An image processing-based heat pipe bending degree automatic measurement method and system

Abstract

The present application relates to the technical field of image processing, in particular to a heat dissipation pipe bending degree automatic measurement method and system based on image processing. The method comprises the following steps: obtaining multiple original images of a heat dissipation pipe to be measured, wherein the multiple original images have different exposure levels; calculating the fusion weight of each pixel point in each original image; fusing the multiple original images based on all fusion weights corresponding to all original images to obtain a fusion image; extracting the contour of the fusion image and calculating the bending degree of the heat dissipation pipe. That is, the scheme of the present application can accurately obtain the bending degree of the heat dissipation pipe.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing. More particularly, the present application relates to a heat dissipation pipe bending degree automatic measurement method and system based on image processing. BACKGROUND

[0002] In modern manufacturing industry, copper and aluminum heat dissipation pipes are the core components of heat exchange systems such as CPU heat sinks, automobile water tanks, air conditioners and household appliances. The bending degree of the heat dissipation pipe is a key quality indicator, which directly affects its assembly performance and heat dissipation efficiency.

[0003] Currently, the detection of the bending degree of the heat dissipation pipe in the industry mainly relies on manual caliper sampling inspection or traditional machine vision method. The former is inefficient and has poor consistency, and cannot meet the full inspection requirements. The latter also faces severe technical challenges when facing smooth metal pipes without surface spraying treatment.

[0004] The traditional method of extracting the profile of the heat dissipation pipe is to take a picture under a single light intensity, extract the profile of the picture, and then calculate the bending degree. However, when industrial light source shines on the surface of high-reflectivity metals such as copper and aluminum, it will form strong specular reflection, which appears as a large area of "highlight" region in the image. The gray value of the pixel points in these highlight regions will quickly reach the saturation state (i.e. overexposure), resulting in distortion of the detailed information in this region, especially the profile information of the pipe as the measurement reference, which will have more or less deviation.

[0005] Specifically, the profile image of the heat dissipation pipe under underexposure can be seen from Figure 1 , the profile image under underexposure can clearly present the image of the top (exposed part), but due to insufficient exposure, the identification effect of the remaining part is not good, i.e. there is a problem of missing the main body; the profile image of the heat dissipation pipe under normal exposure can be seen from Figure 2 , the profile image under normal exposure has distortion (wide) in the top part due to the existence of highlight, and part of the region has a broken situation; the profile image of the heat dissipation pipe under overexposure can be seen from Figure 3 , the profile image under overexposure can clearly identify the complete profile, but due to the overexposure, the identified image size is distorted (too large).

[0006] Therefore, when taking pictures of heat dissipation pipes, no matter how the light intensity is adjusted, the profile identification of a single exposure image will have errors, resulting in a large deviation in the result of the heat dissipation pipe bending degree automatic measurement method, and even cannot be measured. SUMMARY

[0007] The purpose of the present application is to provide an image processing-based heat pipe bending degree automatic measurement method and system to solve the problem of large error in calculating the bending degree of a heat pipe in the prior art. To this end, the present application provides a solution in the following two aspects.

[0008] In the first aspect, the present application provides an image processing-based heat pipe bending degree automatic measurement method, comprising:

[0009] Obtaining multiple original images of a heat pipe to be measured, and preprocessing each original image, wherein the multiple original images have different exposure levels;

[0010] Calculating the fusion weight of each pixel point in each original image;

[0011] Fusing the multiple original images based on all fusion weights corresponding to all original images to obtain a fused image; extracting the contour of the fused image and calculating the bending degree of the heat pipe;

[0012] The fusion weight is positively correlated with the exposure quality and edge potential of any pixel point in the corresponding original image;

[0013] The exposure quality is negatively correlated with the absolute value of the difference between the gray value of the corresponding pixel point and the ideal gray value; the ideal gray value is the gray value corresponding to the peak value of the gradient weighted information histogram;

[0014] The edge potential is the normalized value of the gradient of the highlight probability of the corresponding pixel point; the highlight probability is positively correlated with the gray value of the corresponding pixel point and negatively correlated with the set highlight activation threshold.

[0015] The above solution can effectively overcome the limitations of single exposure by obtaining multiple original images with different exposure levels and fusing the multiple original images based on fusion weights (considering exposure quality and edge potential), significantly improving the accuracy and reliability of the heat pipe bending degree automatic measurement method.

[0016] Optionally, the exposure quality is:

[0017] ;

[0018] wherein, is the exposure quality of the kth original image at position ; is the ideal gray value of the kth original image; is the standard deviation of the gray values of all pixel points in the kth original image, is the gray value of the pixel point of the kth original image at position .

[0019] The exposure quality acquisition can more accurately and smoothly evaluate the exposure level of the pixels, assign a high weight to the pixel points with moderate exposure, and be beneficial to subsequent generation of a high-quality fusion image.

[0020] Optionally, the fusion weight is:

[0021] ;

[0022] wherein, is the exposure quality of the kth original image at the position ; is the edge potential of the kth original image at the position , is an edge priority factor, is a very small positive number.

[0023] The fusion weight can optimize the final contour extraction effect.

[0024] Optionally, the highlight probability is:

[0025] ;

[0026] wherein, is the highlight probability of the kth original image at the position ; is the gray value of the pixel point of the kth original image at the position ; is a highlight activation threshold value; is a steepness factor.

[0027] The highlight probability is calculated in the form of a Sigmoid function, which smoothly maps the pixel gray value to a probability value between 0 and 1, can more stably and continuously describe the possibility of a pixel point belonging to a highlight region, and avoids the logical cliff problem caused by a hard threshold.

[0028] Optionally, the gradient of the highlight probability is:

[0029] ;

[0030] wherein, is the gradient of the highlight probability of the pixel point of the kth original image at the position in the highlight probability map; and are the horizontal direction gradient and the vertical direction gradient of the pixel point of the kth original image at the position in the highlight probability map; the highlight probability map is composed of the highlight probabilities of the pixel points in the corresponding original image.

[0031] Optionally, the gradient weighting information histogram comprises:

[0032] All pixel points of each original image are traversed, the gray scale gradient amplitude of each pixel point is calculated, the sum of the gradient amplitudes of the same gray scale is obtained, and the gradient weighting information histogram corresponding to all gray scales is obtained.

[0033] The gradient weighting information histogram can guide the calculation of exposure quality, and improve the adaptability and accuracy of the whole algorithm.

[0034] Optionally, the preprocessing of the original images comprises: performing denoising processing on the original images by using a bilateral filtering algorithm to obtain denoised original images.

[0035] Denoising the original images can significantly improve the quality of the original images, provide more reliable data sources for subsequent fusion weight calculation and image fusion, and thus improve the accuracy of the final measurement result.

[0036] Optionally, the fusion of the multiple original images based on all fusion weights corresponding to all original images comprises: performing fusion by using a Laplacian pyramid fusion algorithm.

[0037] Optionally, the contour of the fused image is extracted, and the bending degree of the heat dissipation pipe is calculated, comprising:

[0038] The contour of the heat dissipation pipe in the fused image is extracted by using a Canny edge detection algorithm

[0039] A coordinate set of all points on the contour is obtained.

[0040] Least square method is used for circle fitting on the coordinate set to obtain the radius of the fitted circle.

[0041] The radius of the fitted circle is used to calculate a curvature radius value, and the curvature radius value is taken as the bending degree of the heat dissipation pipe.

[0042] In a second aspect, a heat dissipation pipe bending degree automatic measurement system based on image processing comprises:

[0043] A processor;

[0044] A memory storing computer instructions for a heat dissipation pipe bending degree automatic measurement based on image processing, when the computer instructions are run by the processor, the system executes the heat dissipation pipe bending degree automatic measurement method based on image processing.

[0045] The present application has the following beneficial effects:

[0046] The scheme of the present application can effectively inhibit the overexposure influence of high light area and restore the edge information in the high light area from the profile image in underexposure, so as to realize the high-precision measurement of the bending degree of the high-reflective heat dissipation pipe. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 The profile image of the heat dissipation pipe in underexposure is schematically shown;

[0048] Figure 2 The profile image of the heat dissipation pipe in normal exposure is schematically shown;

[0049] Figure 3 The profile image of the heat dissipation pipe in overexposure is schematically shown;

[0050] Figure 4 The step flow chart of the heat dissipation pipe bending degree automatic measurement method based on image processing in the embodiment is schematically shown;

[0051] Figure 5 The profile image of the heat dissipation pipe in underexposure is schematically shown;

[0052] Figure 6 The standard gray scale histogram of the profile image of the heat dissipation pipe in underexposure is schematically shown;

[0053] Figure 7 The gradient weighted information histogram of the profile image of the heat dissipation pipe in underexposure is schematically shown;

[0054] Figure 8 The profile image of the heat dissipation pipe in normal exposure is schematically shown;

[0055] Figure 9 The standard gray scale histogram of the profile image of the heat dissipation pipe in normal exposure is schematically shown;

[0056] Figure 10 The gradient weighted information histogram of the profile image of the heat dissipation pipe in normal exposure is schematically shown;

[0057] Figure 11 The profile image of the heat dissipation pipe in overexposure is schematically shown;

[0058] Figure 12 The standard gray scale histogram of the profile image of the heat dissipation pipe in overexposure is schematically shown;

[0059] Figure 13A gradient weighted information histogram of a profile image of a heat pipe to be measured is illustratively shown when overexposed;

[0060] Figure 14 A structural block diagram of a heat pipe bending degree automatic measurement system based on image processing in the embodiment is illustratively shown. DETAILED DESCRIPTION

[0061] The technical solutions in the embodiments of the present application will be clearly and completely described in combination with the drawings in the embodiments of the present application.

[0062] The embodiment of the present application provides a heat pipe bending degree automatic measurement method based on image processing, aiming at solving the problems of uneven exposure in the image acquisition process, loss of details in high light areas and the like caused by the metal reflection characteristics of the surface of the heat pipe, so that the bending degree cannot be accurately measured due to the incomplete edge profile.

[0063] As shown in Figure 4 , the heat pipe bending degree automatic measurement method based on image processing in the embodiment includes the following steps:

[0064] Step S1: acquiring and pre-processing a plurality of original images of the heat pipe to be measured at different exposure levels.

[0065] In the embodiment, first, an image acquisition device such as an industrial camera is used to take multiple shots of the same heat pipe to be measured at a fixed position and angle, and different exposure parameters (such as exposure time, aperture or sensitivity) are set for each shot to obtain a group of original images with different exposure levels. The group of original images can ensure that each part of the heat pipe (including the normally exposed area, the underexposed area and the overexposed high light area) can be clearly presented in at least one image.

[0066] Further, in order to eliminate the noise introduced by the image acquisition device in the physical shooting process and obtain a clear image for subsequent analysis, it is also necessary to denoise and preprocess each original image collected.

[0067] Specifically, a bilateral filtering method is used to denoise each original image sequence collected to obtain a denoised original image.

[0068] The bilateral filter is a nonlinear filtering method that combines the spatial proximity and pixel value similarity of the image, and can smooth the image while maintaining the clarity of the edge, providing a high-quality image basis for the calculation of the fusion weight and the fusion of multiple original images.

[0069] Step S2: calculating the fusion weight of each pixel point in each original image.

[0070] In order to optimally combine the effective information of multiple original images, a fusion weight needs to be calculated for each pixel point in each original image. The fusion weight comprehensively considers the exposure quality and edge potential of the pixel point, so as to ensure that the pixel point with moderate exposure and located at the edge can be given a higher weight in the fused image.

[0071] Specifically, the fusion weight is positively correlated with the exposure quality and edge potential of any pixel point in the corresponding original image.

[0072] Specifically, the process of obtaining the fusion weight is as follows:

[0073] Step S21: Calculate the exposure quality of each pixel point in each original image.

[0074] The exposure quality in this embodiment is used to measure whether a pixel point is in a "good exposure" state.

[0075] Specifically, the process of obtaining the exposure quality is as follows:

[0076] First, the ideal gray value is obtained.

[0077] In this embodiment, the closer the gray value of a pixel point is to the ideal gray value of the entire image, the higher the exposure quality of the pixel point.

[0078] Specifically, the process of obtaining the ideal gray value is as follows: traverse all pixel points of the kth original image, calculate the gradient amplitude of each pixel point, and calculate the sum of the gradient amplitudes of the same gray level to obtain the gradient weighted information histogram corresponding to all gray levels, and take the gray level with the highest value in the gradient weighted information histogram (peak value) as the ideal gray value of the original image.

[0079] It should be noted that for the kth original image, the gray value of a pixel point is taken as a gray level, i.e., the range of the gray level is 1-254, and the gradient weighted information histogram of the gray level and the gradient amplitude is constructed, so that not only the frequency of each gray level is considered, but also the pixel point located at the edge is given a higher weight; then, the peak value (the information amount is the most) in the gradient weighted information histogram is found as the ideal gray value of the original image.

[0080] Exemplarily, Figure 5 , Figure 8 , Figure 11 are three original images collected by this embodiment, which are the profile images of the heat dissipation pipe under underexposure, normal exposure and overexposure respectively, and the standard gray histograms corresponding to the three original images are shown in Figure 6 , Figure 9 , Figure 12By processing contour images under different exposure levels, corresponding gradient-weighted information histograms can be obtained, thereby yielding ideal grayscale values; specifically, Figure 7 The ideal grayscale value for the underexposed contour image is 159. Figure 10 The ideal grayscale value for the outline image under normal exposure is 20. Figure 13 The ideal grayscale value for the contour image in an overexposed image is 1.

[0081] Secondly, the exposure quality is obtained based on the grayscale value of each pixel and the ideal grayscale value.

[0082] The exposure quality is negatively correlated with the absolute value of the difference between the gray value of the corresponding pixel and the ideal gray value.

[0083] Specifically, the exposure quality is:

[0084] ;

[0085] in, For the k-th original image at position Exposure quality; The ideal grayscale value for the k-th original image; is the standard deviation of the gray values ​​of all pixels in the k-th original image, used to control the width of the Gaussian function. For the k-th original image at position The grayscale value of the pixel.

[0086] From the above formula, it can be seen that when the gray value of a pixel... Compared to the ideal gray value When they are equal, the exposure quality Reaching the maximum value As the square of the difference between the two increases, the exposure quality rapidly decays according to a Gaussian distribution. This effectively assigns higher weight to pixels with proper exposure.

[0087] Step S22: Calculate the edge potential of each pixel in each original image.

[0088] In this embodiment, edge potential is used to characterize the likelihood that a pixel is located at a significant edge, especially at the boundary of a highlight region.

[0089] Since the transition from "well-exposed" to "over-exposed" is a gradual process, setting only one gray threshold to determine whether a pixel belongs to the highlight area will produce a logical cliff, which is also not consistent with the continuity of physical reality. Mapping the gray value to a probability value (highlight probability) to obtain a continuous highlight probability map can more smoothly describe the possibility of a pixel belonging to the highlight area. That is, the greater the gray value of each pixel, the greater the probability of belonging to the highlight area, and vice versa.

[0090] In this embodiment, the highlight probability is:

[0091] ;

[0092] wherein, is the highlight probability of the kth original image at position ; is the gray value of the pixel of the kth original image at position ; is the highlight activation threshold; is the steepness factor, which is used to control the transition steepness of the curve of the highlight probability.

[0093] The above formula is actually a Sigmoid function. When the gray value of the pixel is much smaller than , the highlight probability tends to ; when the gray value of the pixel exceeds , the highlight probability rapidly rises and tends to ; that is, the above highlight probability can reflect the change in the possibility of the transition of the pixel from the non-highlight area to the highlight area.

[0094] After the above calculation, the gray value of the pixel can be converted into a nonlinear and smooth probability value, thereby describing the possibility of the pixel belonging to the over-exposed area and ensuring that smooth and continuous probability output can be obtained regardless of the change in the brightness of the light irradiated on the original image.

[0095] wherein, the calculation method of the gradient of the highlight probability is:

[0096] ;

[0097] ;

[0098] ;

[0099] wherein, is the highlight probability of the kth original image at position the gradient of the highlight probability of the pixel point at position and respectively are the horizontal direction gradient and the vertical direction gradient of the pixel point at position of the kth original image in the highlight probability map, the highlight probability map is composed of the highlight probability of the pixel point in the corresponding original image, , , , respectively are the highlight probabilities of the pixel point at position , , , of the kth original image.

[0100] Since the internal region and the external region belonging to the highlight in the highlight probability map are flat (the highlight probability values are very close), the gradient at the position is 0; only at the boundary, the change rate is the largest, and the gradient is also the largest; therefore, by calculating the gradient condition of the highlight probability map, the edge potential of each pixel point of each original image is obtained, which provides a basis for subsequent calculation of the fusion weight.

[0101] Specifically, the edge potential is the normalized value of the gradient of the highlight probability of the corresponding pixel point. When normalizing the gradient of the highlight probability, the maximum and minimum value normalization method is adopted.

[0102] Step S23, according to the exposure quality and the edge potential, the fusion weight of the corresponding pixel point is obtained.

[0103] Specifically, the fusion weight is:

[0104] ;

[0105] wherein, is the exposure quality of the kth original image at position ; is the edge potential of the kth original image at position , is an edge priority factor, is a very small positive number.

[0106] In this embodiment, the value of the edge priority factor is preferably . When the value of is low, the fusion result is more focused on the well-exposed region; when the value of is high, the edge details of the highlight region are more retained. For example, the is set to The profile of the heat dissipation pipe can be clearly reconstructed while effectively suppressing high light. The value of the weight is 10 -12 to avoid the weight being zero.

[0107] In step S3, the multiple original images are fused based on all the fusion weights corresponding to all the original images to obtain a fused image; the profile of the fused image is extracted, and the bending degree of the heat dissipation pipe is calculated.

[0108] In this embodiment, the multiple original images are fused based on all the fusion weights corresponding to all the original images to obtain a fused image, in combination with a Laplacian pyramid fusion algorithm; the specific process is as follows:

[0109] A Laplacian pyramid is constructed for each original image; a fusion weight map corresponding to each original image is obtained A Gaussian pyramid is constructed; in each layer of the pyramid, the corresponding layer of the Laplacian pyramid is weighted and averaged using the weight of the Gaussian pyramid to generate a fused Laplacian pyramid; the fused Laplacian pyramid is inversely transformed and reconstructed to obtain a final fused image.

[0110] The Laplacian pyramid fusion algorithm can decompose and fuse the original images in multiple scales, effectively avoiding the blocking effect or artifacts generated by direct weighted averaging, so that the fusion result is more natural and smooth.

[0111] Since the Laplacian pyramid fusion algorithm is prior art, it will not be described here.

[0112] The bending degree is obtained as follows:

[0113] The Canny edge detection algorithm is used to extract the profile of the heat dissipation pipe in the fused image;

[0114] A coordinate set of all points on the profile is obtained;

[0115] The least square method is used to fit a circle to the coordinate set to obtain the radius of the fitted circle;

[0116] The radius of the fitted circle is used to calculate the curvature radius value, which is taken as the bending degree of the heat dissipation pipe.

[0117] In industrial measurement, the curvature radius is usually used to represent the bending degree of an object; the greater the radius, the smaller the bending degree; the smaller the radius, the greater the bending degree.

[0118] The scheme of the present application effectively overcomes the influence of high light on the image quality of the metal surface through the multi-exposure fusion technology, generates a globally clear fusion image, and on this basis, through accurate edge extraction and geometric fitting, measures the bending degree of the heat dissipation pipe with high precision and automation.

[0119] The present application also provides a heat dissipation pipe bending degree automatic measurement system based on image processing. Figure 14 As shown in the figure, the system comprises a processor and a memory, and the memory stores computer program instructions, which realize the heat dissipation pipe bending degree automatic measurement method based on image processing according to the present application when executed by the processor.

[0120] The system also comprises a communication bus and a communication interface and other components familiar to those skilled in the art, the settings and functions of which are known in the art, and thus will not be described here.

[0121] In the present application, the aforementioned memory can be any tangible medium containing or storing a program, which can be used or combined with an instruction execution system, device or instrument. For example, the computer readable storage medium can be any appropriate magnetic storage medium or magneto-optical storage medium, such as resistive random access memory (RRAM), dynamic random access memory (DRAM), static random access memory (SRAM), enhanced dynamic random access memory (EDRAM), high bandwidth memory (HBM), hybrid memory cube (HMC) and the like, or any other medium that can be used to store the required information and can be accessed by an application, module or both. Any such computer storage medium can be part of the device or accessible or connectable to the device. Any application or module described in the present application can be implemented by computer readable / executable instructions stored or otherwise held by such computer readable medium.

[0122] In the description of the present application, the meaning of "a plurality of" is at least two, such as two, three or more, etc., unless otherwise explicitly specified.

[0123] While the specification has illustrated and described various embodiments of the application, it will be readily apparent to those of ordinary skill in the art that many modifications, changes and substitutions can be made without departing from the spirit and scope of the application.

Claims

1. An image processing-based automatic measurement method for the bending degree of a heat pipe, characterized in that, The application relates to an image processing-based automatic measurement method for the bending degree of a heat dissipation pipe. The method comprises the following steps: acquiring multiple original images of the heat dissipation pipe to be measured, and pre-processing each original image, wherein the multiple original images have different exposure levels; fusing the multiple original images based on all fusion weights corresponding to all original images to obtain a fusion image; extracting the contour of the fusion image and calculating the bending degree of the heat dissipation pipe. The fusion weight of each pixel point in each original image is calculated, and the fusion weight is: , is the exposure quality of the kth original image at the position , is the edge potential of the kth original image at the position ; is an edge priority factor, and the value is 0.6-0.9, wherein setting the value of to 0.8 can effectively suppress the highlight while clearly reconstructing the outline of the heat dissipation pipe; is a very small positive number, and the value is 10 -12 ; The fusion weight is positively correlated with the exposure quality and edge potential of any pixel point in the corresponding original image. The exposure quality is negatively correlated with the absolute value of the difference between the gray value of the corresponding pixel point and an ideal gray value. The ideal gray value is the gray value corresponding to the peak value of a gradient weighted information histogram. The edge potential is the normalized value of the gradient of the highlight probability of the corresponding pixel point.

2. The method according to claim 1, wherein, The exposure quality is: ； wherein, is the exposure quality of the kth original image at position ; is the ideal gray value of the kth original image; is the standard deviation of the gray values of all pixel points in the kth image, is the gray value of the pixel point of the kth original image at position .

3. The method of claim 1, wherein the method further comprises: The highlight probability is: ； wherein, is the highlight probability of the kth original image at position ; is the pixel value of the kth original image at position ; is the highlight activation threshold; is the steepness factor.

4. The method of claim 1, wherein the method is characterized by: The gradient of the highlight probability is: ； wherein, is the gradient of highlight probability of the pixel point at position in the highlight probability map for the kth original image; and are the horizontal and vertical gradients of the pixel point at position in the highlight probability map for the kth original image, respectively; the highlight probability map is composed of the highlight probabilities of the pixel points in the corresponding original image.

5. The method of claim 1, wherein the method further comprises: The gradient weighted information histogram comprises: The method comprises the following steps: traversing all pixel points of each original image, calculating the gray gradient amplitude of each pixel point, obtaining the sum of the gradient amplitudes of the same gray level, and obtaining the gradient weighted information histogram corresponding to all gray levels, and taking the peak value in the gradient weighted information histogram as the ideal gray value of the original image.

6. The method of claim 1, wherein the method further comprises: The pre-processing of each original image comprises the following steps: adopting a bilateral filtering algorithm to perform denoising processing on the original image to obtain a denoised original image.

7. The method of claim 1, wherein the method further comprises: The fusion of the multiple original images based on all fusion weights corresponding to all original images comprises the following steps: adopting a Laplacian pyramid fusion algorithm to perform fusion.

8. The method according to claim 7, wherein, The method comprises the following steps: A Canny edge detection algorithm is adopted to extract the contour of the heat dissipation pipe in the fusion image. The coordinate set of all points on the contour is obtained. The least square method is used for circle fitting on the coordinate set to obtain the radius of the fitted circle. The curvature radius value is calculated based on the radius of the fitted circle, and the curvature radius value is taken as the bending degree of the heat dissipation pipe.

9. An image processing-based automatic measurement system for the bending degree of a heat pipe, characterized by, The application further discloses a system for image processing-based automatic measurement of the bending degree of a heat dissipation pipe. The system comprises a processor and a memory. The memory stores computer instructions for image processing-based automatic measurement of the bending degree of a heat dissipation pipe, and when the computer instructions are run by the processor, the system executes the image processing-based automatic measurement method for the bending degree of a heat dissipation pipe according to any one of claims 1-8.

Images