Automobile radiator surface corrosion degree evaluation method based on image processing

Abstract

The invention relates to the technical field of image data processing, in particular to an automobile radiator surface corrosion degree evaluation method based on image processing, and the method comprises the steps: obtaining a surface image of an automobile radiator, carrying out the graying processing, obtaining an initial image, carrying out the saliency detection of the initial image through employing an FT algorithm, and obtaining the surface corrosion degree of the automobile radiator. And before saliency detection, the texture degree and the horizontal correlation number of the pixel points are synthesized, and the probability that each pixel point is located in the support strip region is evaluated, so that the pixel points belonging to the support strip region are eliminated before saliency detection, and after elimination is completed, the corrosion region is identified based on the saliency detection result of the remaining pixel points. And evaluating the surface corrosion degree of the radiator through the size of the corrosion area. According to the invention, the accuracy of identifying the corrosion area can be improved, so that the accurate evaluation of the surface corrosion degree of the radiator can be realized.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and more specifically, to a method for evaluating the corrosion degree of a car radiator surface based on image processing. Background Art

[0002] With the global development of the automobile industry and the continuous increase in the number of vehicles owned, automobile radiators, as core components of the engine cooling system, are exposed to complex environmental conditions such as high temperature, high humidity, and salt spray erosion for a long time. Surface corrosion has become a key risk factor causing nonlinear attenuation of heat dissipation efficiency, coolant permeability leakage, and local engine overheating failures. Therefore, how to achieve accurate detection of radiator surface corrosion areas has become a technical problem that needs to be solved urgently.

[0003] With the rapid advancement of computer vision and artificial intelligence technologies, image processing technology has been able to handle complex detection tasks. Among the related technologies, the FT (Frequency-Tuned) algorithm based on frequency domain residual analysis can identify abnormal areas by analyzing the local contrast features of the image.

[0004] However, since the lateral equally spaced support strips (duct baffles) designed on the surface of the radiator to enhance structural stability and optimize aerodynamic characteristics present periodically arranged high-frequency texture features in the image space domain, the inherent frequency domain response of such functional structures will be incorrectly mapped into local contrast anomaly features during the residual calculation of the FT algorithm, thereby generating false positive interference signals with similar significance responses to the actual corrosion areas, resulting in misjudgment of the corrosion areas determined based on the significance detection results output by the algorithm, making it impossible to accurately assess the degree of corrosion on the radiator surface. Summary of the invention

[0005] In order to solve the problem that the detection result of the saliency detection of the radiator surface image based on the FT algorithm cannot accurately evaluate the corrosion degree of the radiator surface due to the presence of the support strip on the radiator surface, the present invention provides a method for evaluating the corrosion degree of the automobile radiator surface based on image processing. The method comprises: Acquire the surface image of the automobile radiator and perform grayscale processing to obtain an initial image; Performing saliency detection on the initial image using the FT algorithm, and removing pixels belonging to the support bar area before the saliency detection. After the removal is completed, identifying the corrosion area based on the saliency detection results of the remaining pixels, and evaluating the degree of corrosion on the radiator surface according to the size of the corrosion area; Among them, the method for identifying pixel points belonging to the support bar area includes: calculating the texture degree of each pixel point according to the gray value of each pixel point and the consistency of the gray values within the neighborhood range of each pixel point, and the texture degree represents the probability that the pixel point is in the support bar area; Taking each pixel point as the center, traversing horizontally to both sides, counting the number of pixel points whose gray value difference from the central pixel point is less than the first threshold, obtaining the horizontal correlation quantity of each pixel point, calculating the high-frequency texture coefficient of each pixel point, the high-frequency texture coefficient is positively correlated with both the texture degree and the horizontal correlation quantity, and defining the pixel points with the high-frequency texture coefficient greater than the set value as the pixel points in the support bar area.

[0006] The present invention utilizes the distribution characteristics of pixel points in the support bar area in the horizontal direction and the performance characteristics of gray values to evaluate the probability that each pixel point belongs to the pixel points in the support bar area, so as to accurately identify the pixel points belonging to the support bar area, and before using the FT algorithm to perform saliency detection on the obtained initial image, eliminating the pixel points belonging to the support bar area, which can avoid the FT algorithm identifying the support bar area as the corrosion area, thereby ensuring the accuracy of corrosion area identification and realizing the precise evaluation of the corrosion degree of the radiator surface.

[0007] Preferably, the high-frequency texture coefficient is also positively correlated with the credibility that each pixel point belongs to the support member area, and the method for obtaining the credibility includes: Taking each pixel point as the center, traversing vertically to both sides, and defining the pixel points whose gray value difference from the central pixel point is less than the second threshold as the same-type pixel points of the corresponding pixel point; Calculating the credibility that each pixel point belongs to the support member area, and the credibility is negatively correlated with the difference between the horizontal correlation quantity of the pixel point and the average horizontal correlation quantity of the same-type pixel points.

[0008] The present invention can avoid the influence of the presence of dirt on the radiator surface on the possibility of evaluating that each pixel point belongs to the support member area by calculating the credibility that each pixel point belongs to the support member area.

[0009] Preferably, the high-frequency texture coefficient satisfies the following relational expression: ; In the formula, is the high-frequency texture coefficient of the th pixel point; is the texture degree of the th pixel point; is the horizontal correlation quantity of the th pixel point; is the average horizontal correlation quantity of all the same-type pixel points of the th pixel point; is a preset hyperparameter; is the absolute value symbol; is the normalization function.

[0010] The present invention synthesizes data from multiple aspects, can accurately evaluate the possibility of each pixel point belonging to the support member area, thereby obtaining an accurate quantization index, that is, the high-frequency texture coefficient, providing a reference standard for the identification of pixel points within the subsequent support member area.

[0011] Preferably, when traversing to both sides along the vertical direction with each pixel point as the center, if a pixel point with a gray value difference from the central pixel point greater than the second threshold appears on either side, the traversal process on that side is stopped.

[0012] Preferably, the method for obtaining the consistency of gray values within the neighborhood range of each pixel point includes: Calculate the average difference between the gray value of each pixel point and the gray values of pixel points within the four-neighborhood range, and use the opposite number of the average difference as the power of the exponential function to perform the power operation to obtain the consistency of gray values within the neighborhood range of each pixel point.

[0013] The present invention calculates the consistency of gray values within the neighborhood range of each pixel point through a negative exponential function, which can effectively suppress the consistency of pixel points with large gray value differences within the neighborhood.

[0014] Preferably, calculating the texture degree of each pixel point satisfies the following relational expression: ; In the formula, is the texture degree of the th pixel point; is the gray value of the th pixel point; is the average value of the gray values of all pixel points in the initial image; is the consistency of gray values within the area around the th pixel point.

[0015] When the present invention measures the high or low of the gray value of each pixel point, it provides a reference standard, that is, the average value of the gray values of all pixel points in the initial image, which can effectively eliminate the influence of the overall brightness change of the image on the analysis result.

[0016] Preferably, when traversing to both sides along the horizontal direction with each pixel point as the center, if two pixel points with gray value differences from the central pixel point greater than the first threshold continuously appear on either side, the traversal process on that side is stopped.

[0017] Preferably, identifying the corrosion area based on the saliency detection results of the remaining pixel points: Obtain a preset significance threshold, screen all pixel points whose significance values are greater than the significance threshold, and use the connected domain composed of all the screened pixel points as the corrosion area.

[0018] Preferably, the corrosion degree of the radiator surface is evaluated by the size of the corrosion area, including: Calculate the ratio of the number of pixel points in the corrosion area to the number of pixel points in the initial image, and use the ratio as the evaluation value of the corrosion degree of the radiator surface.

[0019] Preferably, when graying the surface image of the automotive radiator, the weighted average method is used to convert the RGB channel values into gray values.

[0020] The present invention has the following effects: The present invention calculates the horizontal correlation quantity of each pixel point, can divide the boundary of the support member area into the support member area, and combines the feature that the gray values of the pixel points in the support member area are relatively high to identify the pixel points in the entire support member area, so that before performing significance detection on the initial image using the FT algorithm, all pixel points belonging to the support member area can be excluded, thereby eliminating the influence of the existence of the support member area on the corrosion area identified by the FT algorithm, ensuring the accuracy of the obtained corrosion area, and thus providing a reliable data basis for the evaluation of the corrosion degree of the radiator surface. Description of the Drawings

[0021] Figure 1 is a schematic flow chart of the steps of a method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to an embodiment of the present invention. Detailed Embodiments

[0022] The following will describe in detail the specific embodiments of the present invention with reference to the drawings.

[0023] Refer to Figure 1 , a method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing, includes steps S1 - S3, specifically as follows: S1: Obtain the surface image of the automotive radiator and perform graying processing to obtain an initial image.

[0024] Specifically, a high-definition digital camera can be used to collect an image of the surface of the automotive radiator under uniform illumination conditions to ensure that the image is clear and covers the entire surface of the radiator. Then, in order to simplify the subsequent processing process, the collected surface image can be grayed.

[0025] In an exemplary embodiment of the present invention, when graying the surface image of the automotive radiator, the weighted average method is used to convert the RGB channel values into gray values.

[0026] Optionally, the maximum value method, average value method, etc. may be used to convert the RGB channel value of each pixel in the surface image into a grayscale value, thereby converting the collected surface image into a grayscale image to obtain an initial image. This embodiment does not specifically limit the selected grayscale processing method.

[0027] S2: Perform saliency detection on the initial image using the FT algorithm, and remove pixels belonging to the support strip area before the saliency detection.

[0028] It should be noted that there are usually many horizontally arranged support bars between the cooling fins on the surface of the car radiator, also called air duct baffles. Their function is to enhance the structural stability of the radiator and optimize the air circulation path so that the air can pass through the cooling fins more evenly, thereby improving the heat dissipation efficiency.

[0029] However, when using the FT (Frequency-tuned) algorithm to detect the corrosion area on the surface of the car radiator, due to the large gray value difference between the support bar area and the surrounding pixels, it is characterized by large differences in high-frequency components and local contrast in the frequency domain. The FT algorithm identifies significant areas by calculating the spectral residual in the frequency domain, and these characteristics of the support bar area are easily misjudged as significant features, causing the algorithm to mistakenly mark it as a corrosion area, resulting in false detection. Therefore, before using the FT algorithm for significance detection, the present invention removes the pixels belonging to the support bar area to avoid false detection.

[0030] Specifically, the pixel points within the support strip area can be identified by the following steps: Step 1: Calculate the texture degree of each pixel based on the grayscale value of each pixel and the consistency of the grayscale values within the neighborhood of each pixel. The texture degree represents the probability that the pixel is in the support strip area.

[0031] It should be noted that, through scene analysis, it can be seen that the support bar area is usually made of metal material. Due to the high reflectivity of the metal material, the support bar area presents a higher grayscale value in the image; however, there may be noise pixels with abnormally high grayscale values in the initial image, which will interfere with the recognition of the support bar area; since the positions of the noise pixels are randomly distributed, and the support bar area is continuously distributed in strips, the grayscale values of its pixels in the local neighborhood are highly consistent.

[0032] Therefore, the present invention can evaluate the texture degree of each pixel belonging to the support strip area by combining the high gray value characteristics of the support strip area and the gray consistency characteristics of its neighborhood.

[0033] In an exemplary embodiment of the present invention, the consistency of the grayscale values within the neighborhood of each pixel point can be determined by the following steps: Calculate the average difference between the gray value of each pixel and the gray values of the pixels within the four-neighborhood range, and use the negative of the average difference as the power of the exponential function to perform the power operation to obtain the gray value consistency within the neighborhood range of each pixel.

[0034] Exemplarily, the consistency of the gray values within the area around the th pixel can be denoted as , then satisfies the following relational expression: ; In the formula, is the average difference between the gray value of the th pixel and the gray values of the pixels within the four-neighborhood range. Here, the difference is the absolute value of the numerical difference (the square of the numerical difference can also be used); is the exponential function with the natural constant as the base.

[0035] Optionally, the consistency of the gray values within the neighborhood range of each pixel can also be evaluated based on the variance of the gray values within the four-neighborhood or eight-neighborhood range of each pixel. It should be noted that the determination process of the four-neighborhood range and the eight-neighborhood range of the pixel is the prior art in the image field, and this embodiment will not elaborate on it here.

[0036] Furthermore, after determining the consistency of the gray values within the neighborhood range of each pixel, the texture degree of each pixel can be calculated. Specifically, the texture degree of each pixel satisfies the following relational expression: ; In the formula, is the texture degree of the th pixel; is the gray value of the th pixel; is the average value of the gray values of all pixels in the original image; is the consistency of the gray values within the area around the th pixel.

[0037] Among them, reflects the relative magnitude of the gray value of the th pixel in the entire original image. The larger this value is, the more likely it is that the pixel belongs to the support bar area (it may also be a noise pixel), and the corresponding texture degree of the pixel is larger. It should be noted that when evaluating the gray value level of each pixel in the present invention, the average gray value of the entire image is introduced, which can provide a global reference standard for the gray value evaluation, so as to objectively reflect the gray value level of each pixel.

[0038] The larger the value, the more consistent the gray values in the area around the pixel point, which further indicates that the pixel point is less likely to be a noise pixel point, that is, the pixel point has a greater possibility of belonging to the pixel points in the support bar area, and correspondingly, the texture degree of the pixel point is relatively large.

[0039] Step 2: Taking each pixel point as the center, traverse horizontally to both sides, and count the number of pixel points whose gray value difference from the central pixel point is less than the first threshold, so as to obtain the horizontal correlation number of each pixel point.

[0040] It should be noted that although the interference of noise pixel points has been considered when calculating the texture degree of each pixel point, pixel points located at the edge of the support bar may still be missed. This is because the gray value differences between the pixel points at the boundary of the support bar and the pixel points in the neighborhood are relatively large, resulting in these edge pixel points not being able to be classified into the support bar area. According to the scene research, the support bar area on the surface of the automotive radiator has a significant horizontal distribution feature. Therefore, the present invention utilizes this feature to further evaluate the probability of each pixel point belonging to the support bar area by analyzing the number of neighboring pixel points with relatively small gray value differences in the horizontal direction of each pixel point.

[0041] Exemplarily, taking the th pixel point as an example, the determination process of the horizontal correlation number of each pixel point will be described in detail: Taking the th pixel point as the center, expand horizontally from the position of the th pixel point to both sides. When the gray value difference (the absolute value of the numerical difference) between the expanded pixel point and the th pixel point is less than the first threshold, such as 3, count once until the preset expansion termination condition is reached (such as reaching the image boundary or the gray value differences of multiple consecutive pixel points exceeding the threshold), and take the final counting result as the horizontal correlation number of the th pixel point.

[0042] In an exemplary embodiment of the present invention, when traversing horizontally to the left and right sides with each pixel point as the center, if the gray value differences between any two consecutive pixel points and the central pixel point on either side are greater than the first threshold, the traversal process on that side is stopped.

[0043] It should be noted that the larger the horizontal correlation number of any pixel point, the more pixel points with relatively small gray value differences from the any pixel point in the horizontal direction, which further indicates that the gray value difference situation between the any pixel point and its neighboring pixel points in the horizontal direction is more in line with the support bar area, and the any pixel point has a greater possibility of belonging to the pixel points in the support bar area.

[0044] Step 3: Calculate the high-frequency texture coefficients of each pixel point. The high-frequency texture coefficients are positively correlated with both the texture degree and the horizontal correlation quantity.

[0045] Among them, the high-frequency texture coefficient is an index used to determine whether a pixel point belongs to the support bar region. It should be noted that since the gray values of the pixel points in the support bar region are relatively high, when using the FT algorithm to perform saliency detection on the initial image, the support bar region usually belongs to the high-frequency part in the image frequency domain. Therefore, this index is named the high-frequency texture coefficient here.

[0046] In an exemplary embodiment of the present invention, the high-frequency texture coefficient is also positively correlated with the credibility of each pixel point belonging to the support member region. The determination of the credibility of each pixel point belonging to the support member region can be achieved through the following steps: (1) Taking each pixel point as the center, traverse in both directions along the vertical direction, and define the pixel points with a gray value difference less than the second threshold from the central pixel point as the same-type pixel points of the corresponding pixel point; It should be noted that in practical applications, the dust and dirt attached to the surface of the radiator may cause a significant reduction in the number of pixel points with a small gray value difference from a pixel point in the horizontal direction, thus affecting the accurate recognition of the support bar region. To further solve this problem, this method introduces an analysis in the vertical direction. Specifically: If a pixel point has multiple same-type pixel points (i.e., pixel points with a small gray value difference from this pixel point) in the vertical direction, and the difference between the numbers of pixel points with a small gray value difference corresponding to these same-type pixel points in their respective horizontal directions is small, it indicates that the pixel point being analyzed and its same-type pixel points in the vertical direction are more likely to belong to the same support bar region, and further indicates that the credibility of the pixel point being analyzed belonging to the support bar region is greater.

[0047] Exemplarily, taking the th pixel point as an example, the determination process of the same-type pixel points of each pixel point will be described in detail: Taking the th pixel point as the center, expand in both the upper and lower directions along the vertical direction from the position of the th pixel point. When the expanded pixel point has a gray value difference (the absolute value of the numerical difference) less than the second threshold, such as 3, from the th pixel point, the expanded pixel point is used as the same-type pixel point of the th pixel point until the preset expansion termination condition is reached (such as reaching the image boundary or the gray difference from the central pixel point exceeding the threshold).

[0048] In an exemplary embodiment of the present invention, when traversing on both sides along the vertical direction with each pixel point as the center, if a pixel point with a gray value difference greater than the second threshold from the central pixel point appears on either side, the traversal process on that side is stopped.

[0049] (2) Calculate the credibility of each pixel point belonging to the support member area, and the credibility is negatively correlated with the difference between the horizontal association quantity of the pixel point and the average horizontal association quantity of the same type of pixel points.

[0050] Exemplarily, taking the th pixel point as an example, the calculation of the credibility of each pixel point belonging to the support member area is described in detail: First, determine the horizontal association quantity of the th pixel point, and the average value of the horizontal association quantities of all the same type of pixel points of the th pixel point; then, calculate the difference between the horizontal association quantity of the th pixel point and this average value (the difference here is the absolute value of the numerical difference); after that, the reciprocal of the obtained difference value can be used as the credibility of the th pixel point belonging to the support member area.

[0051] Optionally, other methods can also be used to characterize the negative correlation relationship between the difference value and the credibility here, such as representing the negative correlation relationship not in the form of taking the reciprocal, but through the negative natural exponential function.

[0052] Further, after determining the horizontal association quantity, texture degree, and credibility of each pixel point belonging to the support member area, the high-frequency texture coefficient of each pixel point can be calculated. Specifically, the high-frequency texture coefficient of each pixel point satisfies the following relational expression: ; In the formula, is the high-frequency texture coefficient of the th pixel point; is the texture degree of the th pixel point; is the horizontal association quantity of the th pixel point; is the average horizontal association quantity of all the same type of pixel points of the th pixel point; is a preset hyperparameter, and in this embodiment is used to prevent the denominator from being zero; is the absolute value symbol; is the normalization function.

[0053] Among them, reflects the The credibility that a pixel belongs to the support member area; the larger this value, the greater the likelihood that the pixel belongs to the pixels in the support member area, and the corresponding high-frequency texture coefficient is larger.

[0054] In another embodiment, it can also be through the relational expression: Calculate the high-frequency texture coefficient of each pixel; in the formula, is the exponential function with the natural constant as the base.

[0055] Step 4: Define the pixels with high-frequency texture coefficients greater than the set value as the pixels in the support bar area.

[0056] Optionally, the set value can be set to 0.5. If the high-frequency texture coefficient of any pixel is greater than 0.5, then determine that pixel as a pixel in the support member area and mark the pixel, so that after subsequent saliency analysis, such marked pixels are excluded.

[0057] S3: After the exclusion is completed, identify the corrosion area based on the saliency detection results of the remaining pixels, and evaluate the corrosion degree of the radiator surface through the size of the corrosion area.

[0058] In an exemplary embodiment of the present invention, the identification of the corrosion area can be achieved through the following steps: Obtain a preset saliency threshold, screen all pixels with saliency values greater than the saliency threshold, and use the connected domain composed of all the screened pixels as the corrosion area.

[0059] It should be noted that the process of using the FT algorithm to perform saliency detection on the image to obtain the normalized saliency values of each pixel is a prior art, and this embodiment will not elaborate on it here.

[0060] Optionally, the saliency threshold can be set to 0.8. If the saliency value of any remaining pixel after the exclusion operation is greater than 0.8, then determine that the position of that pixel is the corrosion area, so that all pixels with saliency values greater than the saliency threshold can be screened, and the connected domain composed of all the screened pixels is used as the corrosion area.

[0061] In an exemplary embodiment of the present invention, the evaluation of the corrosion degree of the radiator surface can be achieved through the following steps: Calculate the ratio of the number of pixels in the corrosion area to the number of pixels in the initial image, and use this ratio as the evaluation value of the corrosion degree of the radiator surface.

[0062] Specifically, the evaluation value satisfies the relational expression: ; in the formula, is the evaluation value of the corrosion degree of the radiator surface; is the number of pixel points in the corrosion area; is the number of pixel points in the initial image.

[0063] Optionally, the connectivity of the corrosion area can also be analyzed. For example, calculate the number of connected components in the corrosion area and the size of each connected component. If the corrosion area is dispersed and there are many connected components, it may indicate that the corrosion is relatively severe.

[0064] It should be understood that various alternative solutions to the embodiments of the present invention described herein may be adopted during the practice of the present invention.

Claims

1. An evaluation method for the corrosion degree of the surface of an automotive radiator based on image processing, characterized in that, include: Acquire the surface image of the automobile radiator and perform grayscale processing to obtain an initial image; Performing saliency detection on the initial image using the FT algorithm, and removing pixels belonging to the support bar area before the saliency detection. After the removal is completed, identifying the corrosion area based on the saliency detection results of the remaining pixels, and evaluating the degree of corrosion on the radiator surface according to the size of the corrosion area; The method for identifying pixels belonging to the support strip area includes: calculating the texture degree of each pixel according to the gray value of each pixel and the consistency of the gray value within the neighborhood of each pixel, wherein the texture degree represents the probability that the pixel is in the support strip area; Taking each pixel point as the center, traverse to both sides in the horizontal direction, count the number of pixels whose grayscale value difference with the central pixel point is less than the first threshold, obtain the horizontal association number of each pixel point, and calculate the high-frequency texture coefficient of each pixel point. The high-frequency texture coefficient is positively correlated with the texture degree and the horizontal association number. The pixel point with a high-frequency texture coefficient greater than the set value is defined as the pixel point of the support strip area.

2. The method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to claim 1, wherein, The high-frequency texture coefficient is also positively correlated with the credibility of each pixel belonging to the support area, and the method for obtaining the credibility includes: Taking each pixel point as the center, traverse to both sides along the vertical direction, and define the pixel points whose grayscale value difference with the central pixel point is less than the second threshold as the same type of pixel points as the corresponding pixel point; The credibility of each pixel belonging to the support area is calculated, and the credibility is negatively correlated with the difference between the horizontal association number of the pixel and the horizontal association number of the same type of pixels.

3. According to the method for evaluating the degree of corrosion of the surface of an automobile radiator based on image processing in claim 2, the high-frequency texture coefficient satisfies the following relationship: ； Wherein, is the high-frequency texture coefficient of the th pixel point; is the texture degree of the th pixel point; is the horizontal correlation quantity of the th pixel point; is the average horizontal correlation quantity of all similar pixel points of the th pixel point; is a preset hyperparameter; is the absolute value symbol; is the normalization function.

4. The method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to claim 2, characterized in that, When traversing toward both sides along the vertical direction with each pixel point as the center, if a pixel point appears on either side whose grayscale value difference with the central pixel point is greater than the second threshold, the traversal process on that side is stopped.

5. The method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to claim 1, wherein, The method for obtaining the consistency of the grayscale values within the neighborhood of each pixel point includes: The average difference between the grayscale value of each pixel and the grayscale values of the pixels in the four neighborhoods is calculated, and the inverse of the average difference is used as the power of the exponential function to perform power operations to obtain the consistency of the grayscale values in the neighborhood of each pixel.

6. The method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to claim 5, wherein, The calculation of the texture degree of each pixel satisfies the following relationship: ； Wherein, is the texture degree of the th pixel point; is the gray value of the th pixel point; is the average value of the gray values of all pixel points in the initial image; is the consistency of the gray values within the area around the th pixel point.

7. A method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to claim 1, characterized in that When traversing horizontally toward both sides with each pixel point as the center, if the grayscale value difference between two consecutive pixels on either side and the central pixel point is greater than the first threshold, the traversal process on that side is stopped.

8. A method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to claim 1, characterized in that, The saliency detection result based on the remaining pixels identifies the corrosion area: A preset significance threshold is obtained, all pixels whose significance values are greater than the significance threshold are screened, and a connected domain composed of all the screened pixels is used as a corrosion area.

9. A method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to claim 8, characterized in that, The evaluation of the degree of corrosion on the radiator surface by the size of the corrosion area includes: The ratio of the number of pixels in the corrosion area to the number of pixels in the initial image is calculated, and the ratio is used as an evaluation value of the degree of corrosion on the surface of the radiator.

10. A method for evaluating the corrosion degree of the surface of an automotive radiator based on image processing according to claim 1, characterized in that, When performing grayscale processing on the surface image of an automotive radiator, the weighted average method is used to convert the RGB channel values into grayscale values.

Images