Intelligent detection method for air conditioner compressor accessories

Abstract

The invention relates to the technical field of nondestructive testing, and discloses an intelligent detection method for an air conditioner compressor accessory. The method comprises the following steps of: applying pulse thermal excitation to an air conditioner compressor accessory; continuously acquiring a thermal image sequence by utilizing infrared thermal imaging; establishing a background temperature reference map through multi-frame average; calculating temperature increment based on background temperature and forming thermal diffusion time sequence data; and calculating the change rate in a crack-free reference region to obtain a reference diffusion rate parameter, carrying out normalization evaluation on the diffusion change of the whole image, extracting crack response pixels, and outputting a crack binary mask by combining pixel connectivity and an area threshold value, thereby depicting thermal diffusion evolution and highlighting crack subtle abnormality while inhibiting environmental noise and instrument drift, and further improving the accuracy of the thermal diffusion. Compared with a traditional method, the method is more beneficial to automatic, accurate and reliable detection of the cracks of the air conditioner compressor accessories.

Description

Technical Field

[0001] This invention relates to the field of non-destructive testing technology, specifically to an intelligent testing method for air conditioner compressor parts. Background Technology

[0002] Currently, quality inspection of air conditioner compressor parts mainly relies on visual inspection, random destructive testing, and traditional non-destructive testing methods such as ultrasound and X-rays. Visual inspection can only detect large surface defects and has virtually no ability to detect tiny, hidden microcracks. Furthermore, it is highly dependent on operator experience, resulting in subjective and poor repeatability of test results. While ultrasound and X-ray methods can detect internal defects to some extent, they are poorly suited for compressor parts with complex shapes and uneven wall thicknesses. They also face challenges such as difficulty in applying coupling agent, limited probe placement, and inaccurate coverage of the target area, leading to low inspection efficiency and high equipment costs, making them unsuitable for online or batch automated inspection.

[0003] With the development of infrared thermal imaging technology, some existing technologies have begun to introduce thermal excitation into metal parts, inferring internal defects by observing changes in the surface temperature field. However, these approaches mostly remain at the qualitative or semi-quantitative analysis stage, typically involving thresholding, contrast enhancement, or simple grayscale statistics on single or a few frames of thermal images. They lack systematic modeling of the thermal diffusion process and fail to analyze diffusion differences across different time periods from a temporal perspective. Existing methods often directly use the absolute temperature or simple temperature rise at a specific moment as a criterion, without effectively eliminating the inhomogeneity of the initial temperature field. This leads to environmental fluctuations and differences in the initial state of the workpiece significantly affecting the detection results, resulting in insufficient stability. On the other hand, existing thermal imaging-based detection focuses primarily on the "bright and dark areas" of the temperature field, without constructing a unified, quantifiable diffusion change index to characterize the difference between defect-free and suspicious areas. They typically rely on manual experience to set thresholds or subjective interpretation by observing pseudo-color images, making it difficult to establish standardized judgment criteria. In addition, the publicly available image processing methods mostly focus on general image algorithms such as spatial filtering and edge detection, and lack a processing flow that combines the thermal conductivity of metals and uses physical quantities such as gradient and diffusion rate for anomaly identification. They are not sensitive enough to the subtle diffusion anomalies caused by microcracks, and are prone to missed detection or misjudgment.

[0004] Therefore, this case aims to propose an intelligent detection method for air conditioner compressor parts. First, a background temperature model is constructed to eliminate environmental influences. Then, the thermal diffusion behavior is accurately characterized using temperature increment sequences and three-dimensional diffusion volumes. Next, two-dimensional temperature gradient and intensity images are obtained using discrete difference methods, and the diffusion rate between keyframes is quantitatively calculated to form a diffusion change rate image. Subsequently, a baseline model is established using a crack-free reference area, the relative abnormal diffusion ratio is calculated, and crack response pixels are extracted through statistical discrimination. Finally, connected region analysis is completed based on the relationship between adjacent pixels, and a binary crack mask is output. Summary of the Invention

[0005] This invention provides an intelligent detection method for air conditioner compressor parts, which helps to solve the problems mentioned in the background art.

[0006] This invention provides the following technical solution: an intelligent detection method for air conditioner compressor parts, comprising: Pulsed thermal excitation is applied to the air conditioner compressor parts, and an infrared thermal imager is used to continuously acquire thermal image sequences covering the entire detection process to obtain temperature data of each pixel over time. Several reference frames in the steady-state phase are selected from the thermal imaging sequence, and the temperature of the same pixel in each reference frame is arithmetically averaged to obtain the background temperature reference image. Based on the background temperature reference image and thermal image sequence, the temperature increment of the current temperature of each pixel in each frame thermal image relative to the average background temperature is calculated, and a temperature increment map sequence and the corresponding three-dimensional temperature diffusion data volume are constructed. For the temperature increment image sequence, discrete difference operations are performed on the temperature increment of each pixel in both the row and column directions to obtain the pixel's two-dimensional temperature gradient vector and gradient magnitude, and to form a heat diffusion intensity image sequence. In the thermal diffusion intensity image sequence, select the first key frame and the second key frame with a predetermined time interval. Calculate the diffusion change rate image matrix based on the difference in thermal diffusion intensity of the same pixel in the first key frame and the time interval between the key frames. The diffusion rate of each pixel was collected within the crack-free reference area of ​​the air conditioner compressor components. The diffusion rate and absolute value of the diffusion rate within the crack-free reference area were statistically analyzed to construct the baseline diffusion rate parameters of the crack-free reference area. Based on the diffusion change rate of each pixel and the baseline diffusion rate parameter of the crack-free reference area, the relative abnormal diffusion ratio of all pixels in the image is calculated, and the set of crack response pixels that meet the abnormal conditions is extracted by statistical discrimination. Based on the crack response pixel set, the crack connected sub-regions are divided according to the relationship between adjacent pixels. Connected sub-regions with an area lower than the minimum recognition area threshold are filtered out, and a crack binary mask image is output.

[0007] Optionally, the step of applying pulsed thermal excitation to the air conditioner compressor components and continuously acquiring thermal image sequences covering the entire detection process using an infrared thermal imager to obtain temperature data of each pixel changing over time specifically includes: Place the air conditioner compressor parts flat in the central area of ​​the thermal imaging platform, so that the surface being inspected faces the imaging direction of the infrared thermal imager. Start the infrared thermal imager, set the sampling frequency and total sampling duration, obtain the total number of thermal image frames to be acquired within the entire sampling duration by multiplying the two, and continuously acquire thermal images according to the calculated total number of frames. In the initial frame at the start of sampling, a pulsed heat source is applied to the upper surface of the air conditioning compressor components for a predetermined pulse duration; For each thermal image frame acquired, the temperature values ​​of each pixel are organized into a two-dimensional temperature matrix according to the arrangement of the row and column pixels of the image, so that each thermal image frame corresponds to a temperature matrix composed of the total number of row pixels and the total number of column pixels. A discrete coordinate system is established on the thermal image plane. The center of the pixel at the top left corner of the image is set as the origin. The horizontal direction to the right is set as the positive direction of the column axis, and the vertical direction downward is set as the positive direction of the row axis. The pixel row index is specified to be between zero and the total number of pixels in the row minus one, and the pixel column index is specified to be between zero and the total number of pixels in the column minus one, so as to uniquely identify the position of each pixel.

[0008] Optionally, the step of selecting several reference frames in the steady-state phase of the thermal imaging sequence and performing an arithmetic mean on the temperature of the same pixel in each reference frame to obtain a background temperature reference image specifically includes: The number of reference frames is selected based on the total number of sampled frames. The number of reference frames is greater than or equal to one and less than the total number of sampled frames minus two. For each pixel location in the thermal image sequence, the temperature value of each pixel in all reference frames is obtained from the temperature matrix of each frame corresponding to the selected reference frame. The arithmetic mean of these temperature values ​​is then performed to obtain the background average temperature of each pixel. Based on the average background temperature of each pixel, a background temperature reference image with the same size as a single-frame thermal image is constructed, so that the background temperature reference image stores the corresponding average background temperature at each pixel.

[0009] Optionally, the step of calculating the temperature increment of each pixel in each frame of the thermal image relative to the average background temperature based on the background temperature reference image and the thermal image sequence, and constructing a temperature increment map sequence and the corresponding three-dimensional temperature diffusion data volume, specifically includes: In the thermal imaging sequence, starting from a frame after the reference frame until the end of sampling, the temperature matrix of each frame is selected sequentially. For each pixel position in the selected frame temperature matrix, obtain the pixel's temperature value in the current frame and the average background temperature in the background temperature reference image. Subtract the average background temperature from the current frame temperature to obtain the pixel's temperature increment in the current frame. Based on the temperature increment of all pixels in each frame, a temperature increment map with the same size as the single-frame thermal image is constructed. A temperature increment map sequence is constructed, and the temperature increment map sequence is combined into a three-dimensional temperature diffusion data volume in chronological order.

[0010] Optionally, for the temperature increment image sequence, discrete difference operations are performed on the temperature increment of each pixel in both the row and column directions to obtain the pixel's two-dimensional temperature gradient vector and gradient magnitude, forming a heat diffusion intensity image sequence, specifically including: For any frame in the temperature increment map sequence, for each row of pixels except the top row and the bottom row, obtain the temperature increment of each pixel at the position of the adjacent row above and the adjacent row below, and take half of the difference between the two as the row direction temperature difference component; for pixels located in the top row and the bottom row, set the row direction temperature difference component to zero. For any frame in the temperature increment map sequence, for each column of pixels except the leftmost and rightmost columns, obtain the temperature increment of each pixel at the position of the left adjacent column and the right adjacent column respectively, and take half of the difference between the two as the column direction temperature difference component; for pixels located in the leftmost and rightmost columns, set the column direction temperature difference component to zero. In the current frame, for each pixel, the row-direction temperature difference component and the column-direction temperature difference component are combined to form a two-dimensional temperature gradient vector for the pixel. In the current frame, for each pixel, the square root of the sum of the squares of the row-direction temperature difference components and the squares of the column-direction temperature difference components is obtained to obtain the gradient magnitude of the pixel. The gradient magnitudes of each pixel are then organized into a heat diffusion intensity image with the same size as the single-frame thermal image according to the image position. By repeating the row-direction difference, column-direction difference, and pixel two-dimensional temperature gradient vector construction and gradient magnitude calculation for each frame, a heat diffusion intensity image sequence is formed.

[0011] Optionally, selecting a first keyframe and a second keyframe with a predetermined time interval from the thermal diffusion intensity image sequence, and calculating the diffusion change rate image matrix based on the difference in thermal diffusion intensity of the same pixel in the first and second keyframes and the time interval between the keyframes, specifically includes: In the heat diffusion intensity image sequence, the first key frame and the second key frame at different moments in the temperature diffusion process are selected. The second key frame is later than the first key frame in time. The time interval between the two key frames is calculated based on the sampling frequency and the difference in frame number between the two frames. For each pixel position, obtain the gradient magnitude of each pixel in the first keyframe and the gradient magnitude in the second keyframe, and divide the difference between the two by the time interval between the two keyframes to obtain the diffusion rate of each pixel. Based on the diffusion rate of each pixel, a diffusion rate image matrix with the same size as a single-frame thermal image is constructed across the entire image area.

[0012] Optionally, the step of collecting the diffusion change rate of each pixel within the crack-free reference area of ​​the air conditioner compressor components, statistically analyzing the diffusion change rate and its absolute value within the crack-free reference area, and constructing a baseline diffusion rate parameter for the crack-free reference area specifically includes: A crack-free reference area is pre-defined within the crack-free area on the surface of the air conditioner compressor parts, and at least one pixel with a non-zero diffusion rate is confirmed in the diffusion rate image matrix within the crack-free reference area. The positions of all pixels within the crack-free reference region are used to form a pixel index set, and the number of pixels in the set is used as the area of ​​the crack-free reference region. The average diffusion rate of the crack-free reference region is obtained by taking the arithmetic mean of the diffusion rate of all pixels in the crack-free reference region. The arithmetic mean of the absolute values ​​of the diffusion rate of all pixels in the crack-free reference region is calculated to obtain the average absolute value of the diffusion rate of the crack-free reference region. The average diffusion rate and the average absolute value of the diffusion rate together constitute the reference diffusion rate parameter of the crack-free reference region.

[0013] Optionally, the step of calculating the relative abnormal diffusion ratio of all pixels in the image based on the diffusion change rate of each pixel and the baseline diffusion rate parameter of the crack-free reference region, and extracting the set of crack response pixels that meet the abnormal conditions through statistical discrimination, specifically includes: For each pixel in the diffusion rate image matrix, obtain the pixel's diffusion rate, the average diffusion rate of the crack-free reference area, and the average absolute value of the diffusion rate. Subtract the average diffusion rate from the pixel's diffusion rate and then divide by the average absolute value of the diffusion rate to obtain the pixel's relative abnormal diffusion ratio. Construct a relative anomalous diffusion ratio image matrix with the same size as the single-frame thermal image based on the relative anomalous diffusion ratio of each pixel; In the relative anomalous diffusion ratio image matrix, the arithmetic mean of the relative anomalous diffusion ratio of all pixels is calculated to obtain the mean of the whole image. The square root of the average of the squares of the differences between the relative anomalous diffusion ratio of each pixel and the mean of the whole image is then used to obtain the standard deviation of the whole image. The sum of the mean of the entire image and three times the standard deviation is set as the abnormal diffusion detection threshold. The threshold is compared for each pixel in the image matrix with the relative abnormal diffusion ratio. The pixel positions with the relative abnormal diffusion ratio greater than the abnormal diffusion detection threshold are summarized into the crack response pixel set.

[0014] Optionally, the step of dividing the crack-response pixel set into crack-connected sub-regions according to the relationship between adjacent pixels, filtering out connected sub-regions with areas lower than the minimum recognition area threshold, and outputting a crack binary mask image specifically includes: In the crack response pixel set, two pixels are considered to be adjacent pixels when the sum of the absolute values ​​of the difference between their row coordinates and the absolute values ​​of their column coordinates is equal to one. Based on the relationship between adjacent pixels, a connectivity analysis is performed on the crack response pixel set, and the interconnected pixels are divided into several crack connected sub-regions. The number of connected sub-regions is recorded as the total number of connected sub-regions. When the crack response pixel set is empty, the total number of connected sub-regions is set to zero, and no crack connected sub-regions are generated. For each crack-connected sub-region, count the number of pixels it contains, and use that as the area of ​​the corresponding crack-connected sub-region. Set the minimum recognition area threshold to one pixel, retain all crack connected sub-regions with an area greater than the minimum recognition area threshold, and remove crack connected sub-regions with an area less than or equal to the minimum recognition area threshold. On an image plane with the same size as a single-frame thermal image, pixels belonging to the preserved crack connected sub-regions are assigned a value of one, and the remaining pixels are assigned a value of zero, forming a crack binary mask image, and the crack binary mask image is output.

[0015] The present invention has the following beneficial effects: 1. Replacing traditional constant or continuous heating methods with short-pulse heat source irradiation, combined with high-frequency infrared imaging, achieves highly time-sequential and controllable thermal excitation and imaging acquisition. On the one hand, the pulsed heat source can induce local thermal response differences at minute defects within the material in a very short time, enhancing the temperature contrast of cracks; on the other hand, by flexibly combining the sampling frequency and total duration, both temporal resolution and adaptive adjustment based on component size and thermal inertia are ensured, balancing acquisition efficiency and data integrity. Compared with existing technologies such as simple steady-state thermal imaging or slow heating methods, pulse excitation amplifies defect responses in a short time, reduces overall detection time, mitigates the risk of thermal damage, and improves detection sensitivity and field applicability.

[0016] 2. By selecting multiple reference frames from the steady-state phase of the thermal imaging sequence and performing an arithmetic mean on the same pixel, a background temperature baseline image is generated. This effectively eliminates random noise caused by environmental fluctuations and device drift during thermal imaging, making subsequent temperature increment calculations more accurate and reliable. Compared with existing methods that generally only use single frames or moving averages, this scheme specifies the range of reference frames, ensuring that the reference set can fully cover the steady state while avoiding boundary effects; furthermore, pixel-level averaging eliminates local outliers, constructing a more stable base address model, providing a solid foundation for subsequent extraction of minute temperature differences.

[0017] 3. Based on the background reference image, this scheme calculates the temperature increment of each pixel frame by frame, forming a temperature increment map sequence, and integrates it into a three-dimensional temperature diffusion data volume. By combining the temperature difference information in the time dimension with the spatial location, a three-dimensional diffusion volume is constructed, comprehensively describing the dynamic process from excitation to diffusion. This method overcomes the limitation of traditional two-dimensional thermal images that only reflect static heat distribution, enabling the amplification and identification of weak signals from internal cracks during the diffusion process. Compared with existing techniques that only utilize two-dimensional image differences, this scheme can capture the differences in heat diffusion path and velocity, identify cracks from the overall diffusion trend level, and provide richer data support for quantitative analysis.

[0018] 4. By performing discrete difference operations on the temperature increment image sequence in both row and column directions, the two-dimensional temperature gradient vector and gradient magnitude of each pixel are obtained, and a sequence of heat diffusion intensity images is generated accordingly. Utilizing gradient information to extract the diffusion direction and intensity can intuitively reflect the concentrated heat flow areas at crack tips and defect edges. The algorithm maintains the stability of image boundaries by using zero difference values ​​for different boundary pixels. Compared with existing methods that directly utilize temperature differences or threshold segmentation, the gradient method is more robust to noise, can highlight the detailed features of heat diffusion changes, and improves the accuracy of crack location.

[0019] 5. Select the first and second keyframes from the thermal diffusion intensity sequence, and calculate the diffusion rate change image matrix based on the difference in thermal diffusion intensity and the time interval between the two frames. This accurately measures the local change in diffusion rate at the crack, avoiding redundant calculations for the entire sequence. The keyframe interval can be adaptively adjusted based on material thermal inertia and crack size, balancing detection sensitivity and computational efficiency. Compared to existing methods that traverse all frames one by one, this approach focuses on the most informative moment, reducing computational load and improving the efficiency of extracting crack dynamic information, making it easier for real-time online detection.

[0020] 6. Within a predefined crack-free rectangular region of the component, the diffusion change rate and its absolute average are statistically analyzed to construct a baseline diffusion rate parameter. By using a defect-free region as a healthy control, background diffusion differences caused by variations in the overall workpiece material or manufacturing process are eliminated, ensuring that the subsequent anomaly detection threshold is more targeted. Furthermore, by simultaneously considering both the average and absolute average values, a dual reference for diffusion directionality differences is achieved. Compared to the commonly used global threshold setting in existing technologies, the local adaptive baseline parameter of this scheme greatly improves the ability to distinguish between micro-crack diffusion anomalies.

[0021] 7. Based on the difference between the diffusion change rate per pixel and the baseline parameter, and after normalization, the relative abnormal diffusion ratio is calculated. Pixels meeting the abnormal conditions are then extracted as the crack response pixel set through statistical discrimination. Absolute diffusion anomalies are transformed into a standardized ratio metric, eliminating the influence of the original diffusion rate dimension. Furthermore, a dynamic threshold is established using the mean and standard deviation of the entire image, allowing the extraction process to adaptively adjust for different workpiece batches and environmental conditions. Compared to fixed thresholds or single statistical indicators, this scheme is more flexible and can accurately locate the crack response region.

[0022] 8. The crack response pixel set is divided into connectivity segments based on the relationship between adjacent pixels, and regions smaller than the minimum recognition area threshold are removed. The final output is a binary crack mask image. This method accurately reconstructs the crack connectivity morphology by combining pixel geometric adjacency rules, eliminating noise and isolated pixels. Simultaneously, the minimum recognition area threshold can be flexibly set to distinguish between microcracks and occasional abnormal signals. Compared to traditional morphological filtering or manual post-processing, this scheme's connectivity analysis is fully automated, maintaining the integrity of the crack morphology while quickly removing interference regions, thus improving the stability and visualization of the detection results. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the process of the present invention.

[0024] Figure 2 This is a schematic diagram of the coordinate system of the present invention.

[0025] Figure 3 This is a schematic diagram of the structure of the air conditioner compressor component of the present invention.

[0026] In the diagram: 1-Image, 2-Origin, 3-Column axis, 4-Row axis. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Example, refer to Figure 1 and Figure 3 A method for intelligent detection of air conditioner compressor parts, comprising: Pulsed thermal excitation is applied to the air conditioner compressor parts, and an infrared thermal imager is used to continuously acquire thermal image sequences covering the entire detection process to obtain temperature data of each pixel over time. Several reference frames in the steady-state phase are selected from the thermal imaging sequence, and the temperature of the same pixel in each reference frame is arithmetically averaged to obtain the background temperature reference image. Based on the background temperature reference image and thermal image sequence, the temperature increment of the current temperature of each pixel in each frame thermal image relative to the average background temperature is calculated, and a temperature increment map sequence and the corresponding three-dimensional temperature diffusion data volume are constructed. For the temperature increment image sequence, discrete difference operations are performed on the temperature increment of each pixel in both the row and column directions to obtain the pixel's two-dimensional temperature gradient vector and gradient magnitude, and to form a heat diffusion intensity image sequence. In the thermal diffusion intensity image sequence, select the first key frame and the second key frame with a predetermined time interval. Calculate the diffusion change rate image matrix based on the difference in thermal diffusion intensity of the same pixel in the first key frame and the time interval between the key frames. The diffusion rate of each pixel was collected within the crack-free reference area of ​​the air conditioner compressor components. The diffusion rate and absolute value of the diffusion rate within the crack-free reference area were statistically analyzed to construct the baseline diffusion rate parameters of the crack-free reference area. Based on the diffusion change rate of each pixel and the baseline diffusion rate parameter of the crack-free reference area, the relative abnormal diffusion ratio of all pixels in the image is calculated, and the set of crack response pixels that meet the abnormal conditions is extracted by statistical discrimination. Based on the crack response pixel set, the crack connected sub-regions are divided according to the relationship between adjacent pixels. Connected sub-regions with an area lower than the minimum recognition area threshold are filtered out, and a crack binary mask image is output.

[0029] By applying short-pulse thermal excitation to air conditioner compressor components and continuously acquiring thermal image sequences throughout the process using infrared thermal imaging, this method firstly enables real-time acquisition of the temperature dynamics of each pixel of the component over time, thus avoiding the problems of traditional single-image static analysis being susceptible to environmental temperature drift and measurement errors. Secondly, by selecting multiple reference frames during the steady-state phase and arithmetically averaging the same pixel, a background temperature baseline image is established, making subsequent temperature increment extraction more robust to environmental interference and instrument noise. Thirdly, by combining the constructed temperature increment image sequence with the three-dimensional temperature diffusion data volume, the spatiotemporal changes during the thermal diffusion process can be objectively reflected, overcoming the shortcomings of existing technologies that rely solely on two-dimensional difference or a single frame image to accurately characterize crack propagation information. Finally, by discrete difference analysis in the row and column directions, the two-dimensional temperature gradient and gradient magnitude of the pixels are obtained, forming a thermal... The diffusion intensity sequence distinguishes diffusion hotspots at crack locations from background diffusion, enhancing the detection of subtle anomalies such as crack tips and pores. The first and second keyframes within this sequence are selected to quantitatively calculate the diffusion change rate image matrix, further compressing the data to the two most informative moments, reducing computational load and improving real-time performance. A baseline diffusion rate parameter is established using a crack-free reference region, normalizing the overall diffusion change rate to a relative anomalous diffusion ratio. Crack response pixel sets are extracted through statistical discrimination, avoiding the poor applicability of the fixed threshold method to different workpieces or environments. Finally, connectivity analysis and area filtering strategies are applied to generate a binary crack mask, achieving automation and high-precision positioning throughout the entire process from data acquisition to crack segmentation, effectively improving the accuracy, stability, and applicability of crack detection for air conditioning compressor components.

[0030] Reference Figure 2 The process of applying pulsed thermal excitation to the air conditioner compressor components and continuously acquiring thermal image sequences covering the entire detection process using an infrared thermal imager to obtain temperature data of each pixel over time specifically includes: Place the air conditioner compressor parts flat in the central area of ​​the thermal imaging platform, so that the surface being inspected faces the imaging direction of the infrared thermal imager. Start the infrared thermal imager, set the sampling frequency and total sampling duration, obtain the total number of thermal image frames to be acquired within the entire sampling duration by multiplying the two, and continuously acquire thermal images according to the calculated total number of frames. In the initial frame at the start of sampling, a pulsed heat source is applied to the upper surface of the air conditioning compressor components for a predetermined pulse duration; For each thermal image frame acquired, the temperature values ​​of each pixel are organized into a two-dimensional temperature matrix according to the arrangement of the row and column pixels of the image, so that each thermal image frame corresponds to a temperature matrix composed of the total number of row pixels and the total number of column pixels. A discrete coordinate system is established on the thermal image plane. The center of the pixel at the top left corner of the image is set as the origin. The horizontal direction to the right is set as the positive direction of the column axis, and the vertical direction downward is set as the positive direction of the row axis. The pixel row index is specified to be between zero and the total number of pixels in the row minus one, and the pixel column index is specified to be between zero and the total number of pixels in the column minus one, so as to uniquely identify the position of each pixel.

[0031] Place the metal parts of the air conditioner compressor flat in the center of the thermal imaging platform; Start the infrared thermal imager and set its sampling frequency to [value]. Set the total sampling time to Calculate the total number of thermal image frames acquired during the entire sampling period. ; Discrete frame index in thermal imaging sequences Apply a time duration to the upper surface of the workpiece irradiated by a pulsed heat source; among which, The duration of pulsed heat source irradiation; For each frame of the acquired image, construct the first... The temperature matrix of the frame thermal image is: , ;in, This is the discrete frame index for the thermal imaging sequence; For the first The temperature matrix of the frame thermal image has dimensions of ; In the first In the frame thermal image, the row index is Column index is The temperature value of the pixel; Establish a two-dimensional discrete coordinate system on the thermal image plane, and define the center of the pixel at the top left corner of the image as the origin. The horizontal direction to the right is set as the positive direction of the column axis, and the vertical direction downwards is set as the positive direction of the row axis. (Restricted shares row index) The range of values ​​is Pixel column index The range of values ​​is ;in, This represents the total number of pixels in each row of the thermal image. This represents the total number of columns of pixels in the thermal image.

[0032] The step of selecting several reference frames in the steady-state phase of the thermal imaging sequence and performing an arithmetic mean on the temperature of the same pixel in each reference frame to obtain a background temperature reference image specifically includes: The number of reference frames is selected based on the total number of sampled frames. The number of reference frames is greater than or equal to one and less than the total number of sampled frames minus two. For each pixel location in the thermal image sequence, the temperature value of each pixel in all reference frames is obtained from the temperature matrix of each frame corresponding to the selected reference frame. The arithmetic mean of these temperature values ​​is then performed to obtain the background average temperature of each pixel. Based on the average background temperature of each pixel, a background temperature reference image with the same size as a single-frame thermal image is constructed, so that the background temperature reference image stores the corresponding average background temperature at each pixel.

[0033] Set the number of reference frames used to calculate the average background temperature to 1. ,satisfy ; For each pixel position The background average temperature is calculated as follows: ;in, The average temperature of the background image in pixels Temperature value at; Construct a background temperature reference image as .

[0034] Based on the background temperature reference image and thermal image sequence, the temperature increment of each pixel in each frame of the thermal image relative to the average background temperature is calculated, and a temperature increment map sequence and the corresponding three-dimensional temperature diffusion data volume are constructed, specifically including: In the thermal imaging sequence, starting from a frame after the reference frame until the end of sampling, the temperature matrix of each frame is selected sequentially. For each pixel position in the selected frame temperature matrix, obtain the pixel's temperature value in the current frame and the average background temperature in the background temperature reference image. Subtract the average background temperature from the current frame temperature to obtain the pixel's temperature increment in the current frame. Based on the temperature increment of all pixels in each frame, a temperature increment map with the same size as the single-frame thermal image is constructed. A temperature increment map sequence is constructed, and the temperature increment map sequence is combined into a three-dimensional temperature diffusion data volume in chronological order.

[0035] For each frame The calculated temperature increment graph is as follows: ;in, For the first Frame thermal image at pixel The temperature increment relative to the background temperature; The temperature difference sequence is obtained as follows: , ;in, For the first The temperature increment matrix of the frame, with dimensions of ; All difference plots constitute the temperature diffusion volume data block: ;in, This is a volumetric data block for temperature diffusion.

[0036] The process for generating a temperature increment image sequence involves performing discrete difference operations on the temperature increment of each pixel in both the row and column directions to obtain the pixel's two-dimensional temperature gradient vector and gradient magnitude, thus forming a heat diffusion intensity image sequence. Specifically, this includes: For any frame in the temperature increment map sequence, for each row of pixels except the top row and the bottom row, obtain the temperature increment of each pixel at the position of the adjacent row above and the adjacent row below, and take half of the difference between the two as the row direction temperature difference component; for pixels located in the top row and the bottom row, set the row direction temperature difference component to zero. For any frame in the temperature increment map sequence, for each column of pixels except the leftmost and rightmost columns, obtain the temperature increment of each pixel at the position of the left adjacent column and the right adjacent column respectively, and take half of the difference between the two as the column direction temperature difference component; for pixels located in the leftmost and rightmost columns, set the column direction temperature difference component to zero. In the current frame, for each pixel, the row-direction temperature difference component and the column-direction temperature difference component are combined to form a two-dimensional temperature gradient vector for the pixel. In the current frame, for each pixel, the square root of the sum of the squares of the row-direction temperature difference components and the squares of the column-direction temperature difference components is obtained to obtain the gradient magnitude of the pixel. The gradient magnitudes of each pixel are then organized into a heat diffusion intensity image with the same size as the single-frame thermal image according to the image position. By repeating the row-direction difference, column-direction difference, and pixel two-dimensional temperature gradient vector construction and gradient magnitude calculation for each frame, a heat diffusion intensity image sequence is formed.

[0037] Calculate the first Pixels in a frame Temperature difference component in the row direction Specifically: ; Calculate the first Pixels in a frame Temperature difference component in column direction Specifically: ; For any frame In pixels The gradient vector is constructed as follows: ;in, For the first Pixels in a frame Temperature gradient vector at; Calculate the first Pixels in a frame gradient magnitude Specifically: ; The constructed thermal diffusion intensity image sequence is as follows: ;in, For the first The thermal diffusion intensity matrix of the frame has dimensions of .

[0038] The step of selecting a first keyframe and a second keyframe with a predetermined time interval from the thermal diffusion intensity image sequence, and calculating the diffusion change rate image matrix based on the difference in thermal diffusion intensity of the same pixel in the first keyframe and the time interval between the keyframes, specifically includes: In the heat diffusion intensity image sequence, the first key frame and the second key frame at different moments in the temperature diffusion process are selected. The second key frame is later than the first key frame in time. The time interval between the two key frames is calculated based on the sampling frequency and the difference in frame number between the two frames. For each pixel position, obtain the gradient magnitude of each pixel in the first keyframe and the gradient magnitude in the second keyframe, and divide the difference between the two by the time interval between the two keyframes to obtain the diffusion rate of each pixel. Based on the diffusion rate of each pixel, a diffusion rate image matrix with the same size as a single-frame thermal image is constructed across the entire image area.

[0039] Select two keyframe numbers , ,satisfy The time interval is: ;in, , These are the discrete frame indices for the selected first and second keyframes, respectively; The time interval between two keyframes; Calculate pixels thermal diffusivity ;in, , They were respectively in the second Frame, First Pixels in a frame The gradient magnitude; Obtain the diffusion rate of change image matrix .

[0040] The diffusion rate of each pixel is collected within a crack-free reference area of ​​the air conditioner compressor components. The diffusion rate and its absolute value within the crack-free reference area are statistically analyzed to construct a baseline diffusion rate parameter for the crack-free reference area. Specifically, this includes: A crack-free reference area is pre-defined within the crack-free area on the surface of the air conditioner compressor parts, and at least one pixel with a non-zero diffusion rate is confirmed in the diffusion rate image matrix within the crack-free reference area. The positions of all pixels within the crack-free reference region are used to form a pixel index set, and the number of pixels in the set is used as the area of ​​the crack-free reference region. The average diffusion rate of the crack-free reference region is obtained by taking the arithmetic mean of the diffusion rate of all pixels in the crack-free reference region. The arithmetic mean of the absolute values ​​of the diffusion rate of all pixels in the crack-free reference region is calculated to obtain the average absolute value of the diffusion rate of the crack-free reference region. The average diffusion rate and the average absolute value of the diffusion rate together constitute the reference diffusion rate parameter of the crack-free reference region.

[0041] A defect-free rectangular region is pre-selected, and at least one pixel exists within the pre-selected defect-free rectangular region. satisfy Construct the set of pixel indices corresponding to the selected crack-free reference region. And calculate the area of ​​the defect-free rectangular region. ;in, For set The cardinality; Calculate the average diffusion rate and the absolute average diffusion rate within the defect-free rectangular region respectively: , ;in, This is the arithmetic mean of the diffusion rate within the defect-free rectangular region; This represents the average absolute value of the diffusion rate of change within a defect-free rectangular region. For pixels The absolute value of the diffusion rate at that point.

[0042] The method involves calculating the relative anomalous diffusion ratio of all pixels in the image based on the diffusion change rate of each pixel and the baseline diffusion rate parameter of the crack-free reference region. Then, a set of crack response pixels satisfying the anomalous conditions is extracted using a statistical discrimination method. Specifically, this includes: For each pixel in the diffusion rate image matrix, obtain the pixel's diffusion rate, the average diffusion rate of the crack-free reference area, and the average absolute value of the diffusion rate. Subtract the average diffusion rate from the pixel's diffusion rate and then divide by the average absolute value of the diffusion rate to obtain the pixel's relative abnormal diffusion ratio. Construct a relative anomalous diffusion ratio image matrix with the same size as the single-frame thermal image based on the relative anomalous diffusion ratio of each pixel; In the relative anomalous diffusion ratio image matrix, the arithmetic mean of the relative anomalous diffusion ratio of all pixels is calculated to obtain the mean of the whole image. The square root of the average of the squares of the differences between the relative anomalous diffusion ratio of each pixel and the mean of the whole image is then used to obtain the standard deviation of the whole image. The sum of the mean of the entire image and three times the standard deviation is set as the abnormal diffusion detection threshold. The threshold is compared for each pixel in the image matrix with the relative abnormal diffusion ratio. The pixel positions with the relative abnormal diffusion ratio greater than the abnormal diffusion detection threshold are summarized into the crack response pixel set.

[0043] Calculate the relative anomalous diffusion ratio for all pixels: ;in, For pixels The relative abnormal diffusion ratio; Constructing the relative anomaly diffusion ratio matrix ; right The mean and standard deviation of the entire graph are calculated as follows: , ;in, Within the entire map area The arithmetic mean; Within the entire map area Standard deviation; Set abnormal diffusion detection threshold ; Construct a set of crack response points .

[0044] The method based on the crack response pixel set, dividing the crack into connected sub-regions according to the relationship between adjacent pixels, filtering out connected sub-regions with an area lower than the minimum recognition area threshold, and outputting a crack binary mask image, specifically includes: In the crack response pixel set, two pixels are considered to be adjacent pixels when the sum of the absolute values ​​of the difference between their row coordinates and the absolute values ​​of their column coordinates is equal to one. Based on the relationship between adjacent pixels, a connectivity analysis is performed on the crack response pixel set, and the interconnected pixels are divided into several crack connected sub-regions. The number of connected sub-regions is recorded as the total number of connected sub-regions. When the crack response pixel set is empty, the total number of connected sub-regions is set to zero, and no crack connected sub-regions are generated. For each crack-connected sub-region, count the number of pixels it contains, and use that as the area of ​​the corresponding crack-connected sub-region. Set the minimum recognition area threshold to one pixel, retain all crack connected sub-regions with an area greater than the minimum recognition area threshold, and remove crack connected sub-regions with an area less than or equal to the minimum recognition area threshold. On an image plane with the same size as a single-frame thermal image, pixels belonging to the preserved crack connected sub-regions are assigned a value of one, and the remaining pixels are assigned a value of zero, forming a crack binary mask image, and the crack binary mask image is output.

[0045] In the set In the context, any two pixels are defined. and satisfy: When they are adjacent pixels; Based on this adjacency relationship, in the set Divide the region into connected sub-regions: ;in, Set of crack response points The first in Connected subregions; This represents the number of connected subregions. This represents the number of connected subregions. When the crack response point set When the set is empty, the number of connected subregions A value of 0 will not generate any connected sub-regions. Calculate the first Area of ​​each connected subregion ;in, For set The cardinality; Set minimum recognition area threshold Retain all that meet the requirements ; connected subregions; The crack mask image is constructed as follows: ;in, Crack mask image in pixels The value at the location is 1, which means that the pixel belongs to the final retained crack connected region, and 0 means that the pixel does not belong to the crack region; Output the binary mask image matrix of cracks .

[0046] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0047] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for intelligent detection of air conditioner compressor parts, characterized in that, include: Pulsed thermal excitation is applied to the air conditioner compressor parts, and an infrared thermal imager is used to continuously acquire thermal image sequences covering the entire detection process to obtain temperature data of each pixel over time. Several reference frames in the steady-state phase are selected from the thermal imaging sequence, and the temperature of the same pixel in each reference frame is arithmetically averaged to obtain the background temperature reference image. Based on the background temperature reference image and thermal image sequence, the temperature increment of the current temperature of each pixel in each frame thermal image relative to the average background temperature is calculated, and a temperature increment map sequence and the corresponding three-dimensional temperature diffusion data volume are constructed. For the temperature increment image sequence, discrete difference operations are performed on the temperature increment of each pixel in both the row and column directions to obtain the pixel's two-dimensional temperature gradient vector and gradient magnitude, and to form a heat diffusion intensity image sequence. In the thermal diffusion intensity image sequence, select the first key frame and the second key frame with a predetermined time interval. Calculate the diffusion change rate image matrix based on the difference in thermal diffusion intensity of the same pixel in the first key frame and the time interval between the key frames. The diffusion rate of each pixel was collected within the crack-free reference area of ​​the air conditioner compressor components. The diffusion rate and absolute value of the diffusion rate within the crack-free reference area were statistically analyzed to construct the baseline diffusion rate parameters of the crack-free reference area. Based on the diffusion change rate of each pixel and the baseline diffusion rate parameter of the crack-free reference area, the relative abnormal diffusion ratio of all pixels in the image is calculated, and the set of crack response pixels that meet the abnormal conditions is extracted by statistical discrimination. Based on the crack response pixel set, the crack connected sub-regions are divided according to the relationship between adjacent pixels. Connected sub-regions with an area lower than the minimum recognition area threshold are filtered out, and a crack binary mask image is output.

2. The intelligent detection method for air conditioner compressor parts according to claim 1, characterized in that, The process of applying pulsed thermal excitation to the air conditioner compressor components and continuously acquiring thermal image sequences covering the entire detection process using an infrared thermal imager to obtain temperature data of each pixel over time specifically includes: Place the air conditioner compressor parts flat in the central area of ​​the thermal imaging platform, so that the surface being inspected faces the imaging direction of the infrared thermal imager. Start the infrared thermal imager, set the sampling frequency and total sampling duration, obtain the total number of thermal image frames to be acquired within the entire sampling duration by multiplying the two, and continuously acquire thermal images according to the calculated total number of frames. In the initial frame at the start of sampling, a pulsed heat source is applied to the upper surface of the air conditioning compressor components for a predetermined pulse duration; For each thermal image frame acquired, the temperature values ​​of each pixel are organized into a two-dimensional temperature matrix according to the arrangement of the row and column pixels of the image, so that each thermal image frame corresponds to a temperature matrix composed of the total number of row pixels and the total number of column pixels. A discrete coordinate system is established on the thermal image plane. The center of the pixel at the top left corner of the image is set as the origin. The horizontal direction to the right is set as the positive direction of the column axis, and the vertical direction downward is set as the positive direction of the row axis. The pixel row index is specified to be between zero and the total number of pixels in the row minus one, and the pixel column index is specified to be between zero and the total number of pixels in the column minus one, so as to uniquely identify the position of each pixel.

3. The intelligent detection method for air conditioner compressor parts according to claim 2, characterized in that, The step of selecting several reference frames in the steady-state phase of the thermal imaging sequence and performing an arithmetic mean on the temperature of the same pixel in each reference frame to obtain a background temperature reference image specifically includes: The number of reference frames is selected based on the total number of sampled frames. The number of reference frames is greater than or equal to one and less than the total number of sampled frames minus two. For each pixel location in the thermal image sequence, the temperature value of each pixel in all reference frames is obtained from the temperature matrix of each frame corresponding to the selected reference frame. The arithmetic mean of these temperature values ​​is then performed to obtain the background average temperature of each pixel. Based on the average background temperature of each pixel, a background temperature reference image with the same size as a single-frame thermal image is constructed, so that the background temperature reference image stores the corresponding average background temperature at each pixel.

4. The intelligent detection method for air conditioner compressor parts according to claim 3, characterized in that, Based on the background temperature reference image and thermal image sequence, the temperature increment of each pixel in each frame of the thermal image relative to the average background temperature is calculated, and a temperature increment map sequence and the corresponding three-dimensional temperature diffusion data volume are constructed, specifically including: In the thermal imaging sequence, starting from a frame after the reference frame until the end of sampling, the temperature matrix of each frame is selected sequentially. For each pixel position in the selected frame temperature matrix, obtain the pixel's temperature value in the current frame and the average background temperature in the background temperature reference image. Subtract the average background temperature from the current frame temperature to obtain the pixel's temperature increment in the current frame. Based on the temperature increment of all pixels in each frame, a temperature increment map with the same size as the single-frame thermal image is constructed. A temperature increment map sequence is constructed, and the temperature increment map sequence is combined into a three-dimensional temperature diffusion data volume in chronological order.

5. The intelligent detection method for air conditioner compressor parts according to claim 4, characterized in that, The process for generating a temperature increment image sequence involves performing discrete difference operations on the temperature increment of each pixel in both the row and column directions to obtain the pixel's two-dimensional temperature gradient vector and gradient magnitude, thus forming a heat diffusion intensity image sequence. Specifically, this includes: For any frame in the temperature increment map sequence, for each row of pixels except the top row and the bottom row, obtain the temperature increment of each pixel at the position of the adjacent row above and the adjacent row below, and take half of the difference between the two as the row direction temperature difference component; for pixels located in the top row and the bottom row, set the row direction temperature difference component to zero. For any frame in the temperature increment map sequence, for each column of pixels except the leftmost and rightmost columns, obtain the temperature increment of each pixel at the position of the left adjacent column and the right adjacent column respectively, and take half of the difference between the two as the column direction temperature difference component; for pixels located in the leftmost and rightmost columns, set the column direction temperature difference component to zero. In the current frame, for each pixel, the row-direction temperature difference component and the column-direction temperature difference component are combined to form a two-dimensional temperature gradient vector for the pixel. In the current frame, for each pixel, the square root of the sum of the squares of the row-direction temperature difference components and the squares of the column-direction temperature difference components is obtained to obtain the gradient magnitude of the pixel. The gradient magnitudes of each pixel are then organized into a heat diffusion intensity image with the same size as the single-frame thermal image according to the image position. By repeating the row-direction difference, column-direction difference, and pixel two-dimensional temperature gradient vector construction and gradient magnitude calculation for each frame, a heat diffusion intensity image sequence is formed.

6. The intelligent detection method for air conditioner compressor parts according to claim 5, characterized in that, The step of selecting a first keyframe and a second keyframe with a predetermined time interval from the thermal diffusion intensity image sequence, and calculating the diffusion change rate image matrix based on the difference in thermal diffusion intensity of the same pixel in the first keyframe and the time interval between the keyframes, specifically includes: In the heat diffusion intensity image sequence, the first key frame and the second key frame at different moments in the temperature diffusion process are selected. The second key frame is later than the first key frame in time. The time interval between the two key frames is calculated based on the sampling frequency and the difference in frame number between the two frames. For each pixel position, obtain the gradient magnitude of each pixel in the first keyframe and the gradient magnitude in the second keyframe, and divide the difference between the two by the time interval between the two keyframes to obtain the diffusion rate of each pixel. Based on the diffusion rate of each pixel, a diffusion rate image matrix with the same size as a single-frame thermal image is constructed across the entire image area.

7. The intelligent detection method for air conditioner compressor parts according to claim 6, characterized in that, The diffusion rate of each pixel is collected within a crack-free reference area of ​​the air conditioner compressor components. The diffusion rate and its absolute value within the crack-free reference area are statistically analyzed to construct a baseline diffusion rate parameter for the crack-free reference area. Specifically, this includes: A crack-free reference area is pre-defined within the crack-free area on the surface of the air conditioner compressor parts, and at least one pixel with a non-zero diffusion rate is confirmed in the diffusion rate image matrix within the crack-free reference area. The positions of all pixels within the crack-free reference region are used to form a pixel index set, and the number of pixels in the set is used as the area of ​​the crack-free reference region. The average diffusion rate of the crack-free reference region is obtained by taking the arithmetic mean of the diffusion rate of all pixels in the crack-free reference region. The arithmetic mean of the absolute values ​​of the diffusion rate of all pixels in the crack-free reference region is calculated to obtain the average absolute value of the diffusion rate of the crack-free reference region. The average diffusion rate and the average absolute value of the diffusion rate together constitute the reference diffusion rate parameter of the crack-free reference region.

8. The intelligent detection method for air conditioner compressor parts according to claim 7, characterized in that, The method involves calculating the relative anomalous diffusion ratio of all pixels in the image based on the diffusion change rate of each pixel and the baseline diffusion rate parameter of the crack-free reference region. Then, a set of crack response pixels satisfying the anomalous conditions is extracted using a statistical discrimination method. Specifically, this includes: For each pixel in the diffusion rate image matrix, obtain the pixel's diffusion rate, the average diffusion rate of the crack-free reference area, and the average absolute value of the diffusion rate. Subtract the average diffusion rate from the pixel's diffusion rate and then divide by the average absolute value of the diffusion rate to obtain the pixel's relative abnormal diffusion ratio. Construct a relative anomalous diffusion ratio image matrix with the same size as the single-frame thermal image based on the relative anomalous diffusion ratio of each pixel; In the relative anomalous diffusion ratio image matrix, the arithmetic mean of the relative anomalous diffusion ratio of all pixels is calculated to obtain the mean of the whole image. The square root of the average of the squares of the differences between the relative anomalous diffusion ratio of each pixel and the mean of the whole image is then used to obtain the standard deviation of the whole image. The sum of the mean of the entire image and three times the standard deviation is set as the abnormal diffusion detection threshold. The threshold is compared for each pixel in the image matrix with the relative abnormal diffusion ratio. The pixel positions with the relative abnormal diffusion ratio greater than the abnormal diffusion detection threshold are summarized into the crack response pixel set.

9. The intelligent detection method for air conditioner compressor parts according to claim 8, characterized in that, The method based on the crack response pixel set, dividing the crack into connected sub-regions according to the relationship between adjacent pixels, filtering out connected sub-regions with an area lower than the minimum recognition area threshold, and outputting a crack binary mask image, specifically includes: In the crack response pixel set, two pixels are considered to be adjacent pixels when the sum of the absolute values ​​of the difference between their row coordinates and the absolute values ​​of their column coordinates is equal to one. Based on the relationship between adjacent pixels, a connectivity analysis is performed on the crack response pixel set, and the interconnected pixels are divided into several crack connected sub-regions. The number of connected sub-regions is recorded as the total number of connected sub-regions. When the crack response pixel set is empty, the total number of connected sub-regions is set to zero, and no crack connected sub-regions are generated. For each crack-connected sub-region, count the number of pixels it contains, and use that as the area of ​​the corresponding crack-connected sub-region. Set the minimum recognition area threshold to one pixel, retain all crack connected sub-regions with an area greater than the minimum recognition area threshold, and remove crack connected sub-regions with an area less than or equal to the minimum recognition area threshold. On an image plane with the same size as a single-frame thermal image, pixels belonging to the preserved crack connected sub-regions are assigned a value of one, and the remaining pixels are assigned a value of zero, forming a crack binary mask image, and the crack binary mask image is output.

Images