Real-time monitoring method for aerospace part machining line

Abstract

The invention belongs to the technical field of image communication, and particularly relates to a real-time monitoring method for an aerospace part processing line, and the method comprises the steps: collecting a processing line image, and carrying out the graying processing; dividing each frame of grayscale image into blocks which are not overlapped with each other, and determining the feature importance of the blocks according to the time sequence grayscale change; determining the distribution discrete degree of the blocks based on the pixel gray level distribution; combining the feature importance degree and the distribution dispersion degree to construct a quantitative priority index; a dynamic threshold adaptive mechanism is introduced, a quantization priority division threshold is dynamically adjusted according to a processing state index and a system resource state index, and the quantization number of each block is determined; and finally, carrying out differential compression on the blocks according to the quantization quantity. According to the invention, the balance between the high fidelity of the key area and the overall high efficiency is realized, and reliable real-time monitoring support is provided for the high-precision manufacturing process of aerospace parts.

Description

Technical Field

[0001] The present invention relates to the field of image communication technology, and more particularly to a real-time monitoring method for an aerospace parts processing line. Background Art

[0002] As core products of the high-end manufacturing industry, the machining accuracy and quality of aerospace parts are directly related to the safety and reliability of aircraft. During the machining of key components such as aircraft engine blades and turbine disks, real-time monitoring systems play a vital role in promptly detecting machining anomalies such as tool wear, material delamination, and surface cracks. With the development of intelligent manufacturing technology, vision-based image monitoring methods have gradually become an important means of quality control in high-end manufacturing processes due to their advantages such as non-contact, rich information, and fast response. However, the amount of unprocessed raw image data is huge. If the unprocessed images are directly transmitted to the monitoring server or cloud for analysis, it will bring huge bandwidth pressure and storage burden. Therefore, efficient image compression technology has become a key link in realizing real-time monitoring of aerospace parts machining processes.

[0003] The Absolute Moment Double Threshold Compression (AMBTC) algorithm is currently widely used in industrial visual monitoring due to its computational simplicity and high compression efficiency. The AMBTC algorithm compresses image blocks by calculating the mean of the high and low values ​​within the block as two quantization values ​​and assigning a binary label to each pixel. However, the AMBTC algorithm uses the same two quantization values ​​for all blocks, making it unable to adapt to the differences in information density across different regions. For processing areas with complex textures (such as surface scratches, microcracks, and tool cutting marks), the insufficient quantization level leads to blurring or distortion of key features in the reconstructed image, resulting in a significant loss of detailed information. For smooth areas (such as uniform metal surfaces), the two quantization values ​​result in low compression efficiency and the presence of a large amount of redundant information. This "one-size-fits-all" compression strategy makes it difficult for the AMBTC algorithm to simultaneously meet the dual requirements of high fidelity in key areas and high overall efficiency, affecting the accuracy of real-time monitoring of aerospace parts processing. Summary of the Invention

[0004] To address the technical problem that the AMBTC algorithm described above is unable to simultaneously meet the dual requirements of high fidelity in key areas and overall high efficiency, which affects the accuracy of real-time monitoring of aerospace parts processing, the present invention provides a real-time monitoring method for aerospace parts processing lines, comprising: The system captures images of an aerospace parts processing line and grayscales each captured image frame; divides each grayscale image frame into several non-overlapping blocks, and determines the feature importance of each block based on the grayscale changes of the same block in the grayscale image at different times; determines the distribution discreteness of each block based on the grayscale distribution of pixels in each block; determines the quantization priority of each block based on the distribution discreteness and feature importance of each block; determines the quantization quantity of each block based on a dynamic threshold adaptive mechanism of system status and processing process; compresses the blocks based on the quantization quantity of each block to obtain compressed data; and performs real-time monitoring and abnormality warning based on the compressed data; wherein the dynamic threshold adaptive mechanism includes: dynamically adjusting the quantization priority division threshold according to the current processing status and the resource status of the edge computing device, increasing the quantization quantity of the block to ensure image quality when the processing process is in a high-risk stage, and reducing the quantization quantity of the block to ensure system real-time performance when the edge computing device resources are tight.

[0005] Preferably, the feature importance satisfies the expression: ;in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image The feature importance of each block, Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the size of the block; Indicates the number of historical frames used for reference; is a hyperparameter; represents the hyperbolic tangent function; Indicates the absolute value symbol.

[0006] Preferably, the distribution dispersion satisfies the expression: ;in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image The degree of dispersion of the distribution of blocks; Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the Frame grayscale image The minimum gray value in the blocks, Indicates the Frame grayscale image The maximum gray value in the blocks; Indicates the size of the block; represents the minimum function.

[0007] Preferably, the quantized priority satisfies the expression: ;in, Indicates the Frame grayscale image Quantized priority of each block; Indicates the Frame image The degree of dispersion of the distribution of blocks; Indicates the Frame grayscale image The feature importance of each block; represents the natural exponential function; is a hyperparameter used to adjust the sensitivity of the quantization priority to the feature importance.

[0008] Preferably, the hyperparameters The method of obtaining is: according to The feature importance of all blocks in the frame grayscale image clusters all blocks into two categories, and the mean of the feature importance of all blocks contained in the category is used as the representative feature importance of the category, and the mean of the representative feature importance of the two categories is used as value.

[0009] Preferably, the method of dynamically adjusting the quantization priority division threshold according to the processing status at the current moment and the resource status of the edge computing device includes: determining the processing status index according to the change in the feature importance of each block in the current frame grayscale image relative to the feature importance of the corresponding block in the historical grayscale image, determining the system resource status index in combination with the CPU usage, memory occupancy and bandwidth utilization of the edge computing device, and dynamically adjusting the quantization priority division threshold according to the processing status index and the system resource status index.

[0010] Preferably, the processing state index satisfies the expression: ;in, Indicates the sequence number of the current frame, Indicates the Processing status indicators of frame grayscale images; Indicates the total number of blocks; Indicates the Frame grayscale image The feature importance of each block; and Respectively represent Blocks in the past The mean and standard deviation of feature importance in the frame grayscale image, Indicates the preset reference quantity; represents the hyperbolic tangent function; Indicates the absolute value symbol.

[0011] Preferably, the method for obtaining the system resource status indicator includes: obtaining the maximum value of the CPU usage, memory occupancy, and bandwidth utilization of the edge computing device at the current moment, subtracting the maximum value from 1, and using the result as the system resource status indicator.

[0012] Preferably, the dynamically adjusting the division threshold of the quantization priority according to the processing state indicator and the system resource state indicator includes: modifying the initial division threshold according to the processing state indicator and the system resource state indicator to obtain the instantaneous division threshold of the current frame grayscale image, and performing weighted averaging of the instantaneous division threshold and the historical dynamic division threshold to obtain the dynamic division threshold of the current frame grayscale image; the instantaneous division threshold satisfies the expression: ;in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image An instant partition threshold; Indicates the An initial partition threshold; Indicates the Processing status indicators corresponding to the frame grayscale image; Indicates the system resource status indicators at the current moment; Indicates the normal state threshold of the processing process, Indicates the threshold for sufficient system resources. Indicates the processing state adjustment coefficient, Indicates the system resource adjustment coefficient.

[0013] Preferably, the blocks are compressed according to the quantization number of each block to obtain compressed data, including: when the quantization number of the block is C=1, the grayscale mean of all pixels in the block is calculated as the only quantization value, and no bitmap is generated; when the quantization number of the block is C=2, the high and low means are calculated as the quantization value, and a 1-bit / pixel bitmap is generated; when the quantization number of the block is C=3 or C=4, the K-means clustering algorithm is used to divide the pixels in the block into C clusters, and the grayscale mean of all pixels in each cluster is calculated as the quantization value, and a 2-bit / pixel bitmap is generated at the same time; when the quantization number of the block is C=5, the K-means clustering algorithm is used to divide the pixels in the block into 5 clusters, and the grayscale mean of all pixels in each cluster is calculated as the quantization value, and a 3-bit / pixel bitmap is generated at the same time; the quantization values ​​of all blocks and the bitmaps are combined into compressed data.

[0014] The beneficial effects of the present invention are as follows: the present invention realizes an intelligent strategy of "prioritized fidelity preservation in critical areas and efficient compression in non-critical areas" through a dual evaluation mechanism of feature importance and distribution discreteness, thereby realizing high-precision, high-reliability real-time monitoring of aerospace parts processing.

[0015] The present invention identifies abnormal processing states by analyzing the temporal changes in feature importance, and can accurately capture key abnormal features such as tool cutting into the material interface area, surface cracks, and material delamination.

[0016] The dynamic threshold adaptive mechanism of the present invention can adjust the compression strategy in real time based on the CPU usage, memory utilization, and bandwidth utilization of the edge computing device. When system resources are tight, the quantization of non-critical areas is automatically reduced to ensure stable operation of the monitoring system; when resources are abundant, the quantization of critical areas is appropriately increased to enhance monitoring accuracy.

[0017] By analyzing the normalized deviation of feature importance relative to historical data, this method accurately identifies different stages of the machining process and dynamically adjusts the compression strategy. During abnormal machining phases, the system automatically improves image quality in key areas; during stable machining phases, it automatically increases compression efficiency. This intelligently balances image quality in key areas with overall compression efficiency, ensuring accurate real-time monitoring of aerospace parts machining lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The flowchart schematically illustrates a real-time monitoring method for an aerospace parts processing line according to the present invention. DETAILED DESCRIPTION

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work shall fall within the scope of protection of the present invention.

[0020] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] The embodiment of the present invention discloses a real-time monitoring method for an aerospace parts processing line, referring to Figure 1 , including steps S1 to S7: S1. Collect images of an aerospace parts processing line and perform grayscale processing on each frame of the collected image.

[0022] Specifically, a high-frame-rate industrial camera is used to continuously capture the cutting area of ​​key components such as aircraft engine blades and turbine disks being processed in CNC machine tools, acquiring a dynamic image stream including the contact process between the tool and the workpiece.

[0023] It should be noted that the image acquisition frequency in this embodiment is set to 120 frames per second to fully capture transient phenomena such as tiny vibrations and surface deformations generated during high-speed milling. In other embodiments, the implementation personnel can set the acquisition frequency according to actual implementation conditions.

[0024] Furthermore, the subsequent calculation overhead is reduced, and the present invention performs grayscale processing on each frame of the acquired image.

[0025] It should be noted that grayscale processing can not only reduce the data dimension and subsequent computational overhead, but also enhance the robustness of the image to illumination fluctuations, which is conducive to highlighting the structural and texture changes in images of aerospace parts processing lines.

[0026] S2. Divide each frame of grayscale image into several non-overlapping blocks, and determine the feature importance of the blocks according to the grayscale changes of the same blocks in the grayscale image at different times.

[0027] Specifically, each frame of grayscale image is divided into several non-overlapping In this embodiment, the block size is set , , that is, each block is Pixel size. In other embodiments, implementers can set the block size according to actual implementation conditions.

[0028] It's important to note that in aerospace parts machining, the tool path is fixed, the background area is essentially static, and the grayscale of the cutting area and feed point varies dramatically over time. Therefore, pixel differencing can be used to identify key monitoring areas. This method analyzes the grayscale evolution of each block over time to assess its feature importance in aerospace parts machining.

[0029] Specifically, the feature importance satisfies the expression:

[0030] in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image The feature importance of each block, Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the size of the block; represents the number of historical frames used for reference. In this embodiment, In other embodiments, the implementer can set size; is a hyperparameter used to prevent The larger the value, the closer the feature importance is to 1. , Indicates the Frame grayscale image and previous The average grayscale difference of corresponding pixels in the frame grayscale image. In other embodiments, the implementer can set the hyperparameter according to the actual implementation situation. ; represents the hyperbolic tangent function; Indicates the absolute value symbol.

[0031] When Frame grayscale image The grayscale value of the pixel in the block is consistent with the previous The greater the difference in the grayscale values ​​of the pixels in the corresponding blocks in the frame grayscale image, the The first block is more likely to contain important features that can reflect the changes in the processing state, such as tool cutting, chip generation, and surface morphology mutation. The greater the feature importance of a block, the more attention will be paid to retaining the detailed information of the area in the subsequent compression process to ensure accurate monitoring of key processing states. Frame grayscale image The grayscale value of the pixel in the block is consistent with the previous The smaller the difference in the grayscale values ​​of the pixels in the corresponding blocks in the frame grayscale image, the better the The first block is more likely to contain static information such as background area and fixed fixtures. The smaller the feature importance of a block, the more appropriate the compression quality of the area can be reduced in the subsequent compression process to improve the overall compression efficiency.

[0032] It should be noted that the present invention adopts the hyperbolic tangent function Limit feature importance to The range ensures the consistency and comparability of feature importance evaluation under different lighting conditions and image contrasts, and can effectively adapt to the complex and changing lighting environment of aerospace parts processing sites. In other embodiments, the implementer can choose other appropriate nonlinear mapping functions to replace the hyperbolic tangent function according to the actual implementation situation. .

[0033] S3. Determine the distribution discreteness of each block according to the grayscale distribution of the pixels in each block.

[0034] It should be noted that the AMBTC compression algorithm compresses image blocks by calculating the high and low mean values ​​within the block as two quantization values ​​and assigning a binary label to each pixel. However, all blocks are quantized using two quantization values. This results in significant detail loss in areas with complex textures (such as fine machining textures on the surface of aerospace parts, tool cutting marks, etc.) due to insufficient quantization, and key features in the reconstructed image are blurred or distorted. For smooth areas (such as uniform metal surfaces, etc.), compression efficiency is low and redundant information exists. Therefore, the present invention analyzes the grayscale distribution characteristics of pixels within the block to determine the degree of distribution dispersion of the block, so that different blocks can be differentially quantized based on the degree of distribution dispersion, realizing an adaptive strategy of "fine representation of complex areas and efficient compression of simple areas."

[0035] It should be further explained that in the images of aerospace parts processing lines, the larger the distribution range and the more uniform the distribution (such as processing areas with rich surface textures), the more quantization values ​​are needed to quantize them, thereby reducing information loss and ensuring that tiny surface defects or processing anomalies can be accurately captured; and the smaller the distribution range and the closer the distribution is to the two ends of the grayscale range (such as uniform surface areas), the fewer quantization values ​​can be used to quantize them, thereby improving compression efficiency and reducing the burden of data transmission and storage.

[0036] Specifically, the degree of distribution dispersion satisfies the expression:

[0037] in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image The degree of dispersion of the distribution of blocks; Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the Frame grayscale image The minimum gray value in the blocks, Indicates the Frame grayscale image The maximum gray value in the blocks; Indicates the size of the block; represents the minimum function.

[0038] When Frame grayscale image When the grayscale value distribution range of the pixels in a block is large and relatively evenly distributed in the grayscale interval, it means that the block contains rich texture information. The degree of dispersion of the distribution of blocks Large, in the subsequent compression process, more quantization is required to retain the detail information. Frame grayscale image The grayscale values ​​of the pixels in each block are concentrated at both ends of the grayscale range or converge as a whole, indicating that the information of the block is relatively simple. The degree of dispersion of the distribution of blocks Smaller, in the subsequent compression process, a smaller quantization number can be used to improve compression efficiency.

[0039] It should be noted that the distribution dispersion index used in this invention can effectively reflect the "inward distribution" characteristics of pixel values ​​within the grayscale range. Compared with traditional variance or entropy indicators, it can more accurately guide the quantization strategy of aerospace parts processing line images. For example, a block contains a large number of pixels close to 0 and 255 (such as bright spots and dark areas coexisting). , , then the majority of pixels in the block The item is close to 0, which makes the distribution of the block less discrete and close to 0; while the grayscale of the other block is concentrated between 100 and 170 and is evenly distributed between 100 and 170. , , making the distribution of blocks more discrete. Then the first block is more suitable for quantization with two representative values ​​(high and low mean), while the second block requires more quantization to retain details.

[0040] S4. Determine the quantization priority of each block based on the distribution dispersion and feature importance of each block.

[0041] It should be noted that critical areas reflect core processing information, such as the contact state between the tool and the workpiece, surface quality variations, and potential defects. Therefore, even if the grayscale distribution in these areas is simple, more quantization values ​​are required to ensure accurate monitoring of critical processing states and timely capture of abnormalities. Non-critical areas (such as the background) have lower monitoring value for processing quality and are not a priority. Therefore, even if the grayscale distribution in these areas is complex (such as texture in the background), fewer quantization values ​​can be used to improve compression efficiency. Therefore, the present invention constructs a quantitative priority index based on the degree of distribution dispersion and combines it with feature importance to implement an intelligent strategy of "prioritizing fidelity in critical areas and efficiently compressing non-critical areas."

[0042] Specifically, the quantized priority satisfies the expression:

[0043] in, Indicates the Frame grayscale image Quantized priority of each block; Indicates the Frame image The degree of dispersion of the distribution of blocks; Indicates the Frame grayscale image The feature importance of each block; represents the natural exponential function; is a hyperparameter used to adjust the sensitivity of quantization priority to feature importance.

[0044] In this example, the hyperparameters The method of obtaining is: according to The feature importance of all blocks in the frame grayscale image clusters all blocks into two categories, and the mean of the feature importance of all blocks contained in the category is used as the representative feature importance of the category, and the mean of the representative feature importance of the two categories is used as It should be noted that the clustering algorithm in this embodiment is K-means clustering. In other embodiments, the implementer can select a clustering algorithm based on the actual implementation situation. Since the tool path is fixed in aerospace parts processing, the background area is basically static, while the grayscale of the cutting area, feed point and other positions will change dramatically over time. Therefore, in the two categories obtained by clustering, the category representing the smaller feature importance is the static area such as the background and the fixed fixture, and the category representing the larger feature importance is the area with significant dynamic changes such as the tool contact area and the cutting area.

[0045] It should be noted that when Frame grayscale image The feature importance of each block Greater than hyperparameters , and the larger it is, the The block is more likely to be the key monitoring area such as the tool contact area. Greater than 0, so is a decimal between (0,1). According to the power function characteristics, even if the The degree of dispersion of the distribution of blocks Smaller, quantified priority will still be pulled up, so that more quantitative values ​​will be allocated to the blocks corresponding to the key monitoring areas in the future; on the contrary, when the feature importance Less than the hyperparameter , and the lower it is, the The blocks are more likely to be non-critical areas such as background and fixed fixtures. Less than 0, so Greater than 1, according to the power function characteristics, even if the distribution is discrete Larger, quantified priority It will also be moderately suppressed, allowing the subsequent allocation of fewer quantization values ​​to the blocks corresponding to non-critical areas to improve compression efficiency. Frame grayscale image The feature importance of each block Equal to the hyperparameter When, The block is more likely to be the transition area between the key area and the background. is equal to 0, so =1, according to the power function characteristics, quantify the priority Depends on The degree of dispersion of the distribution of blocks size.

[0046] S5. Determine the quantization quantity of each block based on the dynamic threshold adaptive mechanism of the system state and the processing process.

[0047] It should be noted that for The original data size is Bit. When using When the quantization value is , the amount of compressed data is bits, of which is the number of bits required for quantization, is the number of bits required for the bitmap. When the amount of compressed data is less than the original data, the compression is effective. When the block size is 8×8 pixels, it is calculated that when However, considering the characteristics of aerospace parts processing images and the real-time requirements of edge computing equipment, this invention will The value is limited to the range of {1, 2, 3, 4, 5}, which means that a maximum of 5 quantization values ​​are used. Therefore, 4 thresholds need to be set to divide the quantization priority into 5 intervals, corresponding to 5 quantization quantities.

[0048] In one embodiment, a fixed partition threshold is used. 、 、 、 , according to the quantization priority of the block, determine the quantization quantity of the block, specifically: When the quantization priority of the block is less than When the quantization priority of the block is greater than or equal to , and less than When the quantization priority of the block is greater than or equal to , and less than When the quantization priority of the block is greater than or equal to , and less than When the quantization priority of the block is greater than or equal to When , the quantization number of blocks is 5.

[0049] Among them, the division threshold 、 、 、 The empirical values ​​are 0.15, 0.35, 0.60, and 0.80, respectively. In other embodiments, the implementer may adjust the division threshold settings according to the actual processing materials and process characteristics. For example, for the finishing stage with higher surface quality requirements, the thresholds may be appropriately lowered to increase the overall quantization quantity.

[0050] It should be noted that abnormal conditions during aerospace parts processing (such as tool wear and material delamination) often manifest as sudden changes in specific image regions. These characteristics are early signs of quality deterioration. Traditional fixed-threshold compression methods cannot distinguish abnormal from normal features and are prone to over-compression and loss of critical abnormal information. Furthermore, the resource status of edge computing devices (such as CPU usage, memory utilization, and network bandwidth) changes dynamically over time. Traditional fixed-threshold mapping methods are unable to adapt to these dynamic changes, often leading to system crashes during critical stages due to insufficient information or resource constraints. Therefore, in another embodiment, the present invention introduces a dynamic threshold adaptive mechanism. By monitoring the processing status and system resource status in real time, it dynamically adjusts the threshold for quantization priority division, achieving an intelligent compression strategy that "preserves high fidelity during abnormal stages and maintains operation during resource constraints."

[0051] In another embodiment, an initial partitioning threshold is set; the system resource status indicator at the current moment is determined based on the CPU usage, memory occupancy, and bandwidth utilization of the edge computing device at the current moment; the processing status indicator of the current frame grayscale image is determined based on the change in the feature importance of each block in the grayscale image of the current frame relative to the feature importance of the corresponding block in the historical grayscale image; the initial partitioning threshold is corrected according to the processing status indicator and the system resource status indicator at the current moment to obtain the instantaneous partitioning threshold of the current frame grayscale image, and the instantaneous partitioning threshold and the historical dynamic partitioning threshold are weighted averaged to obtain the dynamic partitioning threshold of the current frame grayscale image; the dynamic partitioning threshold is used to determine the quantization quantity of each block in the current frame grayscale image according to the quantization priority of each block in the current frame grayscale image.

[0052] Specifically, the setting of the initial segmentation threshold includes: In this embodiment, the initial division threshold 、 、 、 The empirical values ​​of are 0.15, 0.35, 0.60, and 0.80 respectively. In other embodiments, the implementers can set them according to the actual implementation situation.

[0053] Furthermore, the system resource status indicator satisfies the expression:

[0054] in, Indicates the system resource status indicators at the current moment; Indicates the CPU usage of the edge computing device at the current moment; Indicates the memory usage of the edge computing device at the current moment; Indicates the bandwidth utilization of the edge computing device at the current moment; Represents the maximum function. Since the system resource status is mainly limited by the most strained resources, this embodiment takes the maximum value of CPU, memory and bandwidth utilization, and subtracts this maximum value from 1 to obtain the resource surplus. When any resource is close to saturation (such as CPU utilization reaches 90%), the overall resource status of the system will become tense, which may cause data processing delays or losses. At this time, the smaller the system resource status indicator is, the more abundant the system is. On the contrary, when all resources are abundant (such as CPU utilization is only 20%, memory occupancy is only 30%, and bandwidth utilization is only 25%), the overall resource status of the system is relatively loose and can handle more computing tasks. At this time, the larger the system resource status indicator is. By constructing a system resource status indicator, this embodiment can effectively reflect the overall resource status of edge devices and facilitate quick decision-making.

[0055] Furthermore, the processing state index satisfies the expression:

[0056] in, Indicates the sequence number of the current frame, Indicates the Processing status indicators of frame grayscale images; Indicates the total number of blocks; Indicates the Frame grayscale image The feature importance of each block; and Respectively represent Blocks in the past The mean and standard deviation of feature importance in the frame grayscale image, represents a preset reference quantity. In this embodiment, In other embodiments, the implementer can set ; represents the hyperbolic tangent function, which is used to limit the processing state index to the range of [0,1]; Indicates the absolute value symbol.

[0057] It should be noted that this embodiment identifies abnormal changes in processing status by analyzing the standardized deviation of feature importance relative to historical data. Aerospace parts processing usually has periodic characteristics (such as turning of rotating workpieces and milling with periodic feed). Feature importance should show regular changes. When feature importance breaks through the historical change pattern, it is more likely to indicate processing abnormality or critical stage. For example, in the milling process of aircraft engine blades, when the tool cuts into the interface area between titanium alloy and composite material, the feature importance of the cutting area will be significantly higher than the historical average, resulting in As the value increases, the system automatically identifies it as a high-risk stage.

[0058] Furthermore, the instantaneous partition threshold satisfies the expression:

[0059] in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image An instant partition threshold, ; Indicates the An initial partition threshold; Indicates the Processing status indicators corresponding to the frame grayscale image; Indicates the system resource status indicators at the current moment; Indicates the normal state threshold of the processing process, Indicates the threshold for sufficient system resources. Indicates the processing state adjustment coefficient, Indicates the system resource adjustment coefficient.

[0060] In this embodiment, the normal state threshold of the processing process The experience value is 0.3, the threshold of sufficient system resources The empirical value is 0.5, which means the processing state adjustment coefficient The experience value is 0.4, the system resource adjustment coefficient The empirical value of is 0.2. In other embodiments, the implementer can set it according to the actual implementation situation.

[0061] When the aerospace parts processing process is at a high-risk stage, the processing status indicators Increase, and When the threshold is immediately divided Lower, so that more blocks get higher quantization numbers (such as from C=3 to C=4), thereby improving the image quality of areas where processing anomalies may exist; when edge computing device resources are tight (such as high CPU usage, high memory usage, high bandwidth utilization), the system resource status indicator Reduced, and When the threshold is immediately divided Improve, at this time, the quantitative number of non-critical areas can be appropriately reduced (such as from C=2 to C=1) to ensure the real-time performance of the system; on the contrary, when the aerospace parts processing process is in a low-risk stage, the processing state indicator Reduced, and When the threshold is immediately divided Increase, so that the number of quantization is reduced (such as from C=2 to C=1), thereby improving the overall compression efficiency; when the edge computing device resources are sufficient (such as low CPU usage, low memory usage, low bandwidth utilization), the system resource status indicator Increase, and When the threshold is immediately divided In this case, the quantization quantity of the key area can be appropriately increased (for example, from C=3 to C=4) to further improve the image quality of the key area.

[0062] Furthermore, the dynamic partitioning threshold satisfies the expression:

[0063] in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image Dynamic partitioning thresholds, ; Indicates the Frame grayscale image A dynamic partitioning threshold; Indicates the Frame grayscale image An instant partition threshold; Represents the smoothing coefficient, which is used to control the smoothness of the threshold change. In this embodiment In other embodiments, the implementation personnel can set it according to the actual implementation situation. It should be noted that the instant threshold It reflects the immediate state requirements of the current frame, but the threshold may jump due to a sudden change in the state indicator. Therefore, this embodiment introduces a threshold smooth transition mechanism to ensure a gradual change in compression quality and avoid sudden changes in image quality.

[0064] It should be noted that, in order to prevent excessive adjustment of the threshold and maintain basic grading logic, this embodiment imposes upper and lower limit constraints on the dynamic division threshold to ensure that the dynamic division threshold is within a reasonable range.

[0065] Specifically, in response to ,Keep unchanged; in response to ,Will The value of ; In response to ,Will The value of ,in, Indicates the An initial partition threshold. It should be noted that since the value range of the quantization priority is [0,1], when Greater than When , it means that the maximum number of quantized blocks is Since the quantization priority cannot be greater than 1, the quantization priority of all blocks is less than the threshold , the corresponding quantization number will not exceed For example, when Adjust to When , the quantization number of all blocks is at most 4, that is, the case of C=5 will not occur.

[0066] Furthermore, the dynamic division threshold is adopted to determine the quantization quantity of each block in the current frame grayscale image according to the quantization priority of each block in the current frame grayscale image, including: When the quantization priority of the block is less than When the quantization priority of the block is greater than or equal to , and less than When the quantization priority of the block is greater than or equal to , and less than When the quantization priority of the block is greater than or equal to , and less than When the quantization priority of the block is greater than or equal to When , the number of quantized blocks is 5. Among them, Indicates the sequence number of the current frame, 、 、 、 Indicates the dynamic segmentation threshold of the current frame grayscale image.

[0067] It should be noted that the dynamic threshold adaptive mechanism of the present invention identifies key processing stages by analyzing the temporal changes in feature importance, can capture abnormal conditions more accurately, can improve the system's adaptability under complex working conditions, and provide a more intelligent and reliable image compression solution for real-time monitoring of aerospace parts processing processes.

[0068] S6. Compress the blocks according to the quantization quantity of each block.

[0069] Specifically, for any block, the quantization process is: When the quantization number of a block is C=1, the grayscale mean of all pixels in the block is calculated and used as the only quantization value. No bitmap is generated and the compressed data size is 8 bits. When the number of quantization blocks is C=2, the high and low mean values ​​q1 and q2 are calculated as quantization values, and a 1-bit / pixel bitmap is generated. The amount of data after compression is Bit; When the number of quantization blocks is C=3 or C=4, the K-means clustering algorithm is used to divide the pixels in the block into C clusters, and the grayscale mean of all pixels in each cluster is calculated as the quantization value. At the same time, a 2-bit / pixel bitmap is generated. The amount of data after compression is Bit; When the number of quantized blocks is C=5, the K-means clustering algorithm is used to divide the pixels in the block into 5 clusters, and the grayscale mean of all pixels in each cluster is calculated as the quantized value. At the same time, a 3-bit / pixel bitmap is generated. The amount of data after compression is bit.

[0070] The quantized values ​​of all blocks and the bitmap are combined into compressed data.

[0071] It should be noted that in other embodiments, implementers may adjust the clustering algorithm or bitmap encoding method based on actual implementation to suit specific hardware platforms or application requirements. For example, for edge devices with extremely limited computing resources, fast threshold segmentation can be used instead of K-means clustering to reduce computational complexity. For scenarios with extremely high requirements for surface defect detection, the maximum value of C can be appropriately increased (e.g., C=6 or C=7) to further improve image quality in key areas.

[0072] S7. Real-time monitoring and abnormal warning based on compressed data.

[0073] Specifically, the edge computing device transmits the compressed data to the cloud or local monitoring server. The cloud or local monitoring server decodes the compressed data, reconstructs the image based on the quantized value and bitmap, and analyzes the characteristic changes in the contact area between the tool and the workpiece in the image. When abnormal surface morphology, tool wear or material stratification are detected, an abnormal alarm is generated to warn the operator.

[0074] It should be noted that this embodiment uses a convolutional neural network to identify features such as surface morphology anomalies, tool wear, or material stratification. In other embodiments, implementers can determine the anomaly detection algorithm based on actual implementation conditions.

Claims

1. A real-time monitoring method for aerospace parts processing line, characterized in that: include: Collect images of aerospace parts processing lines and perform grayscale processing on each frame of the collected images; Each frame of grayscale image is divided into several non-overlapping blocks, and the feature importance of the block is determined according to the grayscale change of the same block in the grayscale image at different times; Determine the distribution discreteness of each block according to the grayscale distribution of pixels in each block; determine the quantization priority of each block based on the distribution discreteness and feature importance of each block; Based on the dynamic threshold adaptive mechanism of the system state and the processing process, the quantization quantity of each block is determined; according to the quantization quantity of each block, the block is compressed to obtain compressed data; Real-time monitoring and abnormal warning based on compressed data; Among them, the dynamic threshold adaptive mechanism includes: dynamically adjusting the quantization priority division threshold according to the current processing status and the resource status of the edge computing device. When the processing process is in a high-risk stage, the quantization number of blocks is increased to ensure image quality. When the resources of the edge computing device are tight, the quantization number of blocks is reduced to ensure the real-time performance of the system.

2. A real-time monitoring method for an aerospace parts processing line according to claim 1, characterized in that: The feature importance satisfies the expression: ； in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image The feature importance of each block, Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the size of the block; Indicates the number of historical frames used for reference; is a hyperparameter; represents the hyperbolic tangent function; Indicates the absolute value symbol.

3. The real-time monitoring method for an aerospace parts processing line according to claim 1, characterized in that: The degree of dispersion of the distribution satisfies the expression: ； in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image The degree of dispersion of the distribution of blocks; Indicates the Frame grayscale image In the block Gray value of each pixel; Indicates the Frame grayscale image The minimum gray value in the blocks, Indicates the Frame grayscale image The maximum gray value in the blocks; Indicates the size of the block; represents the minimum function.

4. The real-time monitoring method for an aerospace parts processing line according to claim 1, characterized in that: The quantized priority satisfies the expression: ； in, Indicates the Frame grayscale image Quantized priority of each block; Indicates the Frame image The degree of dispersion of the distribution of blocks; Indicates the Frame grayscale image The feature importance of each block; represents the natural exponential function; is a hyperparameter used to adjust the sensitivity of the quantization priority to the feature importance.

5. The real-time monitoring method for an aerospace parts processing line according to claim 4, characterized in that: Hyperparameters The method to obtain is: According to The feature importance of all blocks in the frame grayscale image clusters all blocks into two categories, and the mean of the feature importance of all blocks contained in the category is used as the representative feature importance of the category, and the mean of the representative feature importance of the two categories is used as value.

6. The real-time monitoring method for an aerospace parts processing line according to claim 1, characterized in that: The dynamically adjusting the quantization priority division threshold according to the current processing state and the resource state of the edge computing device includes: The processing status index is determined based on the change in the feature importance of each block in the current frame grayscale image relative to the feature importance of the corresponding block in the historical grayscale image. The system resource status index is determined in combination with the CPU usage, memory occupancy and bandwidth utilization of the edge computing device. The quantitative priority division threshold is dynamically adjusted based on the processing status index and the system resource status index.

7. The real-time monitoring method for an aerospace parts processing line according to claim 6, characterized in that: The processing state index satisfies the expression: ； in, Indicates the sequence number of the current frame, Indicates the Processing status indicators of frame grayscale images; Indicates the total number of blocks; Indicates the Frame grayscale image The feature importance of each block; and Respectively represent Blocks in the past The mean and standard deviation of feature importance in the frame grayscale image, Indicates the preset reference quantity; represents the hyperbolic tangent function; Indicates the absolute value symbol.

8. The real-time monitoring method for an aerospace parts processing line according to claim 6, characterized in that: The method for obtaining the system resource status indicator includes: Obtain the maximum value of the CPU usage, memory occupancy, and bandwidth utilization of the edge computing device at the current moment, subtract the maximum value from 1, and use the result as the system resource status indicator.

9. The real-time monitoring method for an aerospace parts processing line according to claim 6, characterized in that: The dynamically adjusting the division threshold of the quantization priority according to the processing status indicator and the system resource status indicator includes: The initial segmentation threshold is modified according to the processing state index and the system resource state index to obtain the instant segmentation threshold of the current frame grayscale image, and the instant segmentation threshold is weighted averaged with the historical dynamic segmentation threshold to obtain the dynamic segmentation threshold of the current frame grayscale image; The instant partition threshold satisfies the expression: ； in, Indicates the sequence number of the current frame, Indicates the Frame grayscale image An instant partition threshold; Indicates the An initial partition threshold; Indicates the Processing status indicators corresponding to the frame grayscale image; Indicates the system resource status indicators at the current moment; Indicates the normal state threshold of the processing process, Indicates the threshold for sufficient system resources. Indicates the processing state adjustment coefficient, Indicates the system resource adjustment coefficient.

10. The real-time monitoring method for an aerospace parts processing line according to claim 1, characterized in that: The step of compressing the blocks according to the quantized quantity of each block to obtain compressed data includes: When the quantization number of a block is C=1, the grayscale mean of all pixels in the block is calculated as the only quantization value, and no bitmap is generated; when the quantization number of a block is C=2, the high and low means are calculated as the quantization value, and a 1-bit / pixel bitmap is generated; when the quantization number of a block is C=3 or C=4, the K-means clustering algorithm is used to divide the pixels in the block into C clusters, and the grayscale mean of all pixels in each cluster is calculated as the quantization value, and a 2-bit / pixel bitmap is generated at the same time; when the quantization number of a block is C=5, the K-means clustering algorithm is used to divide the pixels in the block into 5 clusters, and the grayscale mean of all pixels in each cluster is calculated as the quantization value, and a 3-bit / pixel bitmap is generated at the same time, and the quantization values ​​of all blocks and the bitmap are combined into compressed data.

Images