[0116] in, is the degree of defect, is the illumination intensity change value, is the mean value change value.

[0117] Further, a threshold T is set to distinguish the defect area and the normal area. If T, mark the area as a potential defect area. This indicates that the illumination change is relatively large, while the quality change is small, which is a false judgment caused by the change of ambient light. If T, it is considered that the area has actual defects. This indicates that the quality change is significant, while the illumination change is small, suggesting that there is a real quality problem. If approximately equal to T, re-evaluate. The illumination intensity change value is calculated by measuring the illumination intensity of the decorative hole area of the saddle at different times using a high-precision illuminometer , and then calculating the difference in illumination intensity between adjacent times, i.e. . For the mean value change value , first calculate the mixing degree of the decorative hole area at each time , and then calculate the difference in mixing degree between adjacent times, i.e. . In determining the threshold T, a large amount of sample data is tested. Collect samples of different types and different quality conditions of bicycle saddles, and use the detection method for calculation and analysis. According to the distribution of the defect area and the normal area in the actual detection results, statistical methods such as mean value and standard deviation are used to determine a reasonable threshold T. For example, the mean value of the normal area plus a certain multiple of the standard deviation is set to ensure that the defect area and the normal area can be effectively distinguished in actual application.

[0118] Example 2: Based on Example 1, as shown in Figure 2 , a visual detection system for the processing quality of a bicycle saddle includes:

[0119] Image acquisition and preprocessing module: use a high-resolution camera to acquire a color image of the bicycle saddle, separate the red, green, and blue color channels centered on the decorative hole; normalize the intensity of each channel by the illumination intensity value to identify suspected defect areas;

[0120] Continuity and progression evaluation module: based on the suspected defect area, calculate the change of red, green, and blue channel values to determine the continuity and progression;

[0121] ​The difference degree evaluation module divides the saddle into a detection area and a verification area, compares the color channel value difference in the two areas, evaluates the difference degree, screens out the area exceeding the preset threshold, and performs comparison analysis.

[0122] The detection focus determination module draws a time curve of the illumination intensity, the mixing degree and the difference degree, calculates a mean curve, and determines the final defect area through a verification formula, in combination with the mixing degree and the difference degree.

[0123] The above has carried out the detailed introduction to the application, the principle and the implementation mode of the application have been set forth in the text by applying the specific examples, the above example explanation is only for helping understanding the method and the core thought of the application; simultaneously, for the general technical personnel in the field, according to the thought of the application, there will be changes in the specific implementation mode and the application range, and the above-mentioned is not understood as the limitation of the application.

Claims

1. A method of visually inspecting the quality of a bicycle saddle, characterized in that, The method comprises the following steps: S1: acquiring a color image of a bicycle saddle, taking a decorative hole of the bicycle saddle as a center, and separating red, green and blue color channels in the color image; S2: evaluating continuity and progression in the three color channels in the color image, calculating a mixing degree of a local area based on the continuity and progression, and dividing the local area based on each decorative hole as a center; the continuity in the three color channels in the color image refers to smoothness and similarity of pixel values in the same color channel, the similarity refers to a color value difference between adjacent pixels, and the progression refers to a consistency degree of a change direction of color values of adjacent pixels and a change direction of overall color values in the area in the same color channel; S3: dividing the bicycle saddle into two areas, taking a left side as a detection area and a right side as a verification area, and comparing a difference of corresponding color channel values in the two areas to evaluate a difference degree of the color channels in the corresponding areas; S4: determining a detection focus based on the mixing degree and the difference degree; The acquisition of the color image of the bicycle saddle, taking the decorative hole of the bicycle saddle as the center, and separating the red, green and blue color channels in the color image comprises: acquiring an illumination intensity value of a decorative hole area of the bicycle saddle; extracting red, green and blue color channel intensity values of the hole area within a range of 1 / 2 of a distance from the center of the decorative hole to the nearest adjacent hole; respectively dividing the red, green and blue color channel intensity values by the illumination intensity value to obtain first defect judgment ratios of the decorative hole area; under standard illumination conditions, collecting a non-defective saddle sample, calculating the first defect judgment ratios of the color channels, and taking a maximum value as a preset threshold value; based on the preset threshold value, analyzing a suspected defect area of the bicycle saddle area in a process of illumination intensity change.

2. The method for visual inspection of the quality of a bicycle saddle according to claim 1, characterized in that, The analysis of the suspected defect area of the bicycle saddle area in the process of illumination intensity change based on the preset threshold value comprises: calculating the first defect judgment ratios of the entire bicycle saddle area in the process of illumination intensity change; if the three first defect judgment ratios of a certain area are all greater than the preset threshold value, the area is marked as a suspected defect area; if the first defect judgment ratio of any one channel of a certain area is less than the preset threshold value, the area is not marked as a suspected defect area.

3. The method of claim 1, wherein the method further comprises: The continuity in each color channel refers to smoothness and similarity of pixel values in the same color channel, and comprises: based on the suspected defect area, calculating a change value of red, green and blue channel values in the suspected defect area and defining the change value as a first change value; at the same time, calculating a change value of red, green and blue channel values of an adjacent area of the suspected defect area and defining the change value as a second change value; if the first change value and the second change value are not equal, further calculating a change value of red, green and blue channel values of the adjacent area and defining the change value as a third change value; based on the first change value, the second change value and the third change value, determining the continuity degree of the suspected defect area.

4. The method for visual inspection of the quality of a bicycle saddle according to claim 1, characterized in that, The progression degree refers to the consistency degree of the color value change direction of the adjacent pixels in the same color channel and the overall color value change direction in the region, including: Based on the suspected defect region, the red, green, and blue channel values of all suspected defect regions are calculated; Based on the red, green, and blue channel values, the progression degree of the color channel value between each suspected defect region and the nearest neighbor suspected defect region is calculated; With the rear end of the bicycle saddle as the starting point, the progression degree of the suspected defect region from the rear end of the saddle to the front end of the saddle is obtained.

5. The method for visual inspection of the quality of a bicycle saddle according to claim 1, characterized in that, The evaluation of the continuity and progression degree in the three color channels of the color image is based on the local region centered on each decorative hole, and the mixing degree of the local region is calculated by the continuity and progression degree, including: Obtain the union region of the continuity and progression degree of the decorative hole region; calculate the mixing degree by the mixing degree formula, and the mixing degree formula is as follows, ； wherein, is a degree of mixing, is a degree of progression of the first decorative aperture region; is a degree of progression of the first decorative aperture region; is a degree of progression of the first decorative aperture region; is a degree of progression of the first decorative aperture region; is a number of decorative aperture union regions; is a minimum value.

6. The method of claim 1, wherein the method further comprises: The bicycle saddle is divided into two regions, the left side is the detection region, and the right side is the verification region, and the difference of the corresponding color channel values in the two regions is compared to evaluate the difference degree of the color channel in the corresponding region, including: The three channel values of the decorative hole region in the detection region are calculated by difference, specifically, the difference between the maximum value and the minimum value of the three channel values is calculated, and the difference result is taken as the difference degree of each region; Based on the calculated difference degree of each region, a preset threshold is used to screen out the regions that exceed the preset threshold, and then compared with the verification region on the right side; Based on the comparison result, the overall difference degree of the bicycle saddle is determined.

7. The method of claim 1, wherein the method further comprises: The detection focus is determined based on the mixing degree and the difference degree, including: acquiring a light intensity variation curve at the moment of time when the bicycle saddle is in the light. Drawing a decorative hole area of a bicycle saddle the degree of mixing curve and the degree of difference curve at the moment; Draw the mean curve of the mixing degree curve and the difference degree curve; The final defect region is determined by the verification formula for the suspected defect region and the adjacent region.

8. The method of claim 7, wherein the method further comprises: The final defect region is determined by the verification formula for the suspected defect region and the adjacent region, including: Based on the light intensity change curve, the light intensity change value is obtained; Based on the mean curve, the mean change value is obtained; The final defect region is determined by the verification formula, and the verification formula is as follows, ； wherein is the degree of defect, is the light intensity change value, is the mean change value.

9. A visual inspection system for inspecting the quality of a bicycle saddle, for implementing the method for inspecting the quality of a bicycle saddle according to any one of claims 1 to 8, characterized in that, Including: Image acquisition and preprocessing module: use a high-resolution camera to acquire a color image of a bicycle saddle, separate the red, green, and blue three color channels with the decorative hole as the center; normalize the channel intensity by the light intensity value, and identify the suspected defect region; Continuity and progression degree evaluation module: based on the suspected defect region, calculate the change of the red, green, and blue channel values, and determine the continuity and progression degree; Difference degree evaluation module: divide the saddle into detection and verification regions, compare the difference of the color channel values in the two regions, evaluate the difference degree, screen out the regions that exceed the preset threshold, and perform comparative analysis; Detection focus determination module: combine the mixing degree and the difference degree, draw the time curve of the light intensity, the mixing degree, and the difference degree, calculate the mean curve, and determine the final defect region by the verification formula.

Images