A rectangular and circular feature-based circuit board infrared image and visible light image registration method

Abstract

The application discloses a circuit board infrared image and visible light image registration method based on rectangular and circular features. Circuit board infrared images and circuit board visible light images obtained by shooting at the same time are acquired, and Canny edge detection is performed on the images to obtain edge images. Rectangular and circular detection is performed on the edge images to obtain rectangular features and circular features, and bilateral matching is performed on the feature points to obtain initial matching results. Rotation angle estimation and connection line consistency are applied to the results to remove obvious error matching, and RANSAC algorithm is used to remove a small amount of remaining error matching. After error matching is completed, fine matching based on global iterative optimization is performed to reduce the position deviation of the feature points. The method uses the rectangular components with heat and the circular hollows without heat in the circuit board infrared images, avoids the problem of error matching of similar features in the traditional registration method based on point features and straight line features, and realizes high-precision and high-time-efficiency registration.

Description

TECHNICAL FIELD

[0001] The present application relates to a rectangular and circular feature-based circuit board infrared image and visible light image registration method, belonging to the technical field of infrared image and visible light image hetero-registration. BACKGROUND

[0002] Circuit boards play a crucial role in modern electronic devices. However, due to their complex structure and high-density component layout, fault detection on circuit boards has always been a challenging task. In order to ensure the normal operation and reliability of the circuit board, it is very important to detect and diagnose potential faults.

[0003] In recent years, infrared thermal imaging technology has been widely applied in the field of circuit board fault detection. Infrared thermal imaging technology can measure the heat distribution on the surface of an object and generate corresponding infrared images. By analyzing these infrared images, thermal abnormal areas on the circuit board, such as overheating, short circuits or other faults, can be detected. Therefore, infrared thermal imaging technology has become a very effective method for circuit board fault detection.

[0004] However, infrared images only provide thermal information about the circuit board, lacking the detailed structural information possessed by visible light images. Visible light images can provide details on the circuit board, such as components, wiring and markings, which are crucial for accurate positioning of faults and further analysis.

[0005] Therefore, it is essential to register infrared images with visible light images. Infrared image and visible light image registration refers to aligning images of two different modalities in space so that their corresponding parts are in corresponding positions on the image. By achieving registration of infrared images and visible light images, the advantages of infrared thermal imaging technology and visible light images can be fully utilized, improving the accuracy and reliability of circuit board fault detection.

[0006] Currently, research on infrared image and visible light image registration has made some progress. Methods suitable for circuit board infrared image and visible light image registration mainly include feature-based methods and region-based methods. Region-based methods mainly include mutual information and cross-correlation methods. Due to the different imaging principles of visible light images and infrared images, infrared images contain much less information than visible light images. This type of method has poor robustness in the spatial domain and poor time efficiency in the frequency domain. In contrast, feature-based methods are generally more stable. The features used in these methods include points, line segments and contours. Feature-based methods can generally be described as a process of finding matching points and obtaining a transformation matrix. Point features are widely distributed and easy to detect. Therefore, point feature-based methods are widely used, with the most representative algorithms including SIFT, SURF and ORB.

[0007] However, there are a large number of similar structures on the circuit board, and the point features obtained from the infrared image and the visible light image of the circuit board have similar descriptors, which can cause a large number of false matches in the registration process. Taking a rectangular component as an example, four corner points of the rectangular component are usually extracted when the point features are extracted, and the four corner points have the same angle and principal direction due to the requirement of rotation invariance. On the other hand, there are a large number of circular holes on the circuit board, and the infrared image shows a black circular area due to the lack of heat of the hole. The circle is smooth and has no corner point, so the traditional point feature-based method cannot extract effective feature points. SUMMARY

[0008] To solve the problems existing in the prior art registration algorithm, the present application aims to provide a registration method for infrared images and visible light images of a circuit board based on rectangular and circular features, which extracts rectangular and circular features from the infrared and visible light images of the circuit board to avoid the problem of no matching caused by similar features in the traditional point or straight line feature-based registration algorithm.

[0009] To achieve the above object, the present application adopts the following technical scheme:

[0010] A registration method for infrared images and visible light images of a circuit board based on rectangular and circular features, characterized in that it comprises the following steps:

[0011] Step 1: Obtain the infrared image and the visible light image of the circuit board taken at the same time, and perform edge detection on the infrared and visible light images by using a Canny operator to obtain binary edge images of the infrared and visible light images of the circuit board, wherein the edge pixels are marked as white and the non-edge pixels are marked as black;

[0012] Step 2: Perform rectangular detection on the binary edge images of the infrared and visible light images of the circuit board obtained in step 1, and extract rectangular features, including four corner points of the rectangle, the length of the side and the angle;

[0013] Step 3: Perform circular detection on the binary edge images of the infrared and visible light images of the circuit board obtained in step 1, and extract circular features, including the center and the radius of the circle;

[0014] Step 4: Construct the rectangular feature and the circular feature descriptor by using the rectangular and circular information obtained in steps 2 and 3;

[0015] Step 5: Preliminarily match the infrared image and the visible light image by using bilateral matching to obtain an initial matching feature point pair;

[0016] Step 6: Remove the obviously false matching feature point pairs by using the rotation angle estimation and the consistency theory of the connecting line, and remove the remaining small number of false matching feature point pairs by using the RANSAC algorithm for hundreds of iterations;

[0017] Step 7: Estimate the projection transformation matrix between infrared and visible light images using the coarse matching feature points;

[0018] Step 8: Reduce the position deviation of feature points using fine matching based on global iterative optimization.

[0019] The above step 2 specifically includes the following steps:

[0020] Step 2.1, according to the edge image, initialize an empty contour list;

[0021] Step 2.2, traverse each pixel of the edge image, when a white pixel is found, mark it as the current pixel and start the contour tracking process, in the contour tracking process, according to the connectivity of adjacent pixels and the boundary condition in the edge image, determine how to move to the next pixel, this process will continue until it returns to the starting pixel, forming a closed contour, add the formed contour to the contour list;

[0022] Step 2.3, repeat step 2.2 until the entire edge image is traversed, so that all closed contours in the image can be found and extracted;

[0023] Step 2.4, screen the detected contours, select the contours that meet the rectangular feature, screen according to the area attribute of the contour, use the Shoelace formula to calculate the area of the polygon, the Shoelace formula is based on the vertex coordinates of the polygon, decomposes the polygon into triangles, calculates the area of the triangle, then adds these areas, the formula is as follows:

[0024] ;

[0025] Where n represents the number of edges of the polygon, represents the coordinates of the i-th vertex of the polygon;

[0026] Set the coefficient , if the area of the polygon contour is , delete the contour, where S is the total area of the image;

[0027] Step 2.5, for the screened contour, use the polygon approximation algorithm to approximate the contour to a polygon, use the Douglas-Peucker algorithm, by retaining the key points far away from the original curve to simplify the contour, select the point farthest from the original curve on the contour as the key point, then divide the curve into two sub-curves, respectively, and recursively process them, define a function represents the vertical distance from point to line segment , that is:

[0028] where, The starting point is , The end point of the result curve is

[0029] The starting point on the curve is selected The end point is selected The point with the maximum distance between the starting point and the end point is found by iteratively calculating the perpendicular distance between each point and the starting point and the end point If the maximum distance is greater than a given threshold, the current curve is cut at to obtain two sub-curves; the algorithm is recursively called on the two sub-curves respectively until the maximum distance is less than the threshold or no further cutting is possible; finally, the key points of each sub-curve are connected to obtain the final approximate curve;

[0030] Step 2.6, for the polygon obtained by approximation, judge whether it is a rectangle by calculating the number of sides and angles of the polygon; first, get the number of sides of the polygon according to the number of contour angle points, delete the polygon whose side number is not 4; then calculate the internal angle of the polygon, for the current contour angle point, take the previous contour angle point and the next contour angle point to form two vectors, and calculate the internal angle;

[0031] Step 2.7, use gradient descent method fitting to optimize the corner points, side length and angle of the detected rectangle; initialize the parameters by minimizing the squared distance between the rectangle and the contour, calculate the gradient and update the parameters;

[0032] The above step 3 specifically includes:

[0033] Step 3.1, for the edge image detected by the rectangle, apply Hough circle transformation, and set the x coordinate of the center, the y coordinate of the center and the radius of the circle as the parameter space;

[0034] Step 3.2, traverse each edge point in the edge image, and perform the following operations on each edge point: for each possible combination of center and radius, calculate the center and radius according to the edge point, and perform accumulation operation on the corresponding position in the parameter space to increase the cumulative value of the combination;

[0035] Step 3.3, after the accumulation is completed, determine the detected circle according to the set threshold; the threshold can control the detection sensitivity of the circle, a higher threshold can filter out weaker circle candidates, and a lower threshold can allow more circles to be detected;

[0036] Step 3.4, for each circle that meets the threshold, the center coordinates and radius information can be extracted from the parameter space.

[0037] The above step 4 specifically includes:

[0038] Step 4.1, for each rectangular corner point, take its coordinate as part of the descriptor, in addition, for each corner point, find the two edges connected to it, calculate the angle and edge length of the two edges;

[0039] Step 4.2, normalize the descriptor, normalize the edge length to a specific range [0, 1], and map the angle to a specific range 0 to

[0040] Step 4.3, combine the coordinates of the corner points, the angle and edge length information of the connecting edges into a fixed dimension descriptor vector; the coordinates, angles and edge lengths are connected together to form a longer descriptor vector;

[0041] The above step 6 specifically includes:

[0042] Calculate the scale factor between infrared and visible light images , Equal to the pixel area ratio of the same area in the infrared and visible light images adjusted to the same vertical resolution, convert P2 to the same scale and direction by the following expression:

[0043]

[0044] Where, is the rotation angle calculated in the rotation angle estimation, R is the rotation orthogonal matrix of the infrared and visible light images, the purpose is to rotate the visible light image to the same viewpoint as the infrared image, is the feature point after rotation and scaling transformation.

[0045] The above step 8 specifically includes:

[0046] The projection transformation matrix M is calculated by the least square method, the formula is as follows:

[0047]

[0048] Where, , is the matching point obtained by step 7 rough matching;

[0049] The least square solution of M is called:

[0050]

[0051] The fitting point corresponding to is expressed as:

[0052]

[0053] The residual of and is expressed as:

[0054] ​

[0055] For the matching point pair obtained by step 7 rough matching, the residual is iteratively optimized until it is less than a set threshold.

[0056] The present application has the advantages of:

[0057] Compared with the existing infrared image and visible light image registration based on point or straight line features, the present application proposes a circuit board infrared image and visible light image registration method based on rectangular and circular features, aiming at a large number of rectangular heat-emitting components and circular non-heat holes in the circuit board infrared image. The method successively extracts edge contours, closes the contours, approximates polygons, filters rectangles to obtain rectangular features, and constructs a descriptor unique to the rectangular features. At the same time, the method successively extracts edge contours, accumulates the centers and radii of the circles, filters circular features, and constructs a descriptor unique to the circular features. Compared with the traditional point feature descriptor, our descriptor increases the length of the side, the center of the circle, the radius, etc., and is more unique, avoiding false matching caused by similar feature point descriptors. In addition, in the matching feature point pair stage, our algorithm increases the global optimization algorithm for adjusting the position of the feature points in addition to using the conventional rotation angle estimation, connection line consistency, and RANSAC algorithm, further improving the registration accuracy. BRIEF DESCRIPTION OF DRAWINGS

[0058] Figure 1 Flow chart of the present application based on rectangular and circular feature circuit board infrared image and visible light image registration method;

[0059] Figure 2 Flow chart of extracting rectangular features of circuit board infrared and visible light images;

[0060] Figure 3 Flow chart of extracting circular features of circuit board infrared and visible light images. DETAILED DESCRIPTION

[0061] The present application will be further described below in conjunction with the drawings and specific embodiments.

[0062] As Figure 1 shown, embodiment 1 of the present application provides a circuit board infrared image and visible light image registration method based on rectangular and circular features, including the following steps:

[0063] Step 1: Obtain the circuit board infrared image and the circuit board visible light image taken at the same time, and perform edge detection on the infrared and visible light images using the Canny operator to obtain binary edge images. The edge pixels are marked as white, and the non-edge pixels are marked as black;

[0064] Step 2: Rectangular detection is performed on the infrared and visible light binary edge images of the circuit board obtained in step 1, and rectangular features, including the four corners of the rectangle, the length of the side and the angle, are extracted;

[0065] As Figure 2 shown, in a further preferred but non-limiting embodiment of the present application, step 2 specifically includes:

[0066] Step 2.1, according to the edge image, initialize an empty contour list;

[0067] Step 2.2, traverse each pixel of the edge image. When a white pixel is found, it is marked as the current pixel and the contour tracking process begins. In the contour tracking process, according to the connectivity of adjacent pixels and the boundary conditions in the edge image, it is determined how to move to the next pixel. This process continues until the starting pixel is returned, forming a closed contour. The formed contour is added to the contour list.

[0068] Step 2.3, repeat step 2.2 until the entire edge image is traversed. In this way, all closed contours in the image can be found and extracted.

[0069] Step 2.4, screening the detected contours to select contours that meet the rectangular feature. Here, the area attribute of the contour is screened. Here, the Shoelace formula is selected for polygon area calculation. The Shoelace formula is based on the coordinates of the vertices of the polygon, decomposes the polygon into triangles, calculates the area of the triangle, and then adds the areas. The formula is as follows:

[0070]

[0071] where, n n represents the number of sides of the polygon, represents the coordinates of the i th vertex of the polygon.

[0072] Set the coefficient , if the area of the polygon contour , then delete the contour, where S is the total area of the image.

[0073] Step 2.5, for the screened contour, use the polygon approximation algorithm to approximate the contour to a polygon. We use the Douglas-Peucker algorithm, which simplifies the contour by retaining the key points that are far away from the original curve. The farthest point on the contour is selected as the key point, and then the curve is divided into two sub-curves, which are processed recursively. Define a function to represent the perpendicular distance from point to line segment , that is:

[0074]

[0075] wherein, is the starting point, , is the end point of the resulting curve.

[0076] The starting point on the curve is selected and the end point is selected as the end point of the resulting curve. By iteratively calculating the perpendicular distance of each point between the starting point and the end point, the point with the maximum distance is found If the maximum distance is greater than a given threshold, the current curve is cut at , resulting in two sub-curves. The algorithm is recursively called on both sub-curves until the maximum distance is smaller than the threshold, or no further cutting is possible. Finally, the key points of each sub-curve are connected to obtain the final approximated curve.

[0077] Step 2.6, for the polygon obtained by approximation, whether it is a rectangle is judged by calculating the number of edges of the polygon, the angle. First, according to the number of contour angle corner points, the number of polygon edges is obtained, and the polygon with the number of edges not being 4 is deleted; then the internal angle of the polygon is calculated, for the current contour angle corner point, taking the previous contour angle corner point and the next contour angle corner point, forming two vectors, and calculating the internal angle.

[0078] Step 2.7, using gradient descent method fitting for the detected rectangle to optimize the corner points, side length, angle of the rectangle. The square distance between the rectangle and the contour is minimized as the optimization objective function to initialize the parameters, calculate the gradient, and update the parameters. Objective function:

[0079]

[0080] wherein represents the coordinates of the i th vertex of the polygon.

[0081] Step 3: Perform circle detection on the circuit board infrared and visible light binary edge image obtained in step 1, and extract circular features, including the center of the circle and the radius of the circle;

[0082] As shown in Figure 3 , in the further preferred but non-limiting embodiments of the present application, step 3 specifically comprises:

[0083] Step 3.1, for the edge image obtained by rectangle detection, apply Hough circle transformation, and set the x coordinate representing the center of the circle, the y coordinate of the center of the circle and the radius of the circle as the parameter space.

[0084] Step 3.2, traverse each edge point in the edge image, and for each edge point, do the following: for each possible combination of center and radius, calculate the center and radius according to the edge point. Accumulate the corresponding position in the parameter space to increase the accumulated value of the combination.

[0085] Step 3.3, after the accumulation is completed, determine the detected circles according to the set threshold. The threshold can control the detection sensitivity of the circle, a higher threshold can filter out weaker circle candidates, and a lower threshold can allow more circles to be detected.

[0086] Step 3.4, for each circle that meets the threshold, the center coordinates and radius and other information can be extracted from the parameter space.

[0087] Step 4: using the rectangle and circle information obtained in steps 2 and 3, construct the rectangle feature and circle feature descriptor;

[0088] In further preferred but non-limiting embodiments of the present application, step 4 specifically includes:

[0089] Step 4.1, for each rectangle corner point, its coordinates are taken as part of the descriptor. In addition, for each corner point, find the two edges connected to it, and calculate the angle and length of the two edges.

[0090] Step 4.2, normalize the descriptor, normalize the edge length to a specific range [0, 1], and map the angle to a specific range 0 to .

[0091] Step 4.3, combine the coordinates of the corner points, the angle and length information of the connecting edges into a fixed dimension descriptor vector. The coordinates, angles and lengths are concatenated to form a longer descriptor vector

[0092] Step 5: adopt bilateral matching to preliminarily match the infrared image and the visible light image to obtain initial matching feature point pairs;

[0093] Step 6: apply the rotation angle estimation and connection line consistency theory to remove obvious false matching feature point pairs, and select the RANSAC algorithm for hundreds of iterations to remove the remaining few false matching feature point pairs;

[0094] In further preferred but non-limiting embodiments of the present application, step 6 specifically includes:

[0095] Calculate the scale factor between the infrared and visible light images , which is equal to the ratio of the pixel areas of the same region in the infrared and visible light images adjusted to the same vertical resolution. Convert P2 to the same scale and direction by the following expression:

[0096]

[0097] is the rotation angle calculated in the rotation angle estimation, R is the rotation orthogonal matrix of the infrared and visible light images, and the purpose is to rotate the visible light image to the same viewpoint of the infrared image, is the feature point after the rotation and scaling transformation.

[0098] Step 7: Estimate the projection transformation matrix between the infrared and visible light images by using the coarse matching feature points;

[0099] Step 8: Reduce the position deviation of the feature points by using fine matching based on global iterative optimization.

[0100] In a further preferred but non-limiting embodiment of the present application, step 8 specifically comprises:

[0101] The projection transformation matrix M is calculated by the least square method, and the formula is as follows:

[0102]

[0103] The matching points obtained by the coarse matching of step 7

[0104] The least square solution of M is called:

[0105]

[0106] The fitting point corresponding to is expressed as:

[0107]

[0108] The residual of and is expressed as:

[0109]

[0110] The residual is iteratively optimized for the matching point pair obtained by the coarse matching of step 7 until it is less than a set threshold.

[0111] The present application has the advantages of:

[0112] Compared with the existing infrared image and visible light image registration based on point or straight line features, the application is aimed at a large number of rectangular heat generating components and circular non-heat holes in the circuit board infrared image, and proposes a circuit board infrared image and visible light image registration method based on rectangular and circular features. The method obtains rectangular features by extracting edge contours, closing contours, approximating polygons and screening rectangles in turn, and constructs a unique descriptor for rectangular features. At the same time, the method obtains circular features by extracting edge contours, accumulating the center and radius of the circle, and screening circular features in turn, and constructs a unique descriptor for circular features. Compared with the traditional point feature descriptor, our descriptor increases the information of side length, center, radius, etc., and is more unique, avoiding false matching caused by similar feature point descriptors. In addition, our algorithm increases a global iterative optimization algorithm for adjusting the position of the feature points in the matching feature point pair stage in addition to the conventional rotation angle estimation, connection line consistency, RANSAC algorithm and other matching algorithms, further improving the registration accuracy.

[0113] The applicant of the present application has made a detailed description and explanation of the embodiments of the present application in combination with the drawings of the specification, but those skilled in the art should understand that the above embodiments are only preferred embodiments of the present application, and the detailed description is only to help the reader better understand the spirit of the present application, and is not a limitation on the protection scope of the present application, on the contrary, any improvement or modification based on the spirit of the present application should fall within the protection scope of the present application.

Claims

1. A method for registering infrared and visible light images of a circuit board based on rectangular and circular features, characterized in that, The method comprises the following steps: Step 1: Obtain the infrared image and the visible light image of the circuit board taken at the same time, and perform edge detection on the infrared and visible light images respectively by using a Canny operator to obtain the infrared and visible light binary edge images of the circuit board, wherein the edge pixels are marked as white and the non-edge pixels are marked as black; Step 2: Perform rectangle detection on the infrared and visible light binary edge images of the circuit board obtained in step 1, and extract the rectangle features, including the four corner points, the side length and the angle; Step 3: Perform circle detection on the infrared and visible light binary edge images of the circuit board obtained in step 1, and extract the circle features, including the center and the radius of the circle; Step 4: Construct the rectangle feature and the circle feature descriptors by using the rectangle and circle information obtained in steps 2 and 3; Step 5: Perform bilateral matching to preliminarily match the infrared image and the visible light image to obtain the initial matching feature point pairs; Step 6: Remove the obvious false matching feature point pairs by using the rotation angle estimation and the connection line consistency theory, and remove the remaining small amount of false matching feature point pairs by using the RANSAC algorithm with hundreds of iterations; Step 7: Estimate the projection transformation matrix between the infrared and visible light images by using the coarse matching feature point pairs; Step 8: Reduce the position deviation of the feature points by using the fine matching based on the global iterative optimization.

2. The rectangular and circular feature based circuit board infrared image and visible light image registration method of claim 1, wherein, The step 2 specifically comprises: Step 2.1: According to the edge image, initialize an empty contour list; Step 2.2: Traverse each pixel of the edge image, when a white pixel is found, mark it as the current pixel, and start the contour tracking process, in the contour tracking process, determine how to move to the next pixel according to the connectivity of adjacent pixels and the boundary condition in the edge image, this process will continue until it returns to the starting pixel, forming a closed contour, and the formed contour is added to the contour list; Step 2.3: Repeat step 2.2 until the entire edge image is traversed, and all closed contours in the image are found and extracted; Step 2.

4. Screening the detected contours, selecting the contours that meet the rectangular feature, screening according to the area attribute of the contour, selecting the Shoelace formula for polygon area calculation, the Shoelace formula is based on the vertex coordinates of the polygon, decomposing the polygon into triangles, calculating the area of the triangle, and then adding these areas, the formula is as follows: ; in, This indicates the number of sides of the polygon. Represents the first polygon The coordinates of each vertex; a coefficient is set if the area of ​​the polygon outline... If the outline is deleted, then the outline is deleted, where S is the total area of ​​the image; Step 2.5, for the screened contour, use the polygon approximation algorithm to approximate the contour as a polygon, using the Douglas-Peucker algorithm, simplify the contour by retaining the key points that are far apart in the contour, select the points on the contour that are farthest from the original curve as key points, then divide the curve into two sub-curves, and recursively process them respectively, define a function the perpendicular distance from a point to a line segment , that is: ； Select the starting point on the curve. and end point Using the endpoints of the resulting curve, the point with the maximum distance is found by iteratively calculating the vertical distance from each point to the starting and ending points. If the maximum distance is greater than a given threshold, the current curve will be... The algorithm is then used to cut the curve into two sub-curves. The algorithm is then recursively called on each sub-curve until the maximum distance is less than the threshold or the curve can no longer be cut. Finally, the key points of each sub-curve are connected to obtain the final approximate curve. Step 2.6: For the approximated polygon, judge whether it is a rectangle by calculating the number of sides and the angle of the polygon, first, obtain the number of sides of the polygon according to the number of contour angle corner points, and delete the polygon with a side number of 4; then calculate the internal angle of the polygon, for the current contour angle corner point, take the previous contour angle corner point and the next contour angle corner point to form two vectors, and calculate the internal angle; Step 2.7: Use the gradient descent method to fit the detected rectangle to optimize the corner points, side length and angle of the rectangle; use the square distance between the rectangle and the contour as the optimization objective function to initialize the parameters, calculate the gradient, and update the parameters.

3. The rectangular and circular feature based circuit board infrared image and visible light image registration method of claim 1, wherein, The step 3 specifically comprises: Step 3.1: For the edge image obtained by rectangle detection, apply the Hough circle transformation, and set the x coordinate of the center, the y coordinate of the center and the radius of the circle as the parameter space; Step 3.2: Traverse each edge point in the edge image, and perform the following operations on each edge point: for each possible combination of the center and the radius, calculate the center and the radius according to the edge point, and perform the accumulation operation at the corresponding position in the parameter space to increase the cumulative value of the combination; Step 3.3, after accumulation is completed, determine the detected circle according to the set threshold, the threshold can control the detection sensitivity of the circle, higher threshold can filter out weaker circle candidate, lower threshold can allow more circles to be detected; Step 3.4, for each circle meeting the threshold, extract the center coordinates and radius and other information from the parameter space.

4. The rectangular and circular feature based circuit board infrared image and visible light image registration method of claim 1, wherein, The step 4 specifically comprises: Step 4.1, for each rectangular corner point, take its coordinates as part of the descriptor, in addition, for each corner point, find the two edges connected thereto, and calculate the angle and length of the two edges; Step 4.2, Normalization of the descriptor, normalizing the edge length to a specific range [0, 1] and mapping the angle to a specific range 0 to within; Step 4.3, combine the coordinates of the corner point, the angle and length information of the connecting edge into a fixed dimension descriptor vector, and the coordinates, angle and length are connected together to form a longer descriptor vector.

5. The rectangular and circular feature based circuit board infrared image and visible light image registration method of claim 1, wherein, The step 6 specifically comprises: Computing a scale factor between infrared and visible light images , Equal to the ratio of the pixel areas of the same region in the infrared and visible light images adjusted to the same vertical resolution, P2 is converted to the same scale and orientation by the following expression: ; is the rotation angle calculated in the rotation angle estimation, R is the rotation orthogonal matrix of the infrared and visible light images, and the purpose is to rotate the visible light image to the same viewpoint of the infrared image, is the feature point after the rotation and scaling transformation.

6. The rectangular and circular feature based circuit board infrared image and visible light image registration method of claim 1, wherein, The step 8 specifically comprises: The projection transformation matrix M is calculated by the least square method, and the formula is as follows: ； The least squares solution for M is called: ; corresponding to The fitted points are represented as: ; and The residual representation of ; 、 For the matching point pairs obtained in step 7, the residual is iteratively optimized until it is less than a set threshold.

Images