Flat tube extrusion processing quality detection method and system

Abstract

The invention relates to the technical field of image processing, and discloses a flat tube extrusion processing quality detection method and system. The method comprises the following steps: extracting internal cluster density data of the flat tube through an electromagnetic image sequence, and mapping the internal cluster density data into a yield strength distribution image; geometric deviation images are synchronously obtained, and a comprehensive quality pseudo-color image is generated through fusion; and performing threshold segmentation according to the pixel average value to realize automatic quality grading. According to the method, the problems that the inner cluster hardening mechanism of the flat tube cannot be detected and multi-dimensional quality comprehensive evaluation is lacked in the prior art are solved, and the accuracy and predictability of quality detection of the flat tube are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a flat tube extrusion processing quality detection method and system. BACKGROUND

[0002] In the field of flat tube extrusion processing quality detection, the existing technology mainly relies on manual visual inspection or automatic measurement of a single physical quantity. The manual inspection method is highly subjective, inefficient and prone to missed detection, while the automatic measurement is usually limited to the detection of geometric dimensions (such as wall thickness and width) or the use of single frequency eddy current technology for surface defect scanning. These methods cannot penetrate the material interior and are difficult to quantitatively evaluate the key microstructure (such as cluster structure) that determines the mechanical properties of the flat tube.

[0003] The shortcomings of the existing technology lie in its one-sidedness and lag in detection dimension. First, the traditional method cannot achieve non-destructive real-time monitoring of the internal cluster hardening mechanism of the flat tube. Since the cluster density distribution is closely related to the yield strength of the material, this defect makes it impossible to predict the mechanical property distribution of the flat tube, and only destructive sampling can be performed after product formation, lacking predictability. Second, the existing method usually detects internal structure or external geometric dimensions in isolation, lacking a means to synchronously collect and fuse analyze multi-dimensional quality information (such as strength, geometric tolerance, and material uniformity), resulting in an incomplete quality evaluation system and ignoring the mutual influence between indicators, which may lead to misjudgment or missed judgment. Finally, even if multiple indicators are obtained, there is still a lack of a unified, quantitative comprehensive evaluation standard to automatically grade the product quality, still relying on manual experience, which is inefficient and inconsistent. SUMMARY

[0004] The present application provides a flat tube extrusion processing quality detection method and system to solve the problems of the inability to detect the internal cluster hardening mechanism of the flat tube and the lack of multi-dimensional quality comprehensive evaluation in the existing technology, improving the accuracy and predictability of flat tube quality detection.

[0005] In a first aspect, the present application provides a flat tube extrusion processing quality detection method, which comprises: S1 step: acquiring an electromagnetic response image sequence along the width direction of the flat tube, and extracting cluster density distribution image data reflecting the internal structure state of the material through image analysis; S2 step: performing pixel value mapping processing on the cluster density distribution image data to generate an enhanced contribution value distribution image and calculate a visualized image of the expected yield strength distribution of the flat tube; S3 step: synchronously acquiring a wall thickness variation image and a width variation image of the flat tube, and generating a two-dimensional image of geometric deviation data through image measurement algorithm; S4 step: the image fusion technology is used to superimpose the flat tube expected yield strength distribution image and the geometric deviation image, and a pseudo-color image of comprehensive quality evaluation is output; S5 step: pixel value statistical analysis and threshold segmentation are performed on the pseudo-color image, when the pixel average value is greater than or equal to 90, it is marked as a premium product area, when the pixel average value is greater than or equal to 70 and less than 90, it is marked as a qualified product area, and when the pixel average value is less than 70, it is marked as a waste product area.

[0006] In a second aspect, the application provides a flat tube extrusion processing quality detection system, which comprises: An image acquisition module is configured to acquire an electromagnetic response image sequence along the width direction of the flat tube, and extract a cluster density distribution image data reflecting the internal organization state of the material through image analysis; An image processing module is configured to perform pixel value mapping processing on the cluster density distribution image data, generate a strengthened contribution value distribution image, and calculate a visual image of the flat tube expected yield strength distribution; An image measurement module is configured to synchronously acquire a wall thickness change image and a width change image of the flat tube, and generate a two-dimensional image of geometric deviation data through an image measurement algorithm; An image fusion module is configured to use the image fusion technology to superimpose the flat tube expected yield strength distribution image and the geometric deviation image, and output a pseudo-color image of comprehensive quality evaluation; An image analysis module is configured to perform pixel value statistical analysis and threshold segmentation on the pseudo-color image, when the pixel average value is greater than or equal to 90, it is marked as a premium product area, when the pixel average value is greater than or equal to 70 and less than 90, it is marked as a qualified product area, and when the pixel average value is less than 70, it is marked as a waste product area.

[0007] In a third aspect, a flat tube extrusion processing quality detection device is provided, which comprises a memory and at least one processor, the memory stores instructions; the at least one processor invokes the instructions in the memory, so that the flat tube extrusion processing quality detection device executes the above-mentioned flat tube extrusion processing quality detection method.

[0008] In a fourth aspect, a computer readable storage medium is provided, which stores instructions, when running on a computer, makes the computer execute the above-mentioned flat tube extrusion processing quality detection method.

[0009] In the technical scheme provided in the application, by acquiring the electromagnetic response image sequence along the width direction of the flat tube, the cluster density distribution image data reflecting the internal organization state of the material is extracted through image analysis, the non-destructive and visual detection of the internal microstructure state of the flat tube is realized, the limitation that the traditional detection method cannot detect the internal cluster structure distribution is overcome, further, the pixel value mapping processing is performed on the cluster density distribution image data, the strengthened contribution value distribution image is generated, the visual image of the expected yield strength distribution of the flat tube is calculated, the quantitative relationship between the microstructure information and the macro mechanical property is established, the quality evaluation is upgraded from the appearance parameter detection to the performance prediction evaluation, and the foresight and accuracy of the detection are significantly improved. By synchronously acquiring the wall thickness change image and the width change image of the flat tube, the two-dimensional image of the geometric deviation data is generated through the image measurement algorithm, the time and space synchronization of the internal performance detection and the external geometric size detection is ensured, the misplacement and misjudgment caused by the asynchronous data acquisition are avoided, and a structural consistent information basis is provided for the multi-source data fusion.

[0010] When the image mapping, multi-source image fusion and threshold segmentation algorithm provided in the application is applied, the algorithm features make a core contribution to the scheme. By adopting the image fusion technology to perform superposition processing on the expected yield strength distribution image of the flat tube and the geometric deviation image, the pseudo-color image of the comprehensive quality evaluation is output, the multi-dimensional quality indexes such as the strength, the geometry and the uniformity are organically integrated, the one-sidedness of the single index evaluation is eliminated, and the comprehensive quality evaluation system is formed. Finally, the pixel value statistical analysis and the threshold segmentation are performed on the pseudo-color image, and the automatic grade judgment is performed according to the preset threshold, the objectivity, efficiency and consistency of the quality grading process are realized, the uncertainty and error caused by the human intervention are reduced, and the reliability and efficiency of the quality control are comprehensively improved. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical scheme of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creating labor based on these drawings.

[0012] Figure 1 An embodiment schematic diagram of the flat tube extrusion processing quality detection method in the embodiment of the application; Figure 2 An embodiment schematic diagram of the flat tube extrusion processing quality detection system in the embodiment of the application; Figure 3 An embodiment schematic diagram of the flat tube extrusion processing quality detection system in the embodiment of the application; DETAILED DESCRIPTION

[0013] The embodiment of the present application provides a flat pipe extrusion processing quality detection method and system. The terms "first", "second", "third", "fourth" and the like (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0014] For ease of understanding, the specific process of the embodiment of the present application is described below. Please refer to Figure 1 One embodiment of the flat pipe extrusion processing quality detection method in the embodiment of the present application comprises the following steps. S1 step: Obtain an electromagnetic response image sequence along the width direction of the flat pipe, and extract a cluster density distribution image data reflecting the internal organization state of the material through image analysis; S2 step: Perform pixel value mapping processing on the cluster density distribution image data to generate a reinforced contribution value distribution image, and calculate a visual image of the expected yield strength distribution of the flat pipe; S3 step: Synchronously obtain a wall thickness change image and a width change image of the flat pipe, and generate a two-dimensional image of geometric deviation data through an image measurement algorithm; S4 step: Superimpose the expected yield strength distribution image of the flat pipe and the geometric deviation image by using an image fusion technology, and output a pseudo-color image of comprehensive quality evaluation; S5 step: Perform pixel value statistical analysis and threshold segmentation on the pseudo-color image, mark the excellent product area when the pixel average value is greater than or equal to 90, mark the qualified product area when the pixel average value is greater than or equal to 70 and less than 90, and mark the waste product area when the pixel average value is less than 70.

[0015] It can be understood that the execution subject of the present application can be a flat pipe extrusion processing quality detection system, and can also be a terminal or a server, and the specific place is not limited. The embodiment of the present application takes the server as the execution subject for example.

[0016] Specifically, by converting the traditional physical detection signal into digital image data for analysis and processing, the comprehensive evaluation of the internal organization state and the appearance geometric characteristics of the flat tube is realized. The electromagnetic response image sequence acquisition process uses a high-frequency eddy current sensor array to scan along the width direction of the flat tube. The electromagnetic field generated by the sensor interacts with the internal cluster structure of the flat tube to produce eddy current response signals. These analog signals are converted into digital signals by an analog-to-digital converter and reconstructed into a two-dimensional image matrix. The gray value of each pixel point corresponds to the electromagnetic response intensity at that position. The image analysis algorithm identifies and classifies the spatial distribution pattern of the pixel gray value, extracts the feature parameters reflecting the change of the internal cluster density, and stores the cluster density distribution image data in the form of a pixel array. The pixel value range is usually an 8-bit gray image format with a value range of 0-255.

[0017] The pixel value mapping process uses a lookup table algorithm to establish the corresponding relationship between the cluster density pixel value and the strengthening contribution value. The mapping function is based on the material mechanics theory and experimental calibration data. Each pixel value in the cluster density image is converted into the corresponding strengthening contribution value through a linear or nonlinear transformation function. The transformation process takes into account parameters such as the matrix strength, cluster hardening coefficient, and temperature correction factor of the aluminum alloy material. The generated strengthening contribution value distribution image maintains the same spatial resolution and pixel arrangement structure as the original image. The visualization image of the expected yield strength distribution of the flat tube is obtained by performing pixel-level numerical operations on the strengthening contribution value and the material matrix strength. The operation result is converted into different color representations by using pseudo-color mapping technology. The red area represents the high strength area, and the blue area represents the low strength area.

[0018] The wall thickness change image and the width change image are obtained by using the distance data of the laser triangulation system. The laser ranging data of the upper and lower surfaces are obtained by geometric operation to obtain the wall thickness value. The width data is calculated after identifying the left and right boundary coordinates of the flat tube by the boundary detection algorithm. The image measurement algorithm converts these one-dimensional numerical sequences into two-dimensional image representations. The wall thickness change image constructs a pixel matrix with the length direction of the flat tube as the horizontal axis and the width direction as the vertical axis. The gray value of each pixel represents the wall thickness deviation at that position. The two-dimensional image of the geometric deviation data is generated by performing difference operation on the measured geometric parameters and the design standard value. The positive and negative nature and amplitude of the deviation value are directly mapped to the gray level of the pixel.

[0019] The image fusion technology adopts a weighted average algorithm to superimpose the intensity distribution image and the geometric deviation image at the pixel level. The fusion weight is determined according to the influence degree of the intensity information and the geometric information on the overall quality of the flat tube. The weight coefficient of the intensity distribution image is usually set to be between 0.5 and 0.7, and the weight coefficient of the geometric deviation image is the remaining part. In the superposition process, the pixel value of each corresponding position is calculated according to the weighted formula. The fusion result generates a pseudo-color image for comprehensive quality evaluation. The color coding scheme uses the HSV color space. The hue value represents the quality level, the saturation value represents the quality stability, and the brightness value represents the quality confidence.

[0020] When performing pixel value statistical analysis on the pseudo-color image, the RGB color value is first converted back to a numerical quality index. The algorithm calculates the pixel average value, standard deviation and distribution histogram in the entire image or a specified area. The threshold segmentation algorithm divides the image into different quality areas according to the preset quality level boundary value. When the pixel average value reaches a threshold value of 90 or above, the area is marked as premium and displayed in green. The pixel area between 70 and 90 is marked as qualified and displayed in yellow. The pixel area below 70 is marked as waste and displayed in red. The region marking process uses a connected component analysis algorithm to identify adjacent pixels of the same type to obtain quality classification areas.

[0021] In a specific embodiment, the S1 step further comprises: The electromagnetic response signals obtained by the high-frequency eddy current sensor array are converted into a gray-scale image sequence, and each sensor corresponds to a pixel column in the image; Based on three fixed frequencies of 500 kHz, 1 MHz and 2 MHz, multi-channel image processing is performed on the gray-scale image sequence to obtain electromagnetic characteristic parameter images of different penetration depths; The electromagnetic characteristic parameter images are input into a lookup table-based image mapping algorithm for pixel value conversion to obtain cluster density numerical images of each layer inside the flat tube; The cluster density numerical image is subjected to bilinear interpolation and Gaussian filtering to obtain a smooth and continuous cluster density distribution image data.

[0022] Specifically, the process of converting the electromagnetic response signals acquired by the high-frequency eddy current sensor array into a sequence of gray-scale images converts the analog voltage signals output by each sensor into digital signals through an analog-to-digital converter. The range of the converted digital signal values is usually between 0 and 4095, corresponding to a resolution of 12-bit ADC. Subsequently, a linear mapping function is used to map the digital signal values to a gray-scale level range of 0 to 255. The higher the signal intensity, the larger the corresponding gray-scale value. The spatial position of each sensor in the width direction of the flat tube directly corresponds to the column index in the image matrix. The consecutive sampling points in the time sequence correspond to the row index in the image matrix. Therefore, a two-dimensional gray-scale image data structure is obtained, with time as the vertical axis and spatial position as the horizontal axis. Each frame in the image sequence represents the electromagnetic response state of the entire sensor array at a specific time.

[0023] The multi-channel image processing constructs independent image channels based on electromagnetic signals at three fixed frequencies of 500 kHz, 1 MHz, and 2 MHz. The electromagnetic signals at each frequency have different material penetration depth characteristics. The low-frequency signal at 500 kHz has a shallow penetration depth, mainly reflecting the electromagnetic characteristics of the surface layer of the flat tube. The intermediate-frequency signal at 1 MHz penetrates the middle layer of the flat tube, and the high-frequency signal at 2 MHz can detect the organization state of the inner layer of the flat tube. The image data of the three frequency channels are processed simultaneously in a parallel manner for frequency domain analysis. The frequency domain analysis process converts the time-domain image signals into frequency-domain representations through fast Fourier transform, extracts the amplitude and phase information of each pixel position at the corresponding frequency, and recombines the amplitude and phase information to obtain complex pixel values. The real part of the complex number represents the resistance component, and the imaginary part represents the inductance component. The electromagnetic characteristic parameter images of the three frequency channels finally generate electromagnetic characteristic parameter images corresponding to different penetration depths.

[0024] The image mapping algorithm based on the lookup table converts the pixel values in the electromagnetic characteristic parameter image into cluster density values. The lookup table establishes a correspondence between the electromagnetic characteristic parameters and the cluster densities by pre-calibrating known cluster density samples. During the calibration process, the electromagnetic response characteristics of aluminum alloy samples with different cluster density levels are measured at each frequency. The measurement results are stored in a table form as a mapping reference. During the pixel value conversion process, the algorithm first extracts the complex value of the pixel to calculate its modulus as an electromagnetic response intensity indicator. Then, the closest calibration point to the current electromagnetic response intensity is searched in the lookup table. When the pixel value is between two calibration points, a linear interpolation method is used to calculate the corresponding cluster density value. The interpolation calculation considers the weight distribution of adjacent calibration points, with closer calibration points having larger weights. The cluster density calculation results of the three frequency channels are combined into a single cluster density value image through a depth-weighted average method. The weighting coefficients are determined based on the contribution of different depth layers to the overall performance of the flat tube.

[0025] The bilinear interpolation and Gaussian filtering process converts the discrete cluster density numerical image into a smooth continuous distribution data. The bilinear interpolation algorithm restores the numerical value of the missing or abnormal pixel points in the image. In the interpolation process, the four nearest neighbor pixels around the target pixel are selected as the reference points. The interpolation weight is calculated according to the distance relationship between the target pixel and the four reference points. The closer the reference point, the greater the weight. The interpolation calculation weights the cluster density values of the four reference points according to the weight ratio to obtain the cluster density value of the target pixel. The Gaussian filtering process convolves the interpolated image using a Gaussian kernel function. The standard deviation parameter of the Gaussian kernel controls the filtering strength. The larger the standard deviation, the more obvious the filtering effect but the more image details are lost. In the filtering process, the new value of each pixel is equal to the sum of the product of the pixel value and the corresponding Gaussian weight. The filtering operation removes the high-frequency noise components in the image while maintaining the main features of the cluster density distribution. The filtered image data obtains a smooth and continuous cluster density distribution image.

[0026] In a specific embodiment, the electromagnetic characteristic parameter image is input into a lookup table-based image mapping algorithm for pixel value conversion to obtain the cluster density numerical image of each layer inside the flat tube, including: The electromagnetic characteristic parameter image is subjected to pixel value operation according to the resistivity calculation formula to obtain the resistivity distribution image of each depth layer of the flat tube; A lookup table is established based on the calibration relationship curve between the cluster density and the resistivity to perform pixel value lookup table mapping on the resistivity distribution image to obtain an initial cluster density distribution image; The initial cluster density distribution image is subjected to pixel value correction processing according to the flat tube material composition correction coefficient to obtain a corrected cluster density distribution image; The corrected cluster density distribution image is subjected to weighted image synthesis calculation according to the weight proportion of the inner and outer layers of the flat tube to obtain the cluster density numerical image of each layer inside the flat tube.

[0027] Specifically, the process of performing pixel value operation on the electromagnetic characteristic parameter image according to the resistivity calculation formula converts the complex electromagnetic response data into resistivity values. The resistivity calculation is based on Ohm's law and electromagnetic field theory. The complex value of each pixel includes real and imaginary parts. The real part represents the resistive response, and the imaginary part represents the inductive response. In the resistivity calculation process, the modulus of the complex impedance is first calculated as the total impedance size. Then, the corresponding resistivity value is calculated by combining the geometric parameters of the sensor and the excitation frequency. In the calculation, the geometric factors such as the diameter, the number of turns, and the distance from the flat tube surface of the sensor coil need to be considered, and the frequency-dependent skin effect needs to be compensated. The resistivity distribution image encodes the resistivity value of each pixel position according to the gray level. The higher the resistivity value, the brighter the pixel gray level. A two-dimensional image data reflecting the spatial distribution of the resistivity of each depth layer of the flat tube is obtained.

[0028] The look-up table is established based on a calibration curve of cluster density and resistivity, which is obtained by measuring a series of aluminum alloy standard samples with known cluster density, covering a complete range from low cluster density to high cluster density. The cluster density value of each sample is accurately determined by metallographic microscope observation and image analysis method, and the corresponding resistivity value is measured by four-probe method or eddy current detection method. The calibration data points form a nonlinear curve relationship in the resistivity-cluster density coordinate system. The curve fitting is mathematically modeled by a polynomial function or an exponential function, and the fitting parameters are determined by least squares optimization. The look-up table discretizes the continuous fitting curve into a finite number of data points, with the table index being the resistivity value and the table content being the corresponding cluster density value. In the pixel value look-up table mapping process, the algorithm reads the resistivity value of each pixel in the resistivity distribution image, searches for the closest index item in the look-up table, and calculates the corresponding cluster density value using linear interpolation method when the pixel resistivity value is between two table indexes. The interpolation calculation assigns weight coefficients according to the distance relationship between the pixel resistivity value and the adjacent index items, and finally obtains the initial cluster density distribution image, in which each pixel value directly represents the cluster density at that position.

[0029] The pixel value correction processing of the material composition correction coefficient considers the influence of different composition characteristics of A1100 and A3102 aluminum alloys on the cluster density-resistivity relationship. A1100 aluminum alloy has high purity aluminum content and a small amount of copper and silicon impurities, and A3102 aluminum alloy contains manganese as the main alloying element. Different composition elements have different effects on the conductivity and cluster density mechanism. The correction coefficient is determined according to the alloy chemical composition analysis results. The correction coefficient of aluminum element is set as the reference value. Copper element will reduce the resistivity measurement value due to its good conductivity, and the corresponding correction coefficient is greater than the reference value. The intermetallic compound of manganese element increases the resistivity, and the corresponding correction coefficient is less than the reference value. Silicon element changes the semiconductor characteristics at high temperature to affect the resistivity, and the correction coefficient is dynamically adjusted according to the temperature condition. The correction processing performs multiplication operation on each pixel value of the initial cluster density distribution image. The correction coefficient of each pixel position in the operation is determined according to the material composition distribution at that position. The corrected cluster density distribution image eliminates the systematic error of alloy composition difference on cluster density measurement.

[0030] The weighted image synthesis calculation fuses the corrected cluster density distribution image according to the weight proportion of the inner and outer layers of the flat tube, and the determination of the weight proportion is based on the structural characteristics of the flat tube and the application requirements. The inner layer of the flat tube directly contacts the fluid medium and bears a large pressure and temperature change, so the cluster density of the inner layer has a greater impact on the overall performance of the flat tube, and the corresponding weight coefficient is higher. The middle layer provides structural strength as a bearing layer, and the weight coefficient is set to a medium level. The outer layer plays a protective role, and the weight coefficient is relatively low. The weighted synthesis algorithm performs weighted average calculation on the pixel values of different depth layers at the same spatial position. In the calculation process, the pixel value of the inner layer is multiplied by the weight coefficient of the inner layer, the pixel value of the middle layer is multiplied by the weight coefficient of the middle layer, and the pixel value of the outer layer is multiplied by the weight coefficient of the outer layer. The sum of the three weighted results is the comprehensive cluster density value at that position. The normalization processing of the weight coefficient ensures that the sum of the three weights is equal to 1 to avoid numerical deviation. The generation process of the cluster density value image arranges the weighted synthesis results according to the spatial position to obtain a two-dimensional array structure. The row index of the array corresponds to the length direction position of the flat tube, and the column index corresponds to the width direction position. The array element value represents the comprehensive cluster density value at the corresponding position.

[0031] In a specific embodiment, the S2 step further comprises: The cluster density distribution image data is subjected to pixel value polynomial calculation according to the first-order and second-order coefficients of the cluster density to obtain a basic reinforcement contribution value image; According to the temperature correction factor and the strain correction factor, the basic reinforcement contribution value image is subjected to pixel value linear correction processing according to the extrusion temperature and deformation amount parameters of the flat tube to obtain a corrected reinforcement contribution value image; Based on the material matrix strength value and the corrected reinforcement contribution value image, pixel value addition operation is performed to obtain a local yield strength image of each position of the flat tube; The local yield strength image is subjected to image reconstruction processing according to the length and width coordinates of the flat tube to obtain a complete two-dimensional image of the expected yield strength distribution of the flat tube.

[0032] Specifically, the cluster density distribution image data is subjected to a pixel value polynomial calculation process according to a cluster density first-order term coefficient and a second-order term coefficient, thereby establishing a quantitative mathematical relationship between the cluster density and the strengthening effect. The first-order term coefficient reflects the linear contribution of the increase in the cluster density to the material strengthening, and the second-order term coefficient embodies the nonlinear variation characteristics of the strengthening effect after the cluster density reaches a critical value. The polynomial calculation performs mathematical operations on the cluster density values of each pixel position in the image. In the operation process, the pixel value is first multiplied by the first-order term coefficient to obtain a linear strengthening component, and then the square of the pixel value is multiplied by the second-order term coefficient to obtain a nonlinear strengthening component. The sum of the two components is the basic strengthening contribution value of the pixel position. The first-order term coefficient is usually determined according to the basic strengthening characteristics of the aluminum alloy material, reflecting the direct proportional relationship between the cluster density and the strengthening effect. The second-order term coefficient takes into account the cluster aggregation effect and the interaction mechanism. When the cluster density exceeds the optimal value, the strengthening effect will saturate or even decrease. The basic strengthening contribution value image maintains the same pixel arrangement and spatial resolution as the input cluster density image. Each pixel value in the image represents the strength increment generated by the cluster hardening mechanism at the corresponding position.

[0033] The pixel value linear correction processing of the temperature correction factor and the strain correction factor takes into account the influence of the flat tube extrusion process parameters on the cluster hardening effect. The extrusion temperature affects the cluster obtaining dynamics and thermal stability. High temperature leads to loose cluster structure and reduces the strengthening effect. Low temperature limits the cluster obtaining process. The temperature correction factor is determined according to the deviation of the actual extrusion temperature from the optimal cluster obtaining temperature. The larger the deviation, the smaller the correction factor. The deformation amount parameter includes the strain rate and the total strain amount in the extrusion process. The strain rate affects the uniformity and refinement degree of the cluster distribution. The total strain amount affects the cluster density and size distribution. The strain correction factor is determined based on the strain hardening characteristics of the material and the cluster response mechanism. High strain rate promotes the refinement and uniform distribution of the cluster, corresponding to a larger correction factor. Excessive strain amount destroys the obtained cluster structure, corresponding to a smaller correction factor. The linear correction processing multiplies each pixel value in the basic strengthening contribution value image by the temperature correction factor and the strain correction factor in turn. The correction process uses continuous multiplication operations to ensure that the cumulative effect of each correction factor is accurately reflected. The corrected strengthening contribution value image reflects the actual strengthening effect distribution under the influence of the process parameters.

[0034] The pixel value addition operation of the material matrix strength value and the modified strengthening contribution value image combines the intrinsic strength of the aluminum alloy matrix material and the additional strength generated by the cluster hardening. The matrix strength value represents the yield strength of the aluminum alloy in the state without clusters. This value is determined according to the alloy composition, grain size, and basic heat treatment state. The addition operation uses pixel-level value superposition. The matrix strength value is added to each pixel value in the modified strengthening contribution value image. In the operation process, the interaction effects between different strengthening mechanisms need to be considered. There is a synergistic or competitive relationship between cluster hardening and solid solution hardening, dislocation hardening, etc. When cluster hardening dominates, the contribution of other strengthening mechanisms is relatively small. When the cluster hardening effect is weak, the compensating effect of other strengthening mechanisms needs to be considered. The addition operation result generates a local yield strength image. Each pixel value in the image represents the total yield strength prediction value at the corresponding position. The spatial distribution of pixel values reflects the change law of the carrying capacity at different positions of the flat tube. High pixel value areas represent high strength areas, and low pixel value areas represent relatively weak links.

[0035] The image reconstruction process rearranges the local yield strength image according to the actual geometric coordinates of the flat tube to obtain a two-dimensional strength distribution image. The reconstruction process first establishes the coordinate mapping relationship between the length direction and the width direction of the flat tube. The length direction coordinate corresponds to the time sequence in the detection process, and the width direction coordinate corresponds to the spatial position of the sensor array. The coordinate mapping converts the one-dimensional detection data sequence into a two-dimensional spatial distribution representation. The reconstruction algorithm determines the accurate spatial position of each data point according to the geometric configuration of the detection equipment and the moving trajectory of the flat tube. When there is a spatial interval or missing in the detection data, an interpolation algorithm is used for data supplementation. The interpolation methods include nearest neighbor interpolation, bilinear interpolation, and cubic spline interpolation, etc. The interpolation selection is determined according to the data density and accuracy requirements. The complete two-dimensional image of the expected yield strength distribution of the flat tube is stored and displayed in matrix form. The matrix row index corresponds to the position in the length direction of the flat tube, and the column index corresponds to the position in the width direction. The matrix element value represents the yield strength prediction value at the corresponding position. The image display uses a pseudo-color coding scheme to convert the strength value into an intuitive color representation.

[0036] In a specific embodiment, the S3 step further comprises: Converting the distance data obtained by the laser triangulation system into a depth image, and performing three-dimensional reconstruction on the surface of the flat tube; Based on the depth image, performing pixel value operation according to the difference between the upper surface depth and the lower surface depth to obtain a wall thickness distribution image of the flat tube at each position; According to the edge detection algorithm, the contour of the flat tube is identified, and the width distribution image of the flat tube at each position is calculated; Performing pixel value difference operation on the wall thickness distribution image and the width distribution image of the flat tube at each position with the design standard image respectively to obtain a two-dimensional image of geometric deviation data.

[0037] Specifically, the process of converting the distance data obtained by the laser triangulation system into a depth image reconstructs the one-dimensional distance sequence output by the laser ranging sensor into a two-dimensional spatial representation. The laser triangulation principle calculates the distance value based on the positional offset of the reflected light on the CCD sensor after the laser beam is irradiated onto the surface of the flat tube. The upper and lower sets of laser measurement devices correspond to the upper surface and the lower surface of the flat tube, respectively. Each set of devices includes a laser emitter and a linear array CCD receiver. The linear light beam generated by the laser emitter is vertically irradiated onto the surface of the flat tube to form a light spot. The reflected light is captured by the CCD receiver and converted into an electrical signal. The signal processing circuit converts the analog signal output by the CCD into a digital distance value. The distance data is constructed into a two-dimensional depth image according to the moving direction of the flat tube and the laser scanning direction. The row coordinates of the image correspond to the position in the length direction of the flat tube, and the column coordinates correspond to the position in the width direction. The pixel value represents the depth distance value at the corresponding position. The three-dimensional reconstruction process converts the depth image into three-dimensional point cloud data in combination with the geometric calibration parameters of the laser measurement device. The point cloud data contains the three-dimensional spatial coordinate information of each measurement point on the surface of the flat tube. The pinhole camera model and the laser plane equation are used to calculate the spatial coordinates in the reconstruction algorithm. The pixel resolution of the depth image determines the accuracy and detail performance of the three-dimensional reconstruction.

[0038] The process of calculating the wall thickness distribution image based on the corresponding pixels of the upper and lower surface depth images involves numerical subtraction operations. The wall thickness calculation principle considers that the actual wall thickness at a certain position on the flat tube is equal to the distance from the upper surface of that position to the laser reference plane minus the distance from the lower surface to the reference plane. The pixel value operation uses a pixel-by-pixel traversal method to synchronously process the two depth images. In the operation process, the algorithm first ensures that the upper and lower surface depth images have the same pixel size and spatial correspondence. Then, a subtraction operation is performed for each pixel position. The operation formula is that the wall thickness pixel value is equal to the upper surface depth pixel value minus the lower surface depth pixel value. When there is a spatial registration error between the upper and lower surface images, image correction and geometric transformation are needed. The correction process includes rotation transformation, translation transformation, and scale transformation. The transformation parameters are obtained by measuring the calibration object. Each pixel value in the wall thickness distribution image directly represents the wall thickness value at the corresponding position. The unit of the pixel value is consistent with the measurement unit of the laser ranging system. The gray level of the image reflects the range of wall thickness changes. High gray values correspond to thick wall regions, and low gray values correspond to thin wall regions.

[0039] The process of contour recognition by the edge detection algorithm detects the gray level mutation features of the flat tube boundary using a gradient operator. The edge detection is based on the depth difference between the flat tube region and the background region in the depth image. The depth value of the flat tube boundary position will change sharply to form obvious gradient features. The Sobel operator, as a commonly used edge detection algorithm, calculates the amplitude and direction of the image gradient through convolution operation. The operator includes convolution kernels in the horizontal and vertical directions to detect the edge features in the x and y directions, respectively. The gradient amplitude is calculated by taking the square root of the sum of the horizontal and vertical gradients. The gradient direction is calculated by the arctangent function to obtain the normal angle of the edge. The binary image after edge detection is segmented by thresholding to convert the gradient amplitude into edge pixels and non-edge pixels. The contour recognition algorithm searches for continuous edge pixels in the binary image to form a complete boundary contour. The contour tracking uses an eight-connected domain search method to traverse along the connected path of the edge pixels. The width is calculated by identifying the left and right boundary contours and then calculating the distance between the boundaries at the corresponding positions. The distance calculation considers the proportional conversion from pixel coordinates to actual physical distance. The width distribution image arranges the width values at each length position according to the image rows and columns to form a two-dimensional distribution representation.

[0040] The process of generating a geometric deviation data image by pixel value difference operation compares the measured geometric parameters with the design standard pixel by pixel. The design standard image contains the wall thickness distribution and width distribution data of the flat tube in the ideal state. The standard image is generated based on the design specifications and manufacturing requirements of the flat tube. Each pixel value in the wall thickness standard image represents the design wall thickness value at the corresponding position, and each pixel value in the width standard image represents the design width value at the corresponding position. The difference operation calculates the deviation by subtracting the standard image pixel value from the measured image pixel value. The deviation calculation result includes positive deviation and negative deviation. Positive deviation indicates that the measured value exceeds the design standard, and negative deviation indicates that the measured value is lower than the design standard. The wall thickness deviation image and the width deviation image reflect the manufacturing accuracy of the flat tube in the thickness direction and the width direction, respectively. The pixel value range of the deviation image is usually symmetrically distributed from negative to positive. A zero value pixel indicates that the measured value completely matches the design value. The two-dimensional image of geometric deviation data integrates the wall thickness deviation and width deviation information. The integration method includes a multi-channel image format or a synthesized single-channel image format. In the multi-channel format, the wall thickness deviation is used as one color channel, and the width deviation is used as another color channel. In the synthesized format, the two deviations are weighted and averaged according to the weight coefficient to obtain a comprehensive deviation index.

[0041] In a specific embodiment, the S4 step further comprises: The flat tube expected yield strength distribution image is standardized according to the strength quality index calculation rule to obtain a strength quality distribution image. The two-dimensional image of the geometric deviation data is subjected to image standardization processing according to a geometric quality index calculation rule to obtain a geometric quality distribution image. The pixel variance value of the flat tube expected yield strength distribution image is subjected to statistical analysis processing according to a material uniformity index calculation rule to obtain a material uniformity distribution image. The strength quality distribution image, the geometric quality distribution image and the material uniformity distribution image are subjected to weighted image fusion calculation according to preset weight coefficients to obtain a pseudo-color image of comprehensive quality evaluation.

[0042] Specifically, the process of image standardization processing of the flat tube expected yield strength distribution image according to the strength quality index calculation rule converts the original strength value into a unified quality evaluation index, the strength quality index calculation rule is determined based on the design strength requirement and application standard of the flat tube, the calculation rule contains two dimensions of strength compliance evaluation and strength consistency evaluation, the strength compliance evaluation compares the expected yield strength at each pixel position with the design requirement strength, when the measured strength is equal to or exceeds the design requirement, a full score is obtained, when the measured strength is lower than the design requirement, the score is deducted according to the deviation degree, the deduction function adopts a piecewise linear or exponential decay model, the strength consistency evaluation statistics the distribution uniformity of the strength value in the entire image area, the consistency level is evaluated by calculating the coefficient of variation or standard deviation of the strength value, the image standardization processing maps the strength quality evaluation result to the standard score interval of zero to one hundred using the minimum-maximum normalization method, the normalization formula is that the standardized value is equal to the original value minus the minimum value divided by the difference between the maximum value and the minimum value, and then multiplied by one hundred, each pixel value in the strength quality distribution image represents the strength quality score at the corresponding position, the higher the pixel value, the better the strength quality at the position.

[0043] The process of image standardization processing of the two-dimensional image of the geometric deviation data according to the geometric quality index calculation rule converts the geometric deviation into a quality evaluation score, the geometric quality index calculation rule is established on the basis of the concept of tolerance band, the tolerance band defines the allowed deviation range of the geometric dimension, the deviation exceeding the tolerance band is considered as a quality defect, the calculation rule contains two sub-items of wall thickness quality evaluation and width quality evaluation, the wall thickness quality evaluation compares the measured wall thickness deviation with the wall thickness tolerance band, the position with deviation within the tolerance band gets a full score, the position with deviation exceeding the tolerance band is deducted according to the exceeding degree, the deduction rule adopts a piecewise function model, the deduction is less for slight exceeding and the deduction is more for serious exceeding, the width quality evaluation adopts the same scoring logic to evaluate the quality of the width deviation, the comprehensive calculation of the geometric quality index performs weighted average of the wall thickness quality evaluation and the width quality evaluation according to preset weights, the weight distribution is determined according to the influence degree of the wall thickness and the width on the performance of the flat tube, the image standardization processing also maps the geometric quality evaluation result to the standard score interval of zero to one hundred, the geometric quality distribution image reflects the geometric manufacturing precision of the flat tube at each position.

[0044] The process of statistically analyzing pixel variance values ​​according to the material uniformity index calculation rules quantifies the uniformity of the strength distribution inside the flat tube. Material uniformity evaluation is based on the concept of variance in statistics. Variance values ​​reflect the degree of dispersion of data from the mean. The smaller the variance, the more concentrated the data distribution and the better the material uniformity. The larger the variance, the more dispersed the data distribution and the worse the material uniformity. The statistical analysis first calculates the average value of all pixel values ​​in the expected yield strength distribution image of the flat tube as the central tendency index of the strength distribution. Then, it calculates the deviation of each pixel value from the mean value, squares the deviation, and averages the results to obtain the pixel variance value. The material uniformity index calculation rules convert the variance value into a uniformity score. The conversion relationship uses an inverse proportional function. The smaller the variance value, the higher the uniformity index. The larger the variance value, the lower the uniformity index. The conversion function includes smoothing and boundary constraints to avoid excessive influence of extreme values ​​on the evaluation results. The material uniformity distribution image represents the uniformity score according to spatial location. Each pixel value in the image represents the material uniformity level of a local area centered on that pixel. The size of the local area is determined according to the spatial scale requirements of the uniformity evaluation.

[0045] Weighted image fusion calculation numerically synthesizes three quality distribution images according to preset weight coefficients. The weight coefficient allocation reflects the importance of strength quality, geometric quality, and material homogeneity to the overall quality of the flat tube. The weight allocation needs to be determined according to the specific application scenario and performance requirements of the flat tube. In pressure-bearing applications, strength quality has a higher weight, in precision applications, geometric quality has a higher weight, and in long-term use, material homogeneity has a higher weight. The weighted fusion algorithm calculates a weighted average of corresponding pixels in the three quality distribution images. During the calculation, the strength quality score at each pixel location is multiplied by the strength weight coefficient, and the geometric quality score is multiplied by the geometric weight coefficient. The material uniformity score is multiplied by the uniformity weighting coefficient, and the sum of the three weighted results yields the comprehensive quality score for that location. The normalization of the weighting coefficients ensures that the sum of the three weights equals one, avoiding scoring bias caused by incorrect weight settings. The pseudo-color image generation process converts the comprehensive quality score into an intuitive color representation. The color mapping uses the HSV color space, where hue value represents the quality level, saturation represents quality stability, and luminance value represents quality confidence. High-quality areas are displayed as green or blue, medium-quality areas as yellow, and low-quality areas as red. The pseudo-color coding scheme facilitates the rapid identification and location of quality defects.

[0046] In one specific embodiment, step S5 further includes: The pseudo-color image of the comprehensive quality evaluation is compared with the preset superior product threshold of 90 by pixel value comparison to obtain the binarized image of the superior product area. The pixel value of the pixel region not reaching the premium product standard is compared with a preset qualified product threshold 70 to obtain a binary image of the qualified product region; A logic operation is performed based on the binary image of the premium product region and the binary image of the qualified product region to obtain a color labeling image of the flat tube quality grade classification; The color labeling image of the flat tube quality grade classification is image registered with the flat tube position coordinates to obtain a quality grading visualization image with spatial position information.

[0047] Specifically, in the flat tube extrusion processing quality detection method based on image processing, after the pseudo-color image of comprehensive quality evaluation is generated, a series of image processing operations are required to realize the visualization classification and spatial positioning of the quality grade. First, the RGB value of each pixel in the pseudo-color image is converted back to the corresponding comprehensive quality score value, which has been calculated by the weighted fusion of the intensity quality, geometric quality and material uniformity, and the value ranges between 0 and 100. Then, the quality score value of each pixel is compared with a preset premium product threshold 90: if the score is greater than or equal to 90, the pixel is marked as 1 in the premium product binary image, otherwise it is marked as 0. Similarly, for the pixel region not reaching the premium product standard, it is compared with the qualified product threshold 70: if the score is between 70 and 90, it is marked as 1 in the qualified product binary image, otherwise it is marked as 0. At this time, the waste product region is the pixel with a score lower than 70, and a separate binary image is not needed, which can be indirectly identified by logic operation.

[0048] The two binary images are subjected to a logic operation to generate a color labeling image of the quality grade classification. Specifically, the pixel-level logic "or" operation is used to merge the premium product and qualified product regions, and different colors are assigned through color mapping rules: the premium product region is assigned green (RGB: 0, 255, 0), the qualified product region is assigned yellow (RGB: 255, 255, 0), and the waste product region is assigned red (RGB: 255, 0, 0). This process traverses each pixel position to determine the final color based on its value in different binary images, forming a color labeling image with three color regions. To realize spatial positioning of quality defects, the color labeling image needs to be registered with the physical position coordinates of the flat tube. The registration process is based on the geometric calibration parameters of the detection system to establish a mapping relationship between the image pixel coordinates and the actual position of the flat tube. Specifically, through an affine transformation or perspective transformation model, the row index in the image is mapped to the position in the length direction of the flat tube, and the column index is mapped to the position in the width direction. The transformation parameters are calculated by the sensor installation position and the flat tube moving speed. After registration, each color region has actual spatial coordinate information, and the final quality grading visualization image can be directly used for precise positioning and tracing of defect positions on the production line.

[0049] The flat tube extrusion processing quality detection method in the embodiments of the present application is described above, and the flat tube extrusion processing quality detection system in the embodiments of the present application is described below. Please refer to Figure 2 One embodiment of the flat tube extrusion processing quality detection system in the embodiments of the present application includes: An image acquisition module is configured to acquire a sequence of electromagnetic response images along the width direction of the flat tube, and extract cluster density distribution image data reflecting the internal organization state of the material through image analysis; An image processing module is configured to perform pixel value mapping processing on the cluster density distribution image data to generate a reinforced contribution value distribution image and calculate a visual image of the expected yield strength distribution of the flat tube; An image measurement module is configured to synchronously acquire a wall thickness variation image and a width variation image of the flat tube, and generate a two-dimensional image of geometric deviation data through an image measurement algorithm; An image fusion module is configured to superimpose the expected yield strength distribution image of the flat tube and the geometric deviation image by using an image fusion technology, and output a pseudo-color image of comprehensive quality evaluation; An image analysis module is configured to perform pixel value statistical analysis and threshold segmentation on the pseudo-color image, and mark a superior product region when the average pixel value is greater than or equal to 90, mark a qualified product region when the average pixel value is greater than or equal to 70 and less than 90, and mark a waste product region when the average pixel value is less than 70. The above Figure 2 The flat tube extrusion processing quality detection system in the embodiments of the present application is described in detail from the perspective of modular functional entities, and the flat tube extrusion processing quality detection device in the embodiments of the present application is described in detail from the perspective of hardware processing.

[0050] Refer to Figure 3 In the embodiments of the present application, a flat tube extrusion processing quality detection device is also provided. The flat tube extrusion processing quality detection device can be a server, and its internal structure can be as shown in Figure 3 The flat tube extrusion processing quality detection device includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. The processor of the computer is designed to provide computing and control capabilities. The memory of the flat tube extrusion processing quality detection device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the flat tube extrusion processing quality detection device is used to store the corresponding data in the embodiments. The network interface of the flat tube extrusion processing quality detection device is used to communicate with the external terminal through network connection. The computer program is executed by the processor to implement the above method.

[0051] Those skilled in the art can understand that,Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the flat tube extrusion processing quality detection equipment to which the scheme of the present application is applied.

[0052] The present application also provides a computer readable storage medium, which can be a non-volatile computer readable storage medium, and can also be a volatile computer readable storage medium. The computer readable storage medium stores instructions, and when the instructions are run on a computer, the computer executes the steps of the flat tube extrusion processing quality detection method.

[0053] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, system and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0054] If the integrated unit is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such understanding, the technical scheme of the present application or the whole or part of the technical scheme that essentially contributes to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a flat tube extrusion processing quality detection device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The foregoing storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0055] The above embodiments are only used to illustrate the technical scheme of the present application, but not to limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical scheme recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical scheme deviate from the spirit and scope of the technical scheme of each embodiment of the present application.

Claims

1. An image processing-based flat tube extrusion processing quality detection method, characterized by, The method comprises: S1 step: obtaining an electromagnetic response image sequence along the width direction of the flat tube, and extracting a cluster density distribution image data reflecting the internal organization state of the material through image analysis; S2 step: performing pixel value mapping processing on the cluster density distribution image data to generate a strengthened contribution value distribution image, and calculating a visual image of the flat tube expected yield strength distribution; S3 step: synchronously obtaining a wall thickness change image and a width change image of the flat tube, and generating a two-dimensional image of geometric deviation data through image measurement algorithm; S4 step: adopting image fusion technology to superimpose the flat tube expected yield strength distribution image and the geometric deviation image, and outputting a pseudo-color image of comprehensive quality evaluation; S5 step: performing pixel value statistical analysis and threshold segmentation on the pseudo-color image, marking as excellent product area when the pixel average value is greater than or equal to 90, marking as qualified product area when the pixel average value is greater than or equal to 70 and less than 90, and marking as waste product area when the pixel average value is less than 70.

2. The image processing-based flat tube extrusion processing quality detection method according to claim 1, characterized by, The S1 step further comprises: The electromagnetic response signals obtained by the high-frequency eddy current sensor array are converted into a gray image sequence, and each sensor corresponds to a pixel column in the image; Based on three fixed frequencies of 500 kHz, 1 MHz and 2 MHz, multi-channel image processing is performed on the gray image sequence to obtain electromagnetic characteristic parameter images of different penetration depths; The electromagnetic characteristic parameter images are input into the image mapping algorithm based on the lookup table for pixel value conversion to obtain the cluster density numerical image of each layer inside the flat tube; The cluster density numerical image is subjected to bilinear interpolation and Gaussian filtering processing to obtain a smooth and continuous cluster density distribution image data.

3. The image processing-based flat tube extrusion processing quality detection method according to claim 2, characterized by, The electromagnetic characteristic parameter images are subjected to pixel value operation according to the resistivity calculation formula to obtain the resistivity distribution image of each depth layer of the flat tube; A lookup table is established based on the calibration relationship curve between cluster density and resistivity, and the resistivity distribution image is subjected to pixel value lookup table mapping to obtain an initial cluster density distribution image; The initial cluster density distribution image is subjected to pixel value correction processing according to the flat tube material composition correction coefficient to obtain a corrected cluster density distribution image; The corrected cluster density distribution image is subjected to weighted image synthesis calculation according to the weight proportion of the inner and outer layers of the flat tube to obtain the cluster density numerical image of each layer inside the flat tube. The S2 step further comprises:

4. The image processing-based flat tube extrusion processing quality detection method according to claim 1, characterized by, The cluster density distribution image data is subjected to pixel value polynomial calculation according to the first-order and second-order coefficients of the cluster density to obtain a basic strengthening contribution value image; The basic strengthening contribution value image is subjected to pixel value linear correction processing according to the temperature correction factor and the strain correction factor of the flat tube extrusion temperature and deformation parameters to obtain a corrected strengthening contribution value image; Pixel value addition operation is performed on the material matrix strength value and the corrected strengthening contribution value image to obtain a local yield strength image of the flat tube at each position; ​ The local yield strength image is subjected to image reconstruction processing according to the length and width coordinates of the flat tube, so as to obtain a complete two-dimensional image of the expected yield strength distribution of the flat tube.

5. The image processing-based flat tube extrusion processing quality detection method according to claim 1, characterized by, The S3 step further comprises: The distance data obtained by the laser triangulation system is converted into a depth image, and the surface of the flat tube is three-dimensionally reconstructed; Based on the depth image, pixel value operation is performed according to the depth of the upper surface minus the depth of the lower surface, so as to obtain a wall thickness distribution image of each position of the flat tube; According to an edge detection algorithm, the contour of the flat tube is recognized, and a width distribution image of each position of the flat tube is calculated; The wall thickness distribution image and the width distribution image of each position of the flat tube are subjected to pixel value difference operation with a design standard image respectively, so as to obtain a two-dimensional image of geometric deviation data.

6. The image processing-based flat tube extrusion processing quality detection method according to claim 5, characterized by, The S4 step further comprises: The expected yield strength distribution image of the flat tube is subjected to image standardization processing according to the strength quality index calculation rule, so as to obtain a strength quality distribution image; The two-dimensional image of the geometric deviation data is subjected to image standardization processing according to the geometric quality index calculation rule, so as to obtain a geometric quality distribution image; Based on the pixel variance value of the expected yield strength distribution image of the flat tube, statistical analysis processing is performed according to the material uniformity index calculation rule, so as to obtain a material uniformity distribution image; The strength quality distribution image, the geometric quality distribution image and the material uniformity distribution image are subjected to weighted image fusion calculation according to a preset weight coefficient, so as to obtain a pseudo-color image of comprehensive quality evaluation.

7. The image processing-based flat tube extrusion processing quality detection method according to claim 6, characterized by, The S5 step further comprises: The pseudo-color image of comprehensive quality evaluation is subjected to pixel value comparison operation with a preset excellent product threshold value 90, so as to obtain a binary image of an excellent product region; The pixel regions that do not reach the excellent product standard are subjected to pixel value comparison operation with a preset qualified product threshold value 70, so as to obtain a binary image of a qualified product region; Based on the binary image of the excellent product region and the binary image of the qualified product region, logical operation is performed, so as to obtain a color labeling image of flat tube quality grade classification; The flat tube quality grade classification color labeling image is subjected to image registration processing with the flat tube position coordinates, so as to obtain a quality grading visualized image with spatial position information.

8. An image processing-based flat tube extrusion processing quality detection system, characterized by, A system for implementing the image processing-based flat tube extrusion processing quality detection method according to any one of claims 1-7, the system comprising: An image acquisition module for acquiring an electromagnetic response image sequence along the width direction of the flat tube, and extracting cluster density distribution image data reflecting the internal organization state of the material through image analysis; An image processing module for performing pixel value mapping processing on the cluster density distribution image data, generating a reinforcement contribution value distribution image, and calculating a visualized image of the expected yield strength distribution of the flat tube; An image measurement module for synchronously acquiring a wall thickness change image and a width change image of the flat tube, and generating a two-dimensional image of geometric deviation data through an image measurement algorithm; An image fusion module for superimposing the expected yield strength distribution image of the flat tube and the geometric deviation image by using an image fusion technology, and outputting a pseudo-color image of comprehensive quality evaluation; An image acquisition module for acquiring an electromagnetic response image sequence along the width direction of the flat tube, and extracting cluster density distribution image data reflecting the internal organization state of the material through image analysis; An image processing module for performing pixel value mapping processing on the cluster density distribution image data, generating a reinforcement contribution value distribution image, and calculating a visualized image of the expected yield strength distribution of the flat tube; An image measurement module for synchronously acquiring a wall thickness change image and a width change image of the flat tube, and generating a two-dimensional image of geometric deviation data through an image measurement algorithm; An image fusion module for superimposing the expected yield strength distribution image of the flat tube and the geometric deviation image by using an image fusion technology, and outputting a pseudo-color image of comprehensive quality evaluation; An image analysis module is used to perform pixel value statistical analysis and threshold segmentation on the pseudo-color image, and when the pixel average value is greater than or equal to 90, it is marked as a premium product area, when the pixel average value is greater than or equal to 70 and less than 90, it is marked as a qualified product area, and when the pixel average value is less than 70, it is marked as a waste product area.

9. An image processing-based flat tube extrusion processing quality detection apparatus, characterized by, The computer program is stored in the memory and can be run on the processor, and the processor executes the computer program to realize the image processing-based flat tube extrusion processing quality detection method in any one of claims 1 to 7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, causes the processor to perform the image processing-based flat tube extrusion processing quality detection method in any one of claims 1 to 7.

Images