Intelligent quality detection system based on machine vision

Abstract

The invention discloses an intelligent quality detection system based on machine vision, relates to the technical field of data processing, and solves the problem that light source parameters are difficult to automatically adjust by combining and utilizing a control algorithm according to environmental parameter influence coefficients. After multi-scale fusion alignment is difficult to carry out on image data, optimized image information is obtained by adopting a Q-learning algorithm and combining image quality scores; defect identification detection and three-dimensional deformation detection are difficult to carry out; and the quality abnormal coefficient is difficult to analyze and control feedback and visualization are difficult to carry out. Through adaptive light source control and image optimization, in combination with defect detection and three-dimensional deformation evaluation, efficient and accurate quality detection is realized, and through real-time feedback and visualization, the defect and deformation state of an object can be conveniently and rapidly known.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and specifically relates to an intelligent quality detection system based on machine vision. Background Art

[0002] Machine vision mainly uses a computer to simulate the visual function of humans, extract information from the images of objective things, process and understand it, and finally apply it to actual detection, measurement, and control. With the development of machine vision technology and the continuous progress of image acquisition devices, there are many deficiencies in existing quality detection technologies in aspects such as light source control, image processing, defect detection, deformation analysis, and feedback mechanisms, making it difficult to meet the requirements of efficient, accurate, and automated quality detection in industrial production. By combining the application of deep learning algorithms, its intelligent quality detection can accurately identify and classify complex targets with high precision and in real time and efficiently.

[0003] The following problems exist in the existing intelligent quality detection system based on machine vision: it is difficult to automatically adjust the light source parameters by combining the influence coefficient of environmental parameters and using a control algorithm; it is difficult to perform multi-scale fusion alignment on image data and then obtain optimized image information by using the Q-learning algorithm in combination with image quality scoring; it is difficult to perform defect identification detection and three-dimensional deformation detection; it is difficult to analyze the quality anomaly coefficient and perform control feedback and visualization. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the first aspect of the present invention provides an intelligent quality detection system based on machine vision, including the following modules:

[0005] Data Image Acquisition Module: Combine various types of sensors to collect the environmental parameters of the detected object in real time, and automatically adjust the light source parameters by calculating the influence coefficient of environmental parameters and using a control algorithm; according to the automatic adjustment result, combine an optical camera, an infrared camera, and a laser scanner to collect the image data of the detected object in real time;

[0006] Data Image Preprocessing Module: Perform preprocessing of data cleaning and normalization on environmental parameters; perform preprocessing of filtering denoising and image enhancement on image data; perform multi-scale fusion alignment on image data through a coordinate transformation matrix;

[0007] Quality Detection and Analysis Module: According to the fused image information, train the light source control behavior by using the Q-learning algorithm, combine the analyzed and calculated image quality score, and automatically adjust the light source parameters to obtain optimized image information; perform defect identification detection by using the YOLOv8 model according to the optimized image information; perform three-dimensional deformation detection according to the optimized image information; calculate the quality anomaly coefficient according to the defect and deformation detection results;

[0008] Control feedback module: Perform real-time control feedback based on the quality anomaly coefficient and visually detect the defects and deformation status of the object.

[0009] A further solution is to collect the environmental parameters of the object to be detected in real time by combining multiple types of sensors, and automatically adjust the light source parameters by calculating the environmental parameter influence coefficient and using a control algorithm, including the following steps:

[0010] Use a light sensor, a temperature sensor, a humidity sensor, a laser dust sensor, and an acceleration sensor to collect the environmental parameters of the object to be detected in real time, including: environmental light intensity, environmental temperature, environmental humidity, environmental dust concentration, and object movement speed;

[0011] According to the environmental parameters of the object to be detected, automatically adjust the light source parameters by calculating the environmental parameter influence coefficient and using a control algorithm, including: brightness, angle, and stroboscopic;

[0012] Formula for calculating the environmental parameter influence coefficient:

[0013] Among them, G(t) is the environmental light intensity at time t; T(t) is the environmental temperature at time t; S(t) is the environmental humidity at time t; F(t) is the environmental dust concentration at time t; V(t) is the object movement speed at time t; t is the time variable, representing the time length from the initial time 0 to the current time of collecting environmental parameter data;

[0014] The target brightness is: L 目标 = L base *(1 + H), where L base represents the reference light source brightness; subtract the current brightness from the target brightness to obtain the light source brightness deviation; input the light source brightness deviation into the PID control algorithm, and automatically adjust the current new light source brightness to the current brightness plus the sum of the output values of the PID control algorithm; similarly to the above steps of automatically adjusting the light source brightness, the current new light source angle and new stroboscopic can be obtained.

[0015] A further solution is to perform multi-scale fusion alignment on the image data through a coordinate transformation matrix, including the following steps:

[0016] According to the results of the automatically adjusted light source parameters, collect image data in real time, including: obtaining the appearance image of the object to be detected using an optical camera, capturing the thermal radiation information of the object to be detected using an infrared camera, and obtaining point cloud data by measuring the three-dimensional structure of the object to be detected by emitting a beam using a laser scanner;

[0017] Use the checkerboard calibration method or Zhang Zhengyou calibration method to calibrate the optical camera, infrared camera, and laser scanner respectively, and calculate the internal parameters and external parameters; among them, the internal parameters include: focal length and principal point position, and the external parameters include: rotation matrix and translation vector;

[0018] For the optical camera, infrared camera, and laser scanner, calculate the transformation matrix from the world coordinate system of the three sensors to the target coordinate system respectively. The transformation matrix formula is: where, R is the rotation matrix and X is the translation vector;

[0019] According to the transformation matrix, transform the optical camera image from the optical camera coordinate system to the world coordinate system, and then to the target coordinate system; transform the infrared camera image from the infrared camera coordinate system to the world coordinate system, and then to the target coordinate system; transform the laser scanner point cloud from the laser scanner coordinate system to the world coordinate system, and then to the target coordinate system;

[0020] Perform spatial alignment and temporal alignment on the image data in the target coordinate system. The spatial alignment includes: spatial position and attitude; project the image data onto the same target coordinate system through the coordinate transformation matrix;

[0021] Perform multi-scale fusion on the coordinate-transformed optical image, infrared image, and point cloud data to obtain the fused and aligned image information.

[0022] For a further solution, according to the fused image information, train the light source control behavior by using the Q-learning algorithm, and combine the analyzed and calculated image quality score to automatically adjust the light source parameters to obtain the optimized image information, including the following steps:

[0023] According to the material data of the detected object, including: metal, plastic, glass; use the collected environmental parameters and object material data as the features of the state space of the Q-learning algorithm to construct a state space of the current environment and object state; the dimension and range of the state space are determined according to the accuracy of the sensor;

[0024] Define the adjustable range and step size of the light source parameters to construct an action space; each action in the action space corresponds to an adjustment method for the light source brightness, angle, and stroboscopic;

[0025] According to the multi-scale fused image information, decompose the image into sub-bands of multiple scales, and calculate the structural similarity SSIM, mean square error MSE, and peak signal-to-noise ratio PSNR respectively at each scale; calculate the image quality score through the formula:

[0026] where, H is the environmental parameter influence coefficient; S is the number of decomposed scales; α s, β s and γ s are the weight coefficients of each scale respectively. The sum of α s , β s and γ s is 1, and they are 0.55, 0.25, and 0.2 respectively;

[0027] Design a reward function according to the image quality score to ensure that the reward function reflects the impact of the light source control behavior on the image quality; create an initialized Q-table to store the Q-values under different state and action combinations, and the initial values of the Q-table are set to zero or random values;

[0028] According to the current environmental state, select the action with the largest current Q-value, otherwise randomly select an action; after executing the selected action, update the Q-value in the Q-table according to the update formula of the Q-learning algorithm by observing the new environmental state and reward value; use the Q-learning algorithm to train the light source control behavior, and obtain the optimized image quality score and optimized image information by continuously adjusting the light source parameters.

[0029] For a further solution, perform defect recognition and detection using the YOLOv8 model based on the optimized image information, including the following steps:

[0030] Collect the optimized historical image information training set, which contains various defect types and image quality scores; annotate the image information set, including: defect category and location information; the defect types include: cracks, scratches, bubbles, depressions, and holes;

[0031] Select the YOLOv8 model as the defect detection model to identify and locate the defect types and regions and output the bounding boxes; input the annotated historical image information training set into the YOLOv8 model for training; input the real-time collected and preprocessed image information into the trained YOLOv8 model for defect detection, and the output includes: the category and location bounding box of the defect;

[0032] Introduce online learning technology, continuously collect new defect data, including newly emerged defect types and samples, annotate the newly collected defect data, and supplement the image information training set; deploy an incremental learning framework to continuously update the defect detection model using the new defect data.

[0033] For a further solution, perform three-dimensional deformation detection based on the optimized image information, including the following steps:

[0034] In the initial state or undeformed state, obtain and store optical images, infrared thermal images, and point cloud data, and set them as reference image data; generate a three-dimensional reference model through multi-scale fusion, including geometric shape, surface texture, and thermal distribution; collect optimized image information of the current state at fixed time intervals, including optical images, infrared thermal images, and point cloud data; among them, align the timestamps of the reference image data and the image information of the current state.

[0035] Perform geometric deformation detection based on the point cloud data. Register the current point cloud with the reference point cloud using the ICP algorithm, and calculate the surface strain distribution through finite element analysis or local mesh deformation models; perform thermal radiation anomaly detection based on the infrared images. Align the current infrared image with the reference infrared image to detect the abnormal temperature field area; combine the thermoelastic theory to analyze the thermal expansion coefficient of temperature change and geometric deformation; perform surface texture change detection based on the optical images. Use feature matching algorithms to compare the surface textures of the current and reference optical images; calculate the microscopic displacement field through optical flow method or digital image correlation method; perform weighted summation on the deformation detection results of the above three modalities to obtain the deformation degree value.

[0036] Verify whether the geometric deformation area coincides spatially with the thermal anomaly and texture change areas to exclude interference factors; set a deformation amount threshold. When the deformation degree value exceeds the threshold, it is determined that the object has deformed.

[0037] For a further solution, calculate the quality anomaly coefficient according to the defect and deformation detection results, including the following steps:

[0038] The defect types and positions recognized in real time by the YOLOv8 model are used to output the bounding boxes of the defect areas; use the instance segmentation algorithm in the semantic segmentation algorithm, and input the image area where the defect position is located into the instance segmentation algorithm; the instance segmentation algorithm analyzes and classifies each pixel in this area to determine whether it belongs to the defective part, and generates a pixel-level defect mask; in the defect mask, the pixel values of the defect area are set to 255, while the pixel values of the non-defect area are set to 0; separate the defect area from the bounding box background.

[0039] After obtaining the defect mask, count the number of pixels with a pixel value of 255 in the mask to obtain the number of pixels in the defect area; according to the pixel size information of the image, convert the number of pixels into actual area units and multiply by the number of pixels in the defect area to obtain the defect area; after obtaining the defect mask, use the edge detection algorithm to extract the contour of the defect area, and calculate the sum of the distances between adjacent pixels on the contour edge to obtain the defect perimeter.

[0040] Formula for calculating the quality anomaly coefficient:

[0041] Obtain the quality anomaly coefficient Y; where A q is the area of the i-th defect; C i is the perimeter of the i-th defect; q is the number of defects; B is the deformation degree value; Q represents the image quality score; H is the environmental parameter influence coefficient.

[0042] In a further solution, the control feedback module includes: setting a threshold for the quality anomaly coefficient. When the quality anomaly coefficient exceeds the set threshold, return to step S1, automatically adjust the light source parameters and re-acquire the image; project the detected defect category label and defect location of the object onto a 3D model for visual display; combine the segmentation mask and point cloud data, and use a 3D reconstruction algorithm to generate a 3D model of the detected object, including the geometric shape, surface texture, and thermal distribution information of the detected object; according to the deformation degree value, use the color mapping method to generate a deformation heat map or a 3D deformation model.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] The present invention detects the environmental parameters of the object and uses a control algorithm to automatically adjust the light source parameters to ensure that high-quality image data can still be obtained in a complex environment; the adaptive light source control mechanism significantly improves the stability and accuracy of image acquisition.

[0045] The present invention uses the Q-learning algorithm to train the light source control behavior, combines the image quality score to dynamically optimize the light source parameters, and further improves the usability of the image. At the same time, using the YOLOv8 model for defect recognition and detection can quickly and accurately locate and identify the defects on the surface of the object, greatly improving the detection efficiency and accuracy.

[0046] The present invention performs three-dimensional deformation detection through the optimized image information, can accurately evaluate the deformation of the object, and combines the defect detection result to obtain the quality anomaly coefficient.

[0047] The present invention adjusts the detection process in real time according to the quality anomaly coefficient and visualizes the detection result, which is convenient for the operator to quickly understand the defect and deformation state of the object. This real-time feedback mechanism significantly improves the response speed and operation convenience of the system and reduces the need for manual intervention. Description of the Drawings

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0049] Figure 1 This is the system module diagram of the present invention. Specific implementation manners

[0050] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0051] Please refer to Figure 1 , an embodiment of the first aspect of the present invention provides an intelligent quality detection system based on machine vision, including the following modules:

[0052] Data image acquisition module: Combine various types of sensors to collect the environmental parameters of the detected object in real time, and automatically adjust the light source parameters by calculating the environmental parameter influence coefficient and using the control algorithm; According to the automatic adjustment result, combine the optical camera, infrared camera and laser scanner to collect the image data of the detected object in real time;

[0053] Data image preprocessing module: Perform preprocessing of data cleaning and normalization on the environmental parameters; Perform preprocessing of filtering noise reduction and image enhancement on the image data; Through the coordinate transformation matrix, perform multi-scale fusion alignment on the image data;

[0054] Quality detection and analysis module: According to the fused image information, train the light source control behavior by adopting the Q-learning algorithm, combine the calculated image quality score, and automatically adjust the light source parameters to obtain the optimized image information; Perform defect identification and detection using the YOLOv8 model according to the optimized image information; Perform three-dimensional deformation detection according to the optimized image information; Calculate the quality anomaly coefficient according to the defect and deformation detection results;

[0055] Control feedback module: Perform real-time control feedback according to the quality anomaly coefficient and visualize the defect and deformation states of the detected object.

[0056] Specifically, an integrated infrared thermal imager, temperature and humidity sensors, light intensity sensors, etc. are used to collect real-time environmental temperature, humidity, light conditions and other parameters of the detected object. By calculating the environmental parameter influence coefficient, the light source parameters are dynamically adjusted. Combining an optical camera, an infrared camera and a laser scanner, real-time image data of the detected object is collected. A frame synchronization mechanism is adopted to ensure the temporal consistency of the image data output by multiple sensors. The collected environmental parameters are subjected to data cleaning and normalization to eliminate noise and outliers. A filtering denoising algorithm is used to denoise the image, and image enhancement techniques including contrast adjustment and sharpening are applied to improve the image quality. The images collected by different sensors are multi-scale fused and aligned through a coordinate transformation matrix to ensure the unity of the image data. The Q-learning algorithm is used to train the light source control behavior, and the light source parameters are dynamically adjusted according to the image quality score to obtain an optimized image. The YOLOv8 model is used to detect defects in the optimized image to ensure the accuracy and real-time performance of the detection. Based on the optimized image information, three-dimensional deformation detection is carried out. According to the defect and deformation detection results, the quality anomaly coefficient is calculated to provide a basis for subsequent control feedback. According to the quality anomaly coefficient, a threshold of the quality anomaly coefficient is set, and when the threshold is exceeded, a control action is triggered to further visualize the defect and deformation states of the detected object.

[0057] In one embodiment, multiple types of sensors are combined to collect real-time environmental parameters of the detected object, and the light source parameters are automatically adjusted by calculating the environmental parameter influence coefficient and using a control algorithm, including the following steps:

[0058] An illumination sensor, a temperature sensor, a humidity sensor, a laser dust sensor and an acceleration sensor are used to collect real-time environmental parameters of the detected object, including: environmental light intensity, environmental temperature, environmental humidity, environmental dust concentration and object movement speed;

[0059] According to the environmental parameters of the detected object, the light source parameters are automatically adjusted by calculating the environmental parameter influence coefficient and using a control algorithm, including: brightness, angle and stroboscopic;

[0060] The formula for calculating the environmental parameter influence coefficient is:

[0061] where G(t) is the environmental light intensity at time t; T(t) is the environmental temperature at time t; S(t) is the environmental humidity at time t; F(t) is the environmental dust concentration at time t; V(t) is the object movement speed at time t; t is the time variable, representing the time length from the initial time 0 to the current time of collecting environmental parameter data;

[0062] The target brightness is: L 目标 =L base*(1 + H), where L base represents the brightness of the reference light source; the light source brightness deviation is obtained by subtracting the current brightness from the target brightness; the light source brightness deviation is input into the PID control algorithm, and the current new light source brightness is automatically adjusted to be the current brightness plus the sum of the output values of the PID control algorithm; similarly, for the above steps of automatically adjusting the light source brightness, the current new light source angle and new stroboscopic can be obtained.

[0063] Specifically, according to the real-time collected and preprocessed environmental parameters, substitute them into the environmental parameter influence coefficient formula to calculate the environmental parameter influence coefficient. According to the environmental parameter influence coefficient and the preset brightness reference value, calculate the target brightness. Compare the target brightness with the current light source brightness to obtain the brightness deviation. Input the deviation into the PID control algorithm to calculate the control quantity for adjusting the brightness of the light source. Similarly, according to the environmental parameter influence coefficient and the angle and stroboscopic reference values of the light source, calculate the target angle and target stroboscopic. Through the corresponding control algorithm, adjust the angle and stroboscopic of the light source to adapt to different environmental conditions. After the light source parameters are adjusted, monitor the actual brightness, angle, stroboscopic and other parameters of the light source in real time and compare them with the target parameters. If there is a deviation, adjust the parameters of the control algorithm in time or recalculate the environmental parameter influence coefficient to ensure the accuracy and stability of the light source parameters.

[0064] In one embodiment, through the coordinate transformation matrix, the image data is multi-scale fused and aligned, including the following steps:

[0065] According to the results of automatically adjusted light source parameters, the real-time collected image data includes: using an optical camera to obtain the appearance image of the detected object, using an infrared camera to capture the thermal radiation information of the detected object, and point cloud data obtained by measuring the three-dimensional structure of the detected object by emitting a beam using a laser scanner;

[0066] Adopt the checkerboard calibration method or Zhang Zhengyou calibration method to calibrate the optical camera, infrared camera and laser scanner respectively, and calculate the internal parameters and external parameters; among them, the internal parameters include: focal length and principal point position, and the external parameters include: rotation matrix and translation vector;

[0067] For the optical camera, infrared camera and laser scanner, calculate the transformation matrix from the world coordinate system of the three sensors to the target coordinate system respectively, and the transformation matrix formula is: where, R is the rotation matrix and X is the translation vector;

[0068] According to the transformation matrix, transform the optical camera image from the optical camera coordinate system to the world coordinate system, and then to the target coordinate system; transform the infrared camera image from the infrared camera coordinate system to the world coordinate system, and then to the target coordinate system; transform the laser scanner point cloud from the laser scanner coordinate system to the world coordinate system, and then to the target coordinate system;

[0069] Perform spatial alignment and temporal alignment on the image data in the target coordinate system. The spatial alignment includes: spatial position and attitude; project the image data onto the same target coordinate system through a coordinate transformation matrix;

[0070] Perform multi-scale fusion on the optically imaged, infrared imaged, and point cloud data after coordinate transformation to obtain the fused and aligned image information.

[0071] Specifically, according to the results of the automatically adjusted light source parameters, data from the optical camera, infrared camera, and laser scanner are collected in real time. Ensure that the collected image data is synchronized in time to reduce the complexity of subsequent time alignment. Preprocess the collected image data, including operations such as denoising and filtering, to improve the data quality and processing efficiency. Use the checkerboard calibration method or Zhang Zhengyou calibration method to calibrate the optical camera, infrared camera, and laser scanner respectively. Among them, the checkerboard calibration method uses the checkerboard pattern as the calibration board, and by taking multiple checkerboard images at different angles, the internal and external parameters of the camera are calculated. The Zhang Zhengyou calibration algorithm calculates the internal and external parameters of the camera by taking multiple images of the planar calibration board. Place the checkerboard calibration board within the field of view of the optical camera and take multiple images at different angles. Use the calibration functions in computer vision libraries such as OpenCV to calculate the internal parameters of the optical camera, including: focal length and principal point position, and external parameters, including: rotation matrix and translation vector. Similarly, use the checkerboard calibration board or Zhang Zhengyou calibration method to calibrate the infrared camera to obtain its internal and external parameters. By scanning an object with a known geometric shape, obtain the point cloud data and match it with the known geometric model to calculate the internal and external parameters of the laser scanner. Select a suitable reference point as the origin of the world coordinate system. Usually, a certain feature point of the detected object or a certain corner point of the calibration board can be selected. For the optical camera, infrared camera, and laser scanner, calculate the transformation matrix from their world coordinate system to the target coordinate system respectively. According to the external parameters of the optical camera, construct the transformation matrix from the optical camera coordinate system to the world coordinate system. Then, according to the relationship between the target coordinate system and the world coordinate system, calculate the transformation matrix from the optical camera coordinate system to the target coordinate system. Similarly, according to the external parameters of the infrared camera, calculate the transformation matrix from the infrared camera coordinate system to the target coordinate system. According to the external parameters of the laser scanner, calculate the transformation matrix from the laser scanner coordinate system to the target coordinate system. Using the transformation matrices calculated above, transform the optical camera image, infrared camera image, and laser scanner point cloud from their respective coordinate systems to the target coordinate system. Convert the two-dimensional image pixel coordinates obtained by the optical camera to three-dimensional coordinates in the target coordinate system through the transformation matrix. Similarly, convert the pixel coordinates of the thermal radiation image obtained by the infrared camera to the target coordinate system. Convert the three-dimensional point cloud data obtained by the laser scanner to the target coordinate system through the transformation matrix. In the target coordinate system, perform spatial alignment on the transformed image data. Spatial alignment includes alignment of spatial position and attitude to ensure the consistency of data obtained by different sensors in space. By adjusting the position of the transformed data, make its coordinates in the target coordinate system consistent. Through rotation and translation operations, make the data of different sensors consistent in attitude. Since there may be differences in the acquisition times of different sensors, time alignment is required. Methods such as interpolation can be used to adjust the data collected at different times to the same time point.Perform multi-scale decomposition on the transformed and aligned optical images, infrared images, and point cloud data. Methods such as Gaussian pyramids or Laplacian pyramids can be used to decompose the image data into different resolution levels. According to different application scenarios and requirements, select appropriate fusion strategies including: weighted average, maximum selection, wavelet transform fusion, etc. Starting from the lowest resolution level, gradually fuse upward. At each level, according to the selected fusion strategy, fuse the optical image, infrared image, and point cloud data to obtain the fused image information. Reconstruct the fused multi-scale image information to obtain the final fused and aligned image that contains optical, infrared, and three-dimensional structure information and can more comprehensively reflect the characteristics of the detected object.

[0072] In one embodiment, according to the fused image information, train the light source control behavior by using the Q-learning algorithm, and combine the analyzed and calculated image quality score to automatically adjust the light source parameters to obtain the optimized image information, including the following steps:

[0073] According to the material data of the detected object including: metal, plastic, glass; use the collected environmental parameters and object material data as the features of the state space of the Q-learning algorithm to construct a state space of the current environment and object state; the dimension and range of the state space are determined according to the accuracy of the sensor;

[0074] Define the adjustable range and step size of the light source parameters to construct an action space; each action in the action space corresponds to a way of adjusting the light source brightness, angle, and stroboscopic;

[0075] According to the multi-scale fused image information, decompose the image into sub-bands of multiple scales, and calculate the structural similarity SSIM, mean square error MSE, and peak signal-to-noise ratio PSNR respectively at each scale; the image quality score is calculated by the formula:

[0076] where, H is the environmental parameter influence coefficient; S is the number of decomposition scales; α s , β s and γ s are the weight coefficients of each scale respectively, and the sum of α s , β s and γ s is 1, and they are 0.55, 0.25, and 0.2 respectively;

[0077] Design a reward function according to the image quality score to ensure that the reward function reflects the impact of the light source control behavior on the image quality; create an initialized Q table for storing the Q values under different state and action combinations, and the initial value of the Q table is set to zero or a random value;

[0078] Select the action with the largest current Q-value according to the current environmental state; otherwise, randomly select an action. After executing the selected action, update the Q-value in the Q-table according to the update formula of the Q-learning algorithm by observing the new environmental state and the reward value. Use the Q-learning algorithm to train the light source control behavior, and obtain the optimized image quality score and the optimized image information by continuously adjusting the light source parameters.

[0079] Specifically, environmental parameters of the detected object are collected in real time through a light sensor, a temperature sensor, a humidity sensor, a laser dust sensor, an acceleration sensor, etc., including: environmental light intensity, environmental temperature, environmental humidity, environmental dust concentration, and object movement speed. According to the materials of the detected object, including: metal, plastic, glass, etc., the corresponding material data can be obtained, which can be obtained in real time through a spectral sensor. The collected environmental parameters are normalized and converted to the same numerical range for subsequent state space construction. Corresponding features are extracted according to different object materials. For example, for metal materials, features such as reflectivity and thermal conductivity can be extracted; for plastic materials, features such as transparency and hardness can be extracted; for glass materials, features such as refractive index and light transmittance can be extracted. The dimension of the state space is determined according to the accuracy of the sensor and the diversity of object material features. The environmental light intensity can be divided into 10 levels, the environmental temperature can be divided into 5 levels, the environmental humidity can be divided into 5 levels, the environmental dust concentration can be divided into 5 levels, the object movement speed can be divided into 5 levels, and the object material features can be divided into 3 levels including: metal, plastic, glass. The range of the state space is determined according to the measurement range of the sensor and the value range of object material features. The range of environmental light intensity can be from 0 to 1000 lx, the range of environmental temperature can be from -20°C to 50°C, the range of environmental humidity can be from 0% to 100%, and the range of environmental dust concentration can be from 0 to 1000 μg / m 3, the range of the object's movement speed can be from 0 to 10 m / s. Define the adjustable range and step size of the light source parameters to construct the action space; each action in the action space corresponds to an adjustment method for the light source brightness, angle, and stroboscopic effect; define the adjustable range of the light source brightness as from 0 to 100% and the step size as 5%; define the adjustable range of the light source angle as from 0° to 360° and the step size as 10°; define the adjustable range of the light source stroboscopic effect as from 0 to 100 Hz and the step size as 5 Hz. Determine the dimension of the action space according to the adjustable range and step size of the light source parameters. Decompose the multi-scale fused image information into sub-bands of multiple scales. Methods such as Gaussian pyramid or Laplacian pyramid can be used to decompose the image into different resolution levels. At each scale, calculate the structural similarity SSIM of the optical image, infrared image, and point cloud data, which reflects the structural similarity between the images. The larger the value, the higher the structural similarity. At each scale, calculate the mean squared error MSE of the optical image, infrared image, and point cloud data, which reflects the error degree between the images. The smaller the value, the smaller the error. At each scale, calculate the peak signal-to-noise ratio PSNR of the optical image, infrared image, and point cloud data, which reflects the signal-to-noise ratio of the images. The larger the value, the higher the signal-to-noise ratio. According to historical experience or experimental data, determine the weight coefficients of each scale. The number of decomposed scales S can be 3, and the weight coefficients of each scale can be 0.55, 0.25, and 0.2 respectively. Calculate the image quality score according to the image quality scoring formula. The reward function should be able to accurately reflect the impact of the light source control behavior on the image quality score. When the light source control behavior leads to an increase in the image quality score, give a positive reward; when the light source control behavior leads to a decrease in the image quality score, give a negative reward. The reward function can be designed to be proportional to the difference in the image quality score. The Q-table is a two-dimensional table, where the rows represent the state space and the columns represent the action space. The initial value of the Q-table can be set to zero or a random value. Use data structures such as arrays or hash tables to store the Q-table for quick lookup and update. According to the current environmental state, select the action with the largest current Q value, otherwise randomly select an action. After executing the selected action, observe the new environmental state and the reward value. Update the Q value in the Q-table according to the update formula of the Q-learning algorithm. In each training iteration, select an action according to the current environmental state, observe the new state and the reward value after executing the action, and update the Q-table. Repeat this process until the Q-table converges or reaches the preset number of training times. According to the Q values stored in the Q-table, select the optimal action, that is, the optimal light source parameter adjustment method, and update the light source parameters; by continuously adjusting the light source parameters, obtain the optimized image quality score and the optimized image information.

[0080] In one embodiment, use the YOLOv8 model to perform defect recognition and detection based on the optimized image information, including the following steps:

[0081] Collect the optimized historical image information training set, which contains various defect types and image quality scores; label the image information set, including: defect categories and location information; the defect types include: cracks, scratches, bubbles, dents, and holes.

[0082] Select the YOLOv8 model as the defect detection model to identify and locate the defect types and regions and output the bounding boxes; input the labeled historical image information training set into the YOLOv8 model for training; input the real-time collected and preprocessed image information into the trained YOLOv8 model for defect detection, and the output includes: the category and location bounding box of the defect.

[0083] Introduce online learning technology. By continuously collecting new defect data, including newly emerged defect types and samples, label the newly collected defect data to supplement the image information training set; deploy an incremental learning framework to continuously update the defect detection model using the new defect data.

[0084] Specifically, collect the optimized historical image information training set. These images should contain various defect types, including: cracks, scratches, bubbles, dents, holes, etc., and record the image quality score of each image. Label the image information set, and the labeled content includes: defect categories and location information. The location information is usually represented in the form of a bounding box, including the coordinates of the defect in the image. To increase the diversity of the data set and improve the generalization ability of the model, data augmentation techniques are adopted, such as image rotation, flipping, scaling, cropping, adding noise, etc. Select the YOLOv8 model as the defect detection model. This model performs well in object detection tasks and can quickly and accurately identify and locate defects. Input the labeled historical image information training set into the YOLOv8 model for training. During the training process, adjust the model parameters, such as the learning rate, batch size, etc., to optimize the performance of the model. During the training process, regularly evaluate the performance of the model, such as calculating metrics such as accuracy, recall, and F1 score, to ensure that the model can accurately identify and locate defects. Introduce online learning technology. By continuously collecting new defect data, including newly emerged defect types and samples, label the newly collected defect data to supplement the image information training set. Continuously update the defect detection model using the new defect data. During the incremental learning process, adopt appropriate strategies, such as regular update, batch update, etc., to ensure that the model can adapt to new defect types and samples while avoiding catastrophic forgetting. Real-time collect and preprocess new image information, input the preprocessed new image information into the trained YOLOv8 model for defect detection, and output the detected defect category and location bounding box.

[0085] In one embodiment, perform three-dimensional deformation detection based on the optimized image information, including the following steps:

[0086] In the initial state or undeformed state, obtain and store optical images, infrared thermal images, and point cloud data, and set them as reference image data; generate a three-dimensional reference model through multi-scale fusion, including geometric shape, surface texture, and thermal distribution; collect optimized image information of the current state at fixed time intervals, including optical images, infrared thermal images, and point cloud data; among them, align the timestamps of the reference image data and the image information of the current state.

[0087] Perform geometric deformation detection based on the point cloud data. Register the current point cloud with the reference point cloud using the ICP algorithm, and calculate the surface strain distribution through finite element analysis or local mesh deformation model; perform thermal radiation anomaly detection based on the infrared image. Align the current infrared image with the reference infrared image to detect the abnormal temperature field area; combine the thermoelastic theory to analyze the thermal expansion coefficient of temperature change and geometric deformation; perform surface texture change detection based on the optical image. Use the feature matching algorithm to compare the surface texture of the current and reference optical images; calculate the microscopic displacement field through the optical flow method or digital image correlation method; perform weighted summation on the deformation detection results of the above three modalities to obtain the deformation degree value.

[0088] Verify whether the geometric deformation area coincides spatially with the thermal anomaly and texture change areas to exclude interference factors; set a deformation amount threshold. When the deformation degree value exceeds the threshold, it is determined that the object has deformed.

[0089] Specifically, in the initial state or non-deformed state, obtain and store optical images, infrared thermal images, and point cloud data, and set them as reference image data. Real-time collect the optimized image information of the current state, including: optical images, infrared thermal images, and point cloud data. Align the timestamps of the reference image data and the image information of the current state to ensure the temporal consistency of the data. Through multi-scale fusion technology, fuse the optical image, infrared thermal image, and point cloud data to generate a three-dimensional reference model containing geometric shapes, surface textures, and thermal distributions. Register the current point cloud with the reference point cloud using the ICP algorithm to ensure the spatial consistency of the point cloud data. Calculate the surface strain distribution through finite element analysis or local mesh deformation models. Finite element analysis can be used to simulate the stress and strain distributions of materials under force or deformation. Align the current infrared image with the reference infrared image to detect abnormal temperature field regions. Combine thermoelastic theory to analyze the thermal expansion coefficient of temperature changes and geometric deformations, and evaluate the impact of thermal expansion on deformations. Use feature matching algorithms to compare the surface textures of the current and reference optical images to detect texture changes. Calculate the microscopic displacement field through optical flow methods or digital image correlation methods to analyze the subtle changes in surface textures. Perform a weighted sum of the results of geometric deformation detection, thermal radiation anomaly detection, and surface texture change detection to obtain a deformation degree value; among them, the weights of the results of geometric deformation detection, thermal radiation anomaly detection, and surface texture change detection are 0.55, 0.2, and 0.25 respectively, and the weights can be adjusted through experiments and historical data. Verify whether the geometric deformation region coincides spatially with the thermal anomaly and texture change regions to exclude interference factors. By setting the deformation amount threshold to 0.2, when the deformation degree value exceeds the threshold, it is determined that the object has deformed. Among them, the setting of the deformation degree value threshold needs to be adjusted and determined according to the specific application scenario, the characteristics of the monitoring object, and the statistical analysis of historical data.

[0090] In one embodiment, according to the defect and deformation detection results, calculate the quality anomaly coefficient, including the following steps:

[0091] The type and location of the defects recognized in real time by the YOLOv8 model are output as the bounding boxes of the defect regions; use the instance segmentation algorithm in the semantic segmentation algorithm, and input the image region located at the defect position into the instance segmentation algorithm; the instance segmentation algorithm analyzes and classifies each pixel in this region to determine whether it belongs to the defective part, and generates a pixel-level defect mask; in the defect mask, the pixel values of the defect regions are set to 255, while the pixel values of the non-defective regions are set to 0; separate the defect regions from the bounding box background;

[0092] After obtaining the defect mask, count the pixels with a pixel value of 255 in the mask to obtain the number of pixels in the defect area; according to the pixel size information of the image, convert the number of pixels into actual area units and multiply by the number of pixels in the defect area to obtain the defect area; after obtaining the defect mask, use an edge detection algorithm to extract the contour of the defect area, and calculate the sum of the distances between adjacent pixels on the contour edge to obtain the defect perimeter;

[0093] Formula for calculating the quality anomaly coefficient:

[0094] Obtain the quality anomaly coefficient Y; where A q is the area of the i-th defect; C i is the perimeter of the i-th defect; q is the number of defects; B is the deformation degree value; Q represents the image quality score; H is the environmental parameter influence coefficient.

[0095] Specifically, use the YOLOv8 model to real-time identify the defect type and location, and output the bounding box of the defect area. Collect the optimized image information, including optical images, infrared thermal images, and point cloud data. Input the image area located at the defect position into the instance segmentation algorithm Mask R-CNN. The instance segmentation algorithm analyzes and classifies each pixel in this area to generate a pixel-level defect mask. In the defect mask, the pixel value of the defect area is set to 255, and the pixel value of the non-defect area is set to 0, separating the defect area from the bounding box background. Count the pixels with a pixel value of 255 in the mask to obtain the number of pixels in the defect area. According to the pixel size information of the image, convert the number of pixels into actual area units, and finally multiply by the number of pixels in the defect area to obtain the defect area. Use an edge detection algorithm to extract the contour of the defect area, and calculate the sum of the distances between adjacent pixels on the contour edge to obtain the defect perimeter. Calculate the deformation degree value through a three-dimensional deformation detection method. Use the multi-scale fused image information to calculate the image quality score. According to the collected environmental parameters, calculate the environmental parameter influence coefficient. By substituting into the quality anomaly coefficient formula, obtain the quality anomaly coefficient of the detected object.

[0096] In one embodiment, the control feedback module includes: setting a threshold for the quality anomaly coefficient. When the quality anomaly coefficient exceeds the set threshold, return to step S1, automatically adjust the light source parameters and re-collect the image; project the detected defect category label and defect location of the object onto the three-dimensional model for visual display; combine the segmentation mask and point cloud data, and use a 3D reconstruction algorithm to generate a three-dimensional model of the detected object, including the geometric shape, surface texture, and thermal distribution information of the detected object; according to the deformation degree value, use the color mapping method to generate a deformation heat map or a three-dimensional deformation model.

[0097] Specifically, collect a large amount of historical quality anomaly coefficient data, calculate its mean and standard deviation, and set the threshold as the mean ± 3 times the standard deviation according to the 3-sigma principle, so as to cover the quality anomaly coefficient fluctuations in the vast majority of normal situations and effectively identify abnormal situations. Obtain a reasonable range of quality anomaly coefficient thresholds by referring to relevant industry standards or consulting domain experts, and make appropriate adjustments in combination with the actual situation. By continuously collecting new quality anomaly coefficient data, recalculate the statistical indicators regularly, and dynamically adjust the threshold to adapt to the changes in the data distribution. Use the 3D reconstruction algorithm, combined with the segmentation mask and point cloud data, to generate a three-dimensional model of the detected object, including: the geometric shape, surface texture, and thermal distribution information of the detected object. Traditional computer vision-based 3D reconstruction methods, such as stereo vision and structured light, can be used, or deep learning-based 3D reconstruction algorithms, such as 3D U-Net, can also be used; project the defect category label and defect location of the detected object onto the 3D model, and the visualization display of defect information can be realized by marking the defect location and annotating the defect category label on the 3D model. According to the magnitude of the deformation degree value, select an appropriate color mapping method, such as using a gradient color to represent different deformation degrees, with the color changing from light to dark or from cold color to warm color indicating a gradual increase in the deformation degree. On the two-dimensional image, assign corresponding colors to each pixel point according to the deformation degree value to generate a deformation heat map, which intuitively shows the deformation distribution on the object surface. On the 3D model, assign corresponding colors to each point or each face on the model surface according to the deformation degree value to generate a 3D deformation model, which more comprehensively and intuitively shows the 3D deformation of the object.

[0098] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An intelligent quality inspection system based on machine vision, characterized in that, It includes the following modules: Data image acquisition module: Combine various types of sensors to collect the environmental parameters of the detected object in real time, and automatically adjust the light source parameters by calculating the environmental parameter influence coefficient and using the control algorithm; According to the automatic adjustment result, combine the optical camera, infrared camera and laser scanner to collect the image data of the detected object in real time; Data image preprocessing module: Perform preprocessing of data cleaning and normalization on the environmental parameters; Perform preprocessing of filtering denoising and image enhancement on the image data; Perform multi-scale fusion alignment on the image data through the coordinate transformation matrix; Quality detection and analysis module: According to the fused image information, train the light source control behavior by using the Q-learning algorithm, and combine the calculated image quality score to automatically adjust the light source parameters to obtain the optimized image information; Use the YOLOv8 model to perform defect identification and detection according to the optimized image information; Perform three-dimensional deformation detection according to the optimized image information; Calculate the quality anomaly coefficient according to the defect and deformation detection results; Control feedback module: Perform real-time control feedback according to the quality anomaly coefficient and visualize the defect and deformation states of the detected object.

2. The intelligent quality inspection system based on machine vision according to claim 1, characterized in that, Combine various types of sensors to collect the environmental parameters of the detected object in real time, and automatically adjust the light source parameters by calculating the environmental parameter influence coefficient and using the control algorithm, including the following steps: Adopt a light sensor, temperature sensor, humidity sensor, laser dust sensor and acceleration sensor to collect the environmental parameters of the detected object in real time, including: environmental light intensity, environmental temperature, environmental humidity, environmental dust concentration and object movement speed; According to the environmental parameters of the detected object, automatically adjust the light source parameters by calculating the environmental parameter influence coefficient and using the control algorithm, including: brightness, angle and stroboscopic; Calculation formula for the influence coefficient of environmental parameters: Among them, G(t) is the environmental light intensity at time t; T(t) is the environmental temperature at time t; S(t) is the environmental humidity at time t; F(t) is the environmental dust concentration at time t; V(t) is the object movement speed at time t; t is the time variable, indicating the time length from the initial time 0 to the current time of collecting environmental parameter data; The target brightness is: L 目标 = L base *(1 + H), where L base represents the brightness of the reference light source; the light source brightness deviation is obtained by subtracting the current brightness from the target brightness; the light source brightness deviation is input into the PID control algorithm, and the current new light source brightness is automatically adjusted to be the current brightness plus the sum of the output values of the PID control algorithm; similarly to the above steps of automatically adjusting the light source brightness, the current new light source angle and new stroboscopic effect can be obtained.

3. An intelligent quality inspection system based on machine vision according to claim 1, characterized in that, Perform multi-scale fusion alignment on the image data through the coordinate transformation matrix, including the following steps: According to the result of the automatically adjusted light source parameters, collect the image data in real time, including: obtaining the appearance image of the detected object by using the optical camera, capturing the thermal radiation information of the detected object by using the infrared camera, and obtaining the point cloud data by measuring the three-dimensional structure of the detected object by using the laser scanner to emit a light beam; Adopt the checkerboard calibration method or Zhang Zhengyou calibration method to calibrate the optical camera, infrared camera and laser scanner respectively, and calculate the internal parameters and external parameters; Among them, the internal parameters include: focal length and principal point position, and the external parameters include: rotation matrix and translation vector; For the optical camera, infrared camera, and laser scanner, calculate the transformation matrix from the world coordinate system of the three sensors to the target coordinate system respectively. The transformation matrix formula is as follows: where R is the rotation matrix and X is the translation vector; According to the transformation matrix, transform the optical camera image from the optical camera coordinate system to the world coordinate system, and then to the target coordinate system; transform the infrared camera image from the infrared camera coordinate system to the world coordinate system, and then to the target coordinate system; transform the laser scanner point cloud from the laser scanner coordinate system to the world coordinate system, and then to the target coordinate system; Perform spatial alignment and temporal alignment on the image data in the target coordinate system. The spatial alignment includes: spatial position and attitude; project the image data onto the same target coordinate system through a coordinate transformation matrix; Perform multi-scale fusion on the optically imaged, infrared imaged, and point cloud data after coordinate transformation to obtain the fused and aligned image information.

4. An intelligent quality inspection system based on machine vision according to claim 1, characterized in that According to the fused image information, train the light source control behavior by using the Q-learning algorithm, and combine the analyzed and calculated image quality score to automatically adjust the light source parameters to obtain the optimized image information, including the following steps: Based on the material data of the detected object, including: metal, plastic, glass; use the collected environmental parameters and object material data as the features of the state space of the Q-learning algorithm to construct a state space of the current environment and object state; the dimension and range of the state space are determined according to the accuracy of the sensor; Define the adjustable range and step size of the light source parameters to construct an action space; each action in the action space corresponds to an adjustment method for the light source brightness, angle, and stroboscopic; According to the image information after multi-scale fusion, the image is decomposed into sub-bands of multiple scales, and the structural similarity SSIM, mean square error MSE, and peak signal-to-noise ratio PSNR are calculated separately at each scale; the image quality scoring formula is calculated as: Among them, H is the environmental parameter influence coefficient; S is the number of decomposition scales; α s , β s and γ s are the weight coefficients of each scale respectively. The sum of α s , β s and γ s is 1, which are 0.55, 0.25 and 0.2 respectively; Design a reward function according to the image quality score to ensure that the reward function reflects the impact of the light source control behavior on the image quality; create an initialized Q-table for storing the Q-values under different state and action combinations, and the initial value of the Q-table is set to zero or a random value; According to the current environmental state, select the action with the largest current Q-value, otherwise randomly select an action; after executing the selected action, observe the new environmental state and reward value, and update the Q-value in the Q-table according to the update formula of the Q-learning algorithm; use the Q-learning algorithm to train the light source control behavior, and continuously adjust the light source parameters to obtain the optimized image quality score and optimized image information.

5. An intelligent quality inspection system based on machine vision according to claim 1, wherein Perform defect identification and detection on the optimized image information by using the YOLOv8 model, including the following steps: Collect the training set of the optimized historical image information, which contains various defect types and image quality scores; annotate the image information set, including: defect category and location information; the defect types include: cracks, scratches, bubbles, dents, and holes; Select the YOLOv8 model as the defect detection model to identify and locate the defect types and regions and output the bounding boxes; input the annotated training set of the historical image information into the YOLOv8 model for training; input the real-time collected and preprocessed image information into the trained YOLOv8 model for defect detection and output, including: the category and location bounding box of the defect; Introduce online learning technology. By continuously collecting new defect data, including newly emerged defect types and samples, annotate the newly collected defect data to supplement the image information training set; deploy an incremental learning framework and use the new defect data to continuously update the defect detection model.

6. An intelligent quality inspection system based on machine vision according to claim 1, characterized in that, Perform three-dimensional deformation detection based on the optimized image information, including the following steps: In the initial state or non-deformed state, acquire and store optical images, infrared thermal images, and point cloud data, and set them as reference image data; generate a three-dimensional reference model through multi-scale fusion, including geometric shape, surface texture, and thermal distribution; collect the optimized image information of the current state at fixed time intervals, including optical images, infrared thermal images, and point cloud data; among them, align the time stamps of the reference image data and the image information of the current state. Perform geometric deformation detection based on the point cloud data. Register the current point cloud with the reference point cloud using the ICP algorithm, and calculate the surface strain distribution through finite element analysis or local mesh deformation model; perform thermal radiation anomaly detection based on the infrared image, align the current infrared image with the reference infrared image, and detect the abnormal temperature field area; combine the thermoelastic theory to analyze the thermal expansion coefficient of temperature change and geometric deformation; perform surface texture change detection based on the optical image, and use the feature matching algorithm to compare the surface texture of the current and reference optical images; calculate the microscopic displacement field through the optical flow method or digital image correlation method; perform weighted summation on the deformation detection results of the above three modalities to obtain the deformation degree value. Exclude interference factors by verifying whether the geometric deformation area coincides with the thermal anomaly and texture change areas in space; set a deformation amount threshold, and when the deformation degree value exceeds the threshold, it is determined that the object has deformed.

7. An intelligent quality inspection system based on machine vision according to claim 1, characterized in that, Calculate the quality anomaly coefficient according to the defect and deformation detection results, including the following steps: Through the defect types and positions identified in real time by the YOLOv8 model, output the bounding box of the defect area; use the instance segmentation algorithm in the semantic segmentation algorithm, and input the image area where the defect position is located into the instance segmentation algorithm; the instance segmentation algorithm analyzes and classifies each pixel in this area to determine whether it belongs to the defect part, and generates a pixel-level defect mask; in the defect mask, the pixel value of the defect area is set to 255, while the pixel value of the non-defect area is set to 0; separate the defect area from the bounding box background. After obtaining the defect mask, count the number of pixels with a pixel value of 255 in the mask to obtain the number of pixels in the defect area; according to the pixel size information of the image, convert the number of pixels to the actual area unit and multiply it by the number of pixels in the defect area to obtain the defect area; after obtaining the defect mask, use the edge detection algorithm to extract the contour of the defect area, and calculate the sum of the distances between adjacent pixels on the contour edge to obtain the defect perimeter. Formula for calculating the quality anomaly coefficient: Obtain the quality anomaly coefficient Y; where A q is the area of the i-th defect; C i is the perimeter of the i-th defect; q is the number of defects; B is the deformation degree value; Q represents the image quality score; H is the environmental parameter influence coefficient.

8. An intelligent quality inspection system based on machine vision according to claim 1, characterized in that, The control feedback module includes: setting a threshold for the quality anomaly coefficient, and when the quality anomaly coefficient exceeds the set threshold, returning to step S1 to automatically adjust the light source parameters and re-acquire images; projecting the detected defect category label and defect location of the object onto a three-dimensional model for visual display; combining the segmentation mask and point cloud data, and using a 3D reconstruction algorithm to generate a three-dimensional model of the detected object, including the geometric shape, surface texture, and thermal distribution information of the detected object; generating a deformation heat map or a three-dimensional deformation model using a color mapping method according to the deformation degree value.

Images