Pallet pose estimation method based on monocular camera, storage medium and related equipment
By combining a monocular camera with edge detection, corner detection, and sub-pixel corner detection algorithms, the problems of algorithm complexity and high hardware cost in existing technologies are solved, achieving efficient and low-cost estimation of pallet pose. This is applicable to AGV equipment and improves handling speed and accuracy.
Patent Information
- Application Number
- CN202511162500.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-19
- Publication Date
- 2025-12-05
AI Technical Summary
Existing monocular vision pose estimation methods are complex in industrial applications, have high hardware costs, poor real-time performance, are difficult to adapt to pallets of different sizes and shapes, and lack accuracy.
By employing edge detection, corner detection, and sub-pixel corner detection algorithms based on a monocular camera, combined with feature matching and pose estimation algorithms, the pose of the stack is estimated, reducing computational load and hardware requirements while improving response speed and accuracy.
It achieves efficient and low-cost pose estimation for different types of pallets, is suitable for resource-constrained AGV equipment, improves handling speed and accuracy, and reduces computational load and hardware requirements.
Smart Images
Figure CN121074128A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of warehouse logistics equipment, in particular to a pallet pose estimation method based on a monocular camera, a storage medium and related equipment. BACKGROUND
[0002] In the industrial fields of industrial robots, automatic control and quality detection, the measurement of object position and pose has important application value. In engineering applications, the use of industrial cameras to realize automatic handling of pallets by forklifts saves labor costs to a certain extent, and to realize automatic handling requires data on the relative pose of the camera and the pallet. At present, pose measurement technology is mainly divided into laser radar and visual measurement. In the pose measurement method based on laser radar, the distance to the target object is determined by measuring the flight time of each laser beam. Laser radar has a wide detection range and high distance accuracy, but its cost is relatively high and it is easily affected by electromagnetic waves. In visual measurement, visual measurement methods are divided into monocular visual measurement, binocular visual measurement and multi-view visual measurement according to the number of visual sensors. Among them, binocular vision and multi-view vision systems are limited in industrial applications due to the difficulty of feature point matching. The monocular vision system can avoid this problem and only uses one visual sensor to capture a single image for target measurement. However, the existing monocular visual pose estimation method uses a deep learning-based model to identify the key point positions on the pallet. Different samples need to be collected for training, the algorithm is still relatively complex, the hardware cost of the AGV device is increased, and the inference delay is high, which is less real-time. SUMMARY
[0003] In order to overcome the shortcomings of the prior art, the purpose of the present application is to provide a pallet pose estimation based on a monocular camera, which only needs to use a basic visual algorithm to realize pose estimation of different types of pallets, has small calculation amount, is easy to deploy on resource-limited AGV devices, has fast response speed, and can improve the handling speed of the forklift AGV device.
[0004] To solve the above problems, the technical scheme adopted by the present application is as follows: a pallet pose estimation method based on a monocular camera, comprising the following steps: acquiring a fork entry face image of a pallet by a monocular camera; performing feature matching on the fork entry face image and a pre-stored template image to obtain a pallet region in the fork entry face image; preprocessing the pallet region; detecting edges in the pallet region by an edge detection algorithm; detecting corner points near the vertices of each edge in the pallet region by a corner point detection algorithm, and determining projection corner points corresponding to control corner points on the fork entry face of a preset pallet according to the positions of the corner points. iteratively optimize the preliminary coordinates of the projection corner points through a sub-pixel corner point detection algorithm to obtain projection coordinates of the projection corner points in the first coordinate system; According to the 3D world coordinates of the control corner points in the second coordinate system and the projection coordinates, the pose parameters and the position parameters of the pallet in the first coordinate system are solved through a pose estimation algorithm.
[0005] Compared with the prior art, the beneficial effects of the present application are that: first, the corner points in the fork entry face image of the fork entry face of the pallet are detected through an edge detection algorithm and a corner point detection algorithm, and the projection corner points corresponding to the preset control corner points on the fork entry face of the pallet are determined according to the relative positional relationship between the corner points, and then the relative pose of the monocular camera relative to the pallet can be estimated through a pose estimation algorithm according to the projection coordinates of the projection corner points and the 3D world coordinates of the control corner points, so as to adjust the position of the fork of the equipment relative to the fork entry of the pallet, and realize the handling of the pallet and the goods on the pallet. This method can realize the pose estimation of the pallet only through basic vision algorithms such as edge detection algorithm and corner point detection algorithm, without the need for sample collection and training of different models of pallets of different sizes and shapes in advance, and has small calculation amount, low hardware demand, and is easy to deploy in AGV equipment with limited resources, and has high response speed and high real-time performance. This method further iteratively optimizes the determined projection coordinates through a sub-pixel corner point detection algorithm to obtain high-precision projection coordinates of the projection corner points, thereby improving the accuracy of the pose estimation and solving the precision problem caused by the basic vision algorithm.
[0006] The above-mentioned pallet pose estimation method, in the step of detecting corner points near the vertices of each edge in the pallet region through a corner point detection algorithm and determining projection corner points corresponding to preset control corner points on the fork entry face of the pallet according to the positions of the corner points, the control corner points include a left upper corner point and a left lower corner point of the left fork entry of the pallet on the fork entry face, and a right upper corner point and a right lower corner point of the right fork entry of the pallet on the fork entry face.
[0007] The above-mentioned pallet pose estimation method further comprises the steps of: Calibrating the intrinsic parameters and distortion parameters of the monocular camera; In the step of solving the pose parameters and the position parameters of the pallet in the first coordinate system through a pose estimation algorithm according to the 3D world coordinates of the control corner points in the second coordinate system and the projection coordinates, the intrinsic parameters and the distortion parameters are taken as input parameters of the pose estimation algorithm.
[0008] The above-mentioned pallet pose estimation method, the step of preprocessing the pallet region comprises: graying the pallet region through a weighted average algorithm; The bilaterally filtered algorithm is used to filter the grayscale stack plate region.
[0009] The step of matching the forklift entrance plane image with the pre-stored template image to obtain the stack plate region in the forklift entrance plane image includes: The SIFT operator is used to extract the key point features in the forklift entrance plane image. The FLANN algorithm is used to match the key point features in the forklift entrance plane image and the template image, and the position of the stack plate region in the forklift entrance plane image is determined according to the matching result.
[0010] In the step of detecting the edges in the stack plate region by using the edge detection algorithm, the Canny edge detection algorithm is used to detect the edges in the stack plate region.
[0011] In the step of detecting the corner points near the vertices of each edge in the stack plate region by using the corner detection algorithm, and determining the projection corner points corresponding to the preset control corner points on the forklift entrance plane of the stack plate according to the positions of the corner points, the Shi-Tomasi corner detection algorithm is used to detect the corner points near the vertices of each edge in the stack plate region. In the step of iteratively optimizing the preliminary coordinates of the projection corner points by using the sub-pixel corner detection algorithm to obtain the projection coordinates of the projection corner points in the first coordinate system, the cornerSubPix sub-pixel corner detection algorithm is used to iteratively optimize the preliminary coordinates to obtain the projection coordinates.
[0012] In the step of solving the attitude parameters and position parameters of the stack plate in the first coordinate system according to the 3D world coordinates of the control corner points in the second coordinate system and the projection coordinates by using the pose estimation algorithm, the solvePnP pose estimation algorithm is used to solve the attitude parameters and position parameters of the stack plate in the first coordinate system according to the 3D world coordinates, the projection coordinates, the intrinsic parameters and the distortion parameters.
[0013] A computer-readable storage medium stores a computer program, and when the computer program is called and executed by a processor, the above-mentioned stack plate pose estimation method based on a monocular camera is realized.
[0014] A forklift type handling device includes a monocular camera, a processor and a memory, the monocular camera and the memory are electrically connected with the processor, and the processor can realize the above-mentioned stack plate pose estimation method based on a monocular camera by calling and executing the computer program in the memory.
[0015] The application will be described in further detail below with reference to the drawings and specific embodiments. BRIEF DESCRIPTION OF DRAWINGS
[0016] Figure 1 Flow chart of the stack pose estimation method based on monocular camera for the embodiment of the application.
[0017] Figure 2 Schematic diagram of controlling the position of the corner point for the embodiment of the application.
[0018] Figure 3 Principle schematic diagram of the pose estimation method for the embodiment of the application.
[0019] Figure 4 Forklift type carrying device for the embodiment of the application. DETAILED DESCRIPTION
[0020] The embodiments of the application will be described in detail below with reference to the drawings and specific embodiments. Figure 1 The embodiments of the application provide a stack pose estimation method based on monocular camera, comprising the following steps: acquiring the forking opening plane image of the stack by the monocular camera; performing feature matching on the forking opening plane image and the pre-stored template image to obtain the stack area in the forking opening plane image; preprocessing the stack area; detecting the edges in the stack area by an edge detection algorithm; detecting the corner points near the vertices of each edge in the stack area by a corner point detection algorithm, and determining the projection corner points corresponding to the control corner points on the forking opening plane of the pre-set stack according to the positions of the corner points; iteratively optimizing the preliminary coordinates of the projection corner points by a sub-pixel corner point detection algorithm to obtain the projection coordinates of the projection corner points in the first coordinate system; solving the attitude parameters and position parameters of the stack in the first coordinate system by a pose estimation algorithm according to the 3D world coordinates of the control corner points in the second coordinate system and the projection coordinates.
[0021] The method, by collecting the forklift entry face image of the pallet, and according to the forklift entry face image of the pallet and the actual mechanical parameters of the pallet, realizes the pose estimation of the pallet, simplifies the pose estimation of the 3D pallet to the pose estimation of the forklift entry face, and reduces the calculation amount of the pose estimation algorithm. At the same time, in the whole pose estimation process, only basic visual algorithms such as edge detection algorithm, corner detection algorithm and sub-pixel corner detection algorithm are needed to realize the pose estimation of the pallet, avoiding a large amount of pre-training for pallets of different shapes and sizes. Only the front image of the forklift entry face of the pallet needs to be collected in advance and the position of the control corner point needs to be selected, and the actual relative distance between the control corner points needs to be input, which has stronger applicability. At the same time, compared with the pose estimation method using deep learning algorithm, the method only uses basic visual algorithm, has lower calculation amount, occupies less hardware resources, is easy to deploy in forklift AGV equipment with limited hardware resources, and has faster response speed due to lower calculation amount, which can improve the estimation efficiency of the pose of the pallet. The method also iteratively optimizes the preliminary coordinates of the determined projection corner point through the sub-pixel corner detection algorithm, improves the resolution of the forklift entry face image through the interpolation algorithm, so as to obtain the projection coordinates of the sub-pixel precision of the projection corner point, and makes up for the deviation that may be introduced by the basic visual algorithm to detect the corner point, so as to reduce the calculation amount while ensuring the accuracy of the estimated pose.
[0022] In some embodiments, in order to further improve the adaptability of the pallets of different shapes and sizes, the template images of different models of pallets can be collected in advance and stored in the storage medium of the device to form a template library. The forklift entry face image is matched with each template image in the template library, the template image with the highest matching rate is selected, and then the forklift entry face image is matched with the template image with the highest matching rate to determine the area of the pallet in the forklift entry face image, i.e. the pallet area. The subsequent process only calculates and processes the pallet area, further reduces the calculation amount of the whole process, improves the response speed, and avoids background interference.
[0023] In the embodiment, the key point features in the entry fork face image of the stack are extracted by the SIFT operator of OpenCV, and the key point features in the entry fork face image are matched with the key point features in the template image by the FLANN algorithm, so as to realize the preliminary positioning of the stack region. It can be understood that in some embodiments, the SURF operator, the ORB operator or the AKAZE operator can also be used to extract the key point features in the entry fork face image of the stack. In some embodiments, in order to further improve the matching accuracy, distance filtering can be performed on the matched key point pairs, the distance between the matched key point pairs is calculated, and abnormal matching with a distance exceeding a preset threshold is removed. And through the findHomography() function or the estimateAffinePartial2D() function in OpenCV, combined with the RANSAC algorithm, the geometric consistency of the matched key point pairs is verified, the outliers are removed, and the homography matrix of the stack region is directly obtained, so as to improve the matching accuracy.
[0024] In the embodiment, in order to highlight the edge information in the entry fork face image, the stack region is subjected to grayscale and filtering pretreatment. The grayscale is realized by a weighted average algorithm, and the R, G and B components of each pixel in the entry fork face image are weighted and averaged according to different weights, and the weights of the R, G and B components are usually [0.299, 0.587, 0.114]. The filtering adopts a bilateral filtering algorithm, which eliminates the interference signals in the entry fork face image and retains the edge information in the image.
[0025] In the embodiment, the edges in the pretreated stack region are detected by the Canny edge detection algorithm. Specifically, the pretreated stack region image is first smoothed by a Gaussian filter, and then the image is smoothed by the first derivative of a two-dimensional Gaussian function. It can be understood that the Gaussian kernel size and the standard deviation of the Gaussian filter can be dynamically adjusted according to the Gaussian noise of the stack region, so as to reduce the amount of calculation while ensuring the details of the edge features. For example, when the Gaussian noise is less than a first noise threshold, such as 5, a smaller Gaussian kernel size of 3x3 or 5x5 is selected, and a standard deviation of 0.5-1.5 is used; when the Gaussian noise is between the first noise threshold and a second noise threshold, such as greater than or equal to 5 and less than or equal to 20, a Gaussian kernel size of six times the standard deviation plus 1 is selected, and a standard deviation of 1.5-3.0 is used; when the Gaussian noise is above the second noise threshold, a larger Gaussian kernel size of 15x15 is selected, and a standard deviation of 3.0 or more is used for Gaussian filtering.
[0026] After filtering, the Sobel horizontal and vertical operators are used to convolve with the stack area image after Gaussian filtering to calculate the gradient amplitude and direction of the image, and the gradient image of the stack area is obtained. Then, the non-maximum elements in the gradient amplitude are suppressed, and the non-maximum suppression is performed on the amplitude along the gradient direction to more accurately determine the boundary of the stack fork. Finally, the edges in the gradient image are detected by double threshold detection, which reduces the number of false edges and ensures the closure of the stack fork face edge after edge connection. The two threshold values of the double threshold detection can be set according to experience, or the basic values of the high and low threshold values can be determined according to the threshold() function or histogram analysis results, or according to the mean value (meanGrad) and standard deviation (stdDevGrad) of the gradient amplitude of the gradient image, and then the high threshold value = meanGrad + k*stdDevGrad, the low threshold value = ratio*highThreshold, the initial values of the coefficients k and ratio are set according to experience, and the coefficients k and ratio are adjusted according to the edge detection effect, such as the number of edge pixels per unit area and the number of breakpoints of the edge.
[0027] It can be understood that the position of the control corner point can be selected from the corners of the stack or the fork on the fork face, and in this embodiment, the upper left corner point and the lower left corner point of the left fork on the fork face, and the upper right corner point and the lower right corner point of the right fork on the fork face are selected as the control corner points, i.e. Figure 2 P0, P1, P2 and P3 in the formula. The curvature and gradient of each pixel point of the stack area are calculated by the Shi-Tomasi algorithm in OpenCV to detect the corner points. It can be understood that the mechanical parameters of the fork face of the stack corresponding to the template image are stored in the storage medium, including the number of corner points and the horizontal and vertical distances between the corner points, and the maximum number of corner points and the minimum distance between the corner points of the Shi-Tomasi algorithm are determined according to the mechanical parameters.
[0028] After detecting the corner points by the Shi-Tomasi algorithm, the projection corner points corresponding to the control corner points need to be screened from the detected corner points. The edge profile where the detected corner points are located can be matched with the edge profile where the control corner points are located, the four corner points with the highest matching degree are determined as the projection corner points corresponding to the control corner points, and the order of the four projection corner points p0, p1, p2 and p3 is given according to the spatial order of the four control corner points P0, P1, P2 and P3, so as to facilitate the subsequent PnP pose estimation. In this embodiment, the initial coordinates of the four projection corner points p0, p1, p2 and p3 are iteratively optimized by the cornerSubPix sub-pixel corner point detection algorithm to obtain the projection coordinates of the four projection corner points with sub-pixel accuracy.
[0029] In the embodiment, the pose parameters and the position parameters of the monocular camera relative to the stacker are solved according to the 3D world coordinates of the control corner points in the second coordinate system, i.e., the 3D world coordinate system, and the projection coordinates of the projection corner points in the first coordinate system, i.e., the monocular camera coordinate system, by using a solvePnP or solvePnPRansac pose estimation algorithm, and the pose of the stacker is obtained. Since the solvePnP or solvePnPRansac pose estimation algorithm needs to input the intrinsic parameters and the distortion parameters of the camera, the intrinsic parameters and the distortion parameters of the monocular camera also need to be calibrated. In the embodiment, the intrinsic parameters and the distortion parameters of the monocular camera are calibrated by using a Zhang Zhengyou calibration method, and the intrinsic parameters and the distortion parameters of the camera are calibrated by collecting 15 to 20 images of the checkerboard calibration board at different observation angles and distances by using the monocular camera, and extracting the corner points of the checkerboard.
[0030] Referring to Figure 2 and Figure 3 In the embodiment, the control corner point P0 is taken as the origin of the 3D world coordinate system, and the 3D world coordinates (0, 0, 0), (W, 0, 0), (W, H, 0) and (0, H, 0) of the four control corner points P0, P1, P2 and P3 are determined according to the horizontal distance W and the vertical distance H between the control corner points, and the four control corner points, the corresponding four projection coordinates, the intrinsic parameters and the distortion parameters of the monocular camera are input into the solvePnP or solvePnPRansac pose estimation algorithm to solve the pose parameters R and the position parameters T of the monocular camera relative to the stacker, and the pose estimation of the stacker is completed.
[0031] In some embodiments, if the fork entrance face image of the stacker has been collected once and the pose estimation has been performed once according to the collection result, the pose estimation result of the last time can be taken as the initial value and input into the solvePnP pose estimation algorithm, and useExtrinsicGuess is set to True to improve the convergence speed and the accuracy of the pose estimation.
[0032] Based on the same inventive concept, the embodiments of the present application also provide a computer-readable storage medium, and the storage medium stores a computer program. When the computer program is executed by a processor, the computer program can implement the above-mentioned stacker pose estimation method based on a monocular camera.
[0033] In some possible implementation manners, each aspect of the stacker pose estimation method based on a monocular camera provided by the present application can also be implemented in the form of a program product, which includes program code for causing the control device to perform the steps in the stacker pose estimation method based on a monocular camera according to various exemplary embodiments of the present application described above in the specification when the program product runs on the device.
[0034] Based on the same inventive concept, referring to Figure 4 Embodiments of the present application also provide a forklift truck for implementing the above-mentioned stack board pose estimation method based on a monocular camera, comprising a monocular camera, a processor and a memory, the monocular camera is arranged on a mast of a fork, and the memory and the monocular camera are electrically connected to the processor. The processor is used to execute the computer program stored in the memory, so as to implement the above-mentioned stack board pose estimation method based on a monocular camera.
[0035] In a possible design, the processor can include one or more processing units, and the processor and the memory can be implemented on the same chip or implemented on separate chips respectively. The processor can be a general-purpose processor, such as a central processing unit (CPU), a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, and can implement or execute the methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the stack board pose estimation method based on a monocular camera disclosed in the embodiments of the present application can be directly embodied as execution completed by a hardware processor, or executed by a combination of hardware and software modules in the processor.
[0036] The memory is a non-volatile computer readable storage medium, which can be used to store non-volatile software programs, non-volatile computer executable programs and modules. The memory can include at least one type of storage medium, for example, can include flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. The memory is any other medium capable of carrying or storing desired program code in the form of instructions or data structures and capable of being accessed by a computer, but is not limited to this. The memory in the embodiments of the present application can also be a circuit or other any device capable of realizing a storage function, used to store program instructions and / or data.
[0037] The processor can be programmed to solidify the code corresponding to the single camera based stack board pose estimation method introduced in the foregoing embodiments into the chip, so that the chip can execute the steps of the single camera based stack board pose estimation method of the embodiments of the present application at runtime. How to program the processor is a known technology to those skilled in the art, which will not be described here.
[0038] It should be noted that, in the description of the present application, if there is a description of orientation, such as the orientation or position relationship indicated by up, down, front, back, left, right, etc., it is based on the orientation or position relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, and does not indicate or imply that the device or element referred to must have a particular orientation, be constructed or operated in a particular orientation, and cannot be understood as a limitation on the present application.
[0039] In the description of the present application, one or more is meant to be one or more, two or more is meant to be two or more, greater than, less than, more than, etc. are understood to not include the number, above, below, etc. are understood to include the number. If there is a description of first or second, etc., it is only for the purpose of distinguishing technical features, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features or the sequence of indicated technical features.
[0040] The present application is described with reference to flowcharts and / or block diagrams according to the methods, devices (systems) and computer program products of the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks.
[0041] These computer program instructions can also be stored in a computer readable memory that can guide the computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer readable memory produce a product including instruction devices, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks.
[0042] These computer program instructions can also be loaded into a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 Figure 1
[0043] In the description of the present application, unless otherwise explicitly defined, the words such as setting, installing, connecting, etc. should be understood in a broad sense, and the person skilled in the art can reasonably determine the specific meaning of the above words in the present application in combination with the specific content of the technical solutions.
[0044] The above embodiments are only preferred embodiments of the present application, and cannot be used to limit the protection scope of the present application. Any non-essential changes and replacements made by the person skilled in the art on the basis of the present application shall fall within the protection scope of the present application.
Claims
1. A monocular camera based stack plate pose estimation method, characterized in that, The method comprises the following steps: acquiring a pallet entry port plane image by a monocular camera; performing feature matching between the entry port plane image and a pre-stored template image to obtain a pallet area in the entry port plane image; preprocessing the pallet area; detecting edges in the pallet area by an edge detection algorithm; detecting corner points near vertices of each edge in the pallet area by a corner detection algorithm, and determining projection corner points corresponding to preset control corner points on a pallet entry port plane according to positions of the corner points; iteratively optimizing preliminary coordinates of the projection corner points by a sub-pixel corner point detection algorithm to obtain projection coordinates of the projection corner points in a first coordinate system; solving attitude parameters and position parameters of a pallet in the first coordinate system by a pose estimation algorithm according to 3D world coordinates of the control corner points in a second coordinate system and the projection coordinates.
2. The pallet pose estimation method of claim 1, wherein, In the step of detecting corner points near vertices of each edge in the pallet area by a corner detection algorithm, and determining projection corner points corresponding to preset control corner points on a pallet entry port plane according to positions of the corner points, the control corner points comprise a left upper corner point and a left lower corner point of a left pallet entry port on the entry port plane, and a right upper corner point and a right lower corner point of a right pallet entry port on the entry port plane.
3. The pallet pose estimation method of claim 1, wherein, The method further comprises the following steps: calibrating intrinsic parameters and distortion parameters of the monocular camera; in the step of solving attitude parameters and position parameters of a pallet in the first coordinate system by a pose estimation algorithm according to 3D world coordinates of the control corner points in a second coordinate system and the projection coordinates, the intrinsic parameters and the distortion parameters are taken as input parameters of the pose estimation algorithm.
4. The pallet pose estimation method of claim 1, wherein, The step of preprocessing the pallet area comprises: graying the pallet area by a weighted average algorithm; filtering the graying pallet area by a bilateral filtering algorithm.
5. The pallet pose estimation method according to any one of claims 1 to 4, characterized in that, The step of performing feature matching between the entry port plane image and a pre-stored template image to obtain a pallet area in the entry port plane image comprises: extracting key point features in the entry port plane image by a SIFT operator; matching key point features in the entry port plane image and the template image by a FLANN algorithm, and determining a position of the pallet area in the entry port plane image according to a matching result.
6. The pallet pose estimation method according to any one of claims 1 to 4, characterized in that, In the step of detecting edges in the pallet area by an edge detection algorithm, the edges in the pallet area are detected by a Canny edge detection algorithm.
7. The pallet pose estimation method of claim 1, wherein, In the step of detecting corner points near the vertices of each edge in the pallet region by a corner point detection algorithm, and determining a projection corner point corresponding to a preset control corner point on the entry fork face of the pallet according to the position of the corner point, the corner points near the vertices of each edge in the pallet region are detected by a Shi-Tomasi corner point detection algorithm. In the step of iteratively optimizing the preliminary coordinates of the projection corner point by a sub-pixel corner point detection algorithm to obtain the projection coordinates of the projection corner point in the first coordinate system, the preliminary coordinates are iteratively optimized by a cornerSubPix sub-pixel corner point detection algorithm to obtain the projection coordinates.
8. The pallet pose estimation method of claim 3, wherein, In the step of solving the attitude parameter and the position parameter of the pallet in the first coordinate system by a pose estimation algorithm according to the 3D world coordinates of the control corner point and the projection coordinates, the attitude parameter and the position parameter of the pallet in the first coordinate system are solved by a solvePnP pose estimation algorithm according to the 3D world coordinates, the projection coordinates, the intrinsic parameter and the distortion parameter.
9. A computer readable storage medium storing a computer program, characterized in that, The computer program is called and executed by the processor to implement the monocular camera-based pallet pose estimation method according to any one of claims 1 to 8.
10. A fork truck handling device characterized by, The monocular camera-based pallet pose estimation method according to any one of claims 1 to 8 is implemented by calling and executing the computer program in the memory by the processor. The monocular camera-based pallet pose estimation method according to any one of claims 1 to 8 is implemented by calling and executing the computer program in the memory by the processor.