Rapid cylindrical surface back projection expansion method

Abstract

The invention discloses a rapid cylindrical surface back projection expansion method, which relates to the technical field of automation, and comprises the following steps of: wrapping a flexible calibration plate on the surface of a cylinder, extracting related feature point coordinates, establishing theoretical feature point expansion coordinates according to the features of the calibration plate, and establishing a grid perspective transformation model. The model solves the problem of expansion transformation in x and y directions at one time. In the actual use process, information such as the width and the offset position of a target object is calculated, related feature point coordinates are corrected, then the model is used for carrying out cylinder back projection expansion transformation, the position of the target object does not need to be relatively fixed, expansion transformation in the x-axis direction and the y-axis direction is achieved through one-time transformation, and efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the field of automation technology, and particularly relates to a method for rapid cylindrical back-projection unfolding. Background Art

[0002] Currently, the cylindrical back-projection unfolding transformation in scientific research papers of scientific research institutions and universities is basically carried out based on the basic projection and back-projection formulas given in the basic textbooks of computational photography. This formula requires that the position of the target object has no offset, requires calculating various parameters of the camera lens, and requires a large amount of calculation for each pixel, resulting in low efficiency and unable to meet the real-time requirement of automation. In the prior art, the quadratic curve theory is mostly adopted, and two quadratic curve back-projection models for the x-axis and y-axis are established and stored. During the actual application process, the coefficients of the two quadratic curve models are respectively corrected, and the x-axis unfolding and y-axis unfolding are respectively carried out to meet the movement of the target object within a small range and solve the problem that the position of the target object cannot be offset. Therefore, it is particularly important to invent a method for rapid cylindrical back-projection unfolding.

[0003] The prior art such as the invention application patent with the publication number of CN106023115A discloses a method for cylindrical back-projection transformation. According to the perspective projection principle, the straight line located on the cylinder is projected into a quadratic curve, and the imaging characteristics are characterized by fitting the quadratic curve. The correction processing is respectively carried out in the vertical direction and the horizontal direction, and then the cylindrical image back-projection transformation is realized. When the target object has displacements in the front-back direction and the left-right direction relative to the camera, the coefficients of the quadratic curve are corrected to adapt to the change of its imaging characteristics without re-calibration. The experimental results show that the invention only needs one calibration to meet the requirements of image transformation when the target object moves within a small range, improves the scene adaptability, and the accuracy is equivalent to that of the traditional method.

[0004] The prior art also has the following defects, specifically reflected in: 1. In the prior art, in various scenarios involving image processing related to cylinders, the relative position of the object must be fixed. If the position of the cylinder moves, it is difficult for the image processing system based on the existing algorithms and technologies to accurately identify and process the image information of the cylinder, which will lead to deviations in the image analysis results, inaccurate product detection results, and even the entire image processing process cannot run normally.

[0005] 2. In the prior art, when adopting the quadratic curve theory, two quadratic curve back-projection models for the x-axis and y-axis need to be established respectively, and the coefficients need to be corrected and unfolded respectively in actual application, which increases the complexity of the process, increases the amount of calculation, increases the performance requirements for the hardware device, and leads to a large deviation between the result of the image back-projection unfolding and the actual requirement, seriously affecting the quality of image processing and the accuracy of subsequent analysis. Summary of the Invention

[0006] The object of the present invention is to provide a fast cylindrical back-projection unfolding method, which solves the problems existing in the background art.

[0007] To solve the above technical problems, the present invention adopts the following technical solutions: The present invention provides a fast cylindrical back-projection unfolding method, including: Step 1, model establishment.

[0008] Step 2, acquiring an image of an object.

[0009] Step 3, calculating the boundary, width and other information of the object.

[0010] Step 4, correcting the model.

[0011] Step 5, performing a cylindrical back-projection unfolding transformation using a correction template to generate a new image.

[0012] Preferably, for the model establishment, the specific implementation method is: S1: fabricating a target object, purchasing or printing a calibration board, tightly winding it on the surface of the target object, placing it on an image acquisition platform, and acquiring an image.

[0013] S2: extracting feature points, and calculating the coordinate information, row index and column index of the feature points.

[0014] S3: generating theoretical coordinate information of the feature points, and generating the theoretical coordinates after cylindrical back-projection unfolding of each feature point according to the resolution of the acquired image and the size information of the calibration board.

[0015] S4: calculating a cylindrical back-projection unfolding grid mapping table, establishing a cylindrical back-projection unfolding model, calculating the perspective transformation matrix of every four points according to the perspective principle, and then generating a grid mapping table, thereby establishing a cylindrical back-projection unfolding model.

[0016] S5: saving the model and parameters.

[0017] Preferably, for the perspective principle, the specific analysis method is: If the coordinates of the original image are (x, y) and the coordinates after perspective transformation are (x′, y′), then the perspective transformation matrix is where is the perspective transformation matrix, and the coordinate calculation formula after perspective transformation is:

[0018] The beneficial effects of the present invention are as follows: 1. In the present invention, the absolute fixed relative position of the object during the image processing process is avoided. Without the need for operators to spend a lot of time and effort on precisely calibrating and fixing the object position, the image processing work of cylindrical back-projection unfolding can be carried out quickly and efficiently, reducing the operation difficulty, improving the work efficiency, and significantly broadening the applicable scenario range of this method.

[0019] 2. In the present invention, each image only needs to undergo a one-time inverse mapping transformation, which can simultaneously overcome the problems of unfolding transformation in both the x-axis and y-axis directions. In practical application scenarios, the processing time of each image is shortened, the waiting and conversion time losses caused by multiple operations are reduced, the work efficiency is improved, and strong technical support is provided for the development of related industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] 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 drawings in the following description 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.

[0021] Figure 1 It is a schematic flowchart of the implementation steps of the method of the present invention.

[0022] Figure 2 It is a flowchart for establishing a model.

[0023] Figure 3 It is a schematic diagram of the calibrated board image collected.

[0024] Figure 4 It is a schematic diagram of extracting the feature points and boundaries of the target object.

[0025] Figure 5 It is a schematic diagram of the effect after the cylindrical inverse projection unfolding transformation.

[0026] Figure 6 It is a flowchart for using the model.

[0027] Figure 7 It is a schematic diagram of the image of the collected object.

[0028] Figure 8 It is a schematic diagram of the image after the cylindrical inverse projection unfolding. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0030] Refer to Figure 1 As shown, the present invention provides a fast cylindrical inverse projection unfolding method, including: Step 1, model establishment.

[0031] In a specific embodiment, the model is established with reference to Figure 2 , and the specific implementation method is as follows: S1: Produce the target object, purchase or print a calibration board, tightly wind it around the surface of the target object, and place it on the image acquisition platform to acquire images. With reference to Figure 3 .

[0032] S2: Extract feature points and calculate the coordinate information, row index, and column index of the feature points. The feature points of the checkerboard calibration board are the intersections of black and white blocks, and the feature points of the dot calibration board are the centers of the dots. With reference to Figure 4 , in order to improve the accuracy of the cylindrical back-projection unfolding algorithm, after initially extracting the characteristic coordinates, the sub-pixel corner detection function provided in the OpenCV open-source image processing library is used to extract more accurate corner coordinates.

[0033] It should be noted that for calculating the coordinate information, row index, and column index of the feature points, for the checkerboard calibration board, its feature points are the intersections of black and white blocks. After identifying the checkerboard pattern in the image, the findChessboardCorners function in OpenCV can be used to detect the corners. For the detected corners, their coordinates are the feature point coordinates; after detecting the corners of the checkerboard, the corners are sorted in the order from left to right and top to bottom. Assuming the checkerboard has m rows and n columns, then the row index of the first corner is 0 and the column index is 0; the column index of the second corner in the same row is 1, and so on. When the column index reaches n - 1, the row index of the next corner increases by 1 and the column index is reset to 0. In this way, row indices and column indices are assigned to each feature point.

[0034] S3: Generate the theoretical coordinate information of the feature points, with reference to Figure 5 , and generate the theoretical coordinates after the cylindrical back-projection unfolding of each feature point according to the resolution of the acquired image and the size information of the calibration board.

[0035] S4: Calculate the grid mapping table for cylindrical back-projection unfolding, establish the cylindrical back-projection unfolding model, calculate the perspective transformation matrix for every four points according to the perspective principle, and then generate the grid mapping table, thereby establishing the cylindrical back-projection unfolding model.

[0036] S5: Save the model and parameters.

[0037] In a specific embodiment, for the perspective principle, the specific analysis method is as follows: If the coordinates of the original image are (x, y) and the coordinates after perspective transformation are (x′, y′), then the perspective transformation matrix is where is the perspective transformation matrix, and the coordinate calculation formula after perspective transformation is:

[0038]

[0039] In the present invention, the absolute fixed relative position of an object during the image processing is avoided. Without the need for an operator to spend a great deal of time and effort on precisely calibrating and fixing the position of the object, the image processing work of cylindrical back-projection unfolding can be carried out quickly and efficiently, reducing the operation difficulty, improving the work efficiency, and significantly broadening the applicable scenario range of this method.

[0040] Step Two: Collect the image of the object. Refer to Figure 7 , place the product object to be processed at the collection position for image collection.

[0041] Step Three: Calculate the boundary, width and other information of the object. The product needs to have a large difference from the background so that it is easy to extract the boundary, width and other information of the product. If the product does not have a large difference from the background, the extraction of the product boundary is inaccurate, and subsequent steps cannot be carried out or subsequent steps will be incorrect.

[0042] It should be noted that the other information refers to the center of gravity of the object. By calculating the weighted average of the coordinates of all pixel points within the object contour, the center of gravity coordinates of the object can be obtained, which is very useful when analyzing the position offset or rotation of the object. When calculating the boundary, width and other information of the object, a contour search algorithm is used, such as the findContours function in OpenCV to find all contours in the image. Since there may be some small interfering contours, it is necessary to screen according to the characteristics of the contour area, perimeter, etc., and retain the contour with a larger area and a shape that conforms to the object characteristics. This contour is the boundary of the object. If a calibration board is used during image collection and the proportional relationship between the actual distance of the feature points on the calibration board and the distance of the corresponding feature points in the image is known, the actual width of the object can be calculated by measuring the pixel distance between two points on the object boundary in the image and combining the calibration ratio.

[0043] Step Four: Correct the model. Because the position of the product changes in the front, back, left and right directions, it is necessary to calculate according to the current product information and model parameters, such as deviation, width scaling ratio, etc. Using the calculated information, operations such as offset and scaling are performed on the coordinate of the feature points of the model to generate a new model.

[0044] Step Five: Use the corrected template to perform cylindrical back-projection unfolding transformation, refer to Figure 8 , to generate a new image.

[0045] In the present invention, each image only needs to be subjected to a one-time inverse mapping transformation to simultaneously overcome the unfolding transformation problems in both the x-axis and y-axis directions. In the actual application scenario, the processing time of each image is shortened, the waiting and conversion time loss caused by multiple operations is reduced, the work efficiency is improved, and strong technical support is provided for the development of related industries.

[0046] The above content is only an example and illustration of the concept of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the described specific embodiments or use similar methods for substitution, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and they should all fall within the protection scope of the present invention.

Claims

1. A fast cylindrical back-projection expansion method, characterized in that: include: Step 1: Model building; Step 2: Collecting an image of the object; Step 3: Calculate the boundary, width and other information of the object; Step 4: Correct the model; Step 5: Use the correction template to perform cylindrical back-projection transformation to generate a new image.

2. A fast cylindrical back projection expansion method according to claim 1, characterized in that: The model is established and the specific implementation method is as follows: S1: Make the target object, purchase or print the calibration plate, wrap it tightly around the surface of the target object, place it on the image acquisition platform, and collect images; S2: Extract feature points and calculate the coordinate information, row index and column index of the feature points; S3: Generate theoretical coordinate information of feature points. According to the acquired image resolution and calibration plate size information, generate the theoretical coordinates of each feature point after cylindrical back projection expansion; S4: Calculate the cylindrical back-projection unfolding grid mapping table, establish the cylindrical back-projection unfolding model, calculate the perspective transformation matrix of each four points according to the perspective principle, and then generate the grid mapping table, so as to establish the cylindrical back-projection unfolding model; S5: Save the model and parameters.

3. A fast cylindrical back projection expansion method according to claim 2, characterized in that: The perspective principle and specific analysis method are as follows: If the coordinates of the original image are (x, y) and the coordinates after perspective transformation are (x′, y′), then the perspective transformation matrix is in is the perspective transformation matrix, and the coordinate calculation formula after perspective transformation is:

Images