Data processing method, device and equipment for intelligent navigation and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a data processing method, device and equipment for intelligent navigation and a storage medium, and the method comprises the following steps: obtaining first point cloud data at a first moment and second point cloud data at a second moment; if the target identification image is scanned at the second moment, decoding the target identification image to obtain target pose data corresponding to the target identification image; and processing the first point cloud data and the second point cloud data based on the target pose data to determine target navigation map information. According to the method, the high-precision target pose data is analyzed through the identification image, and the point cloud data is updated through the target pose data to obtain the high-precision point cloud data, so that the high-precision navigation map information is determined through the high-precision point cloud data.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and particularly relates to a data processing method, device, equipment and storage medium for intelligent navigation. Background Art

[0002] An AGV (Automated Guided Vehicle) cart can, in a factory building with a relatively stable equipment layout environment, start from a fixed point and travel along a predefined route according to system scheduling instructions in the formulated operation map, and perform cargo transfer and handling between different devices. When performing tasks in an unknown environment, it is necessary to determine the operation map before corresponding tasks can be executed. The existing data processing methods for intelligent navigation mainly use a matching processing method, that is, directly matching the point cloud data collected at two adjacent moments to determine the map. However, when collecting point cloud data, it is affected by external factors, resulting in a low-precision determined map. Summary of the Invention

[0003] The present application aims to at least solve one of the technical problems existing in the related art. For this purpose, embodiments of the present application provide a data processing method, device, equipment and storage medium for intelligent navigation, which can determine high-precision navigation map information.

[0004] In a first aspect, an embodiment of the present application provides a data processing method for intelligent navigation, including:

[0005] Obtaining first point cloud data at a first moment and second point cloud data at a second moment;

[0006] If a target identification image is scanned at the second moment, decoding the target identification image to obtain target pose data corresponding to the target identification image;

[0007] Processing the first point cloud data and the second point cloud data based on the target pose data to determine target navigation map information.

[0008] In a second aspect, an embodiment of the present application provides a data processing device for intelligent navigation, including:

[0009] An obtaining module, configured to obtain first point cloud data at a first moment and second point cloud data at a second moment;

[0010] A decoding module, configured to, if a target identification image is scanned at the second moment, decode the target identification image to obtain target pose data corresponding to the target identification image;

[0011] A navigation map information determination module, configured to process the first point cloud data and the second point cloud data based on the target pose data to determine target navigation map information.

[0012] In a third aspect, an embodiment of the present application further provides an electronic device, including a memory storing multiple computer programs; a processor loads computer programs from the memory to execute any one of the data processing methods for intelligent navigation provided by the embodiments of the present application.

[0013] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium storing multiple computer programs, and the computer programs are suitable for being loaded by a processor to execute any one of the data processing methods for intelligent navigation provided by the embodiments of the present application.

[0014] In a fifth aspect, an embodiment of the present application further provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements any one of the data processing methods for intelligent navigation provided by the embodiments of the present application.

[0015] In the embodiments of the present application, high-precision target pose data is parsed from the identification image, and then the point cloud data is updated through the target pose data to obtain high-precision point cloud data, so that high-precision navigation map information can be determined through the high-precision point cloud data. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present application. For those skilled in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0017] Figure 1 is a schematic flowchart of the data processing method for intelligent navigation provided in the embodiments of the present application;

[0018] Figure 2 is a schematic flowchart of time information alignment provided in the embodiments of the present application;

[0019] Figure 3 is a schematic diagram of the overall framework of Cartographer provided in the embodiments of the present application;

[0020] Figure 4 is a schematic diagram of the error pose provided in the embodiments of the present application;

[0021] Figure 5 is a schematic diagram of scheduling provided in the embodiments of the present application;

[0022] Figure 6 It is a schematic diagram of the overall solution process provided by the embodiments of the present application;

[0023] Figure 7 It is a schematic structural diagram of a data processing device for intelligent navigation provided by the embodiments of the present application;

[0024] Figure 8 It is a schematic structural diagram of an electronic device provided by the embodiments of the present application. Detailed implementation manners

[0025] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application. At the same time, in the description of the embodiments of the present application, terms such as "first" and "second" are only used for distinguishing descriptions, and cannot be understood as indicating or implying relative importance. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more features. In the description of the embodiments of the present application, "a plurality of" means two or more, unless otherwise specifically defined.

[0026] The embodiments of the present application provide a data processing method, device, equipment and storage medium for intelligent navigation. Specifically, the embodiments of the present application will be described from the perspective of a data processing device for intelligent navigation. The data processing device for intelligent navigation may include intelligent operation devices such as SLAM (Simultaneous Localization and Mapping) intelligent cars and SLAM intelligent robots.

[0027] It should be noted that the description order of the following embodiments does not limit the preferred order of the embodiments. Although the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown in the drawings.

[0028] The embodiments of the present application take a data processing device for intelligent navigation as an execution subject for illustration. The embodiments of the present application take a SLAM intelligent car as an example. The SLAM intelligent car can be applied to logistics, warehousing, etc. Therefore, the navigation map information determined in the embodiments of the present application may include logistics plant scene navigation map information, warehouse plant scene navigation map information, etc.

[0029] The following will be described in detail with reference to the accompanying drawings respectively. Refer to Figure 1 , Figure 1It is a schematic flowchart of the data processing method for intelligent navigation provided in the embodiments of the present application. The specific process of the data processing method for intelligent navigation provided in the embodiments of the present application can be as follows: steps 101 to 103, including:

[0030] Step 101: Obtain the first point cloud data at the first moment and the second point cloud data at the second moment.

[0031] It should be noted that the SLAM intelligent vehicle in the embodiments of the present application includes multiple data sensors. The multiple data sensors include lidar, cameras, inertial measurement units (IMUs), etc. Therefore, the SLAM intelligent vehicle can collect point cloud data of the current environment through the lidar, collect images of the environment ahead during travel through the cameras, and obtain the attitude (direction) and acceleration of the intelligent vehicle in three-dimensional space through the IMU inertial measurement unit.

[0032] Optionally, at the first moment, the SLAM intelligent vehicle collects the first point cloud data of the current environment through the lidar. At the second moment, which is the next moment after the first moment, the SLAM intelligent vehicle collects the second point cloud data of the current environment through the lidar. Among them, the time difference between the first moment and the second moment can be set according to the actual situation. For example, the difference between the first moment and the second moment is 5s, 10s, etc.

[0033] Step 102: If a target identification image is scanned at the second moment, decode the target identification image to obtain the target pose data corresponding to the target identification image.

[0034] Optionally, during the process of collecting the point cloud data of the current environment through the lidar, due to the influence of external factors, the accuracy of the collected point cloud data is affected. Therefore, the SLAM intelligent vehicle detects the image of the environment ahead collected through the camera at the second moment to determine whether there is an identification image in the image of the environment ahead collected through the camera at the second moment. In one embodiment, the identification image can be a QR code image.

[0035] Optionally, if a target identification image is scanned in the image of the environment ahead collected through the camera at the second moment, the SLAM intelligent vehicle decodes the target identification image to obtain the target pose data corresponding to the target identification image. The specific process is described in steps 1021 to 1024. Among them, the pose data corresponding to each identification image is manually measured and calibrated, so the accuracy of the pose data corresponding to the identification image is high.

[0036] Step 103: Process the first point cloud data and the second point cloud data based on the target pose data to determine the target navigation map information.

[0037] Optionally, the SLAM intelligent vehicle processes the first point cloud data and the second point cloud data according to the target pose data to determine the target navigation map information. The processing may include update processing and matching processing. Therefore, it can be understood that the SLAM intelligent vehicle performs update processing and matching processing on the first point cloud data and the second point cloud data according to the target pose data to determine the target navigation map information, specifically as in Steps 1031 to 1033.

[0038] In the embodiment of the present application, highly accurate target pose data is parsed from the identification image, and then the point cloud data is updated by the target pose data to obtain highly accurate point cloud data, so that highly accurate navigation map information is determined through the highly accurate point cloud data.

[0039] In an optional embodiment, processing the first point cloud data and the second point cloud data based on the target pose data to determine the target navigation map information includes:

[0040] Step 1031: Perform update processing on the second point cloud data based on the target pose data to obtain the target point cloud data at the second moment.

[0041] Optionally, the SLAM intelligent vehicle performs update processing on the second point cloud data according to the target pose data to obtain the target point cloud data at the second moment, specifically as described in Steps 10311 to 10314.

[0042] Optionally, if the target identification image is not scanned in the image of the environment ahead of the vehicle collected by the camera at the second moment, the SLAM intelligent vehicle obtains the IMU data of the intelligent vehicle in the three-dimensional space through the IMU inertial measurement unit, and performs update processing on the second point cloud data through the IMU data to obtain the target point cloud data at the second moment.

[0043] Step 1032: Perform matching processing based on the target point cloud data and the first point cloud data to determine the local sub-map information at the second moment.

[0044] Optionally, the SLAM intelligent vehicle performs matching processing according to the target point cloud data and the first point cloud data to determine the local sub-map information at the second moment.

[0045] Step 1033: Determine the target navigation map information based on the local sub-map information at all moments.

[0046] Optionally, the SLAM intelligent vehicle fuses the local sub-map information at all times into the sub-navigation map information to determine the target navigation map information, which is specifically described in steps 10331 to 10333.

[0047] In the embodiment of the present application, highly accurate target pose data is parsed from the identification image, and then the point cloud data is updated through the target pose data to obtain highly accurate point cloud data, so that highly accurate navigation map information is determined through the highly accurate point cloud data.

[0048] Optionally, before decoding the target identification image to obtain the target pose data corresponding to the target identification image if the target identification image is scanned at the second moment, it further includes:

[0049] Performing edge detection on the acquired image according to the gradient to obtain the gradient values of each pixel point in the image, and determining each edge information in the image according to the gradient values of each pixel point;

[0050] Eliminating the non-linear edge information in each edge information to obtain the target edge information in each edge information;

[0051] Encoding the target edge information, and determining a plurality of polygon images based on the nested relationship between the encoded target edge information;

[0052] Scanning based on the convex hull area and the polygon area of the plurality of polygon images until the target identification image is scanned.

[0053] It should be noted that in the embodiment of the present application, AprilTag (a two-dimensional code marker vision benchmark system) is used for identifying and detecting the identification in the image, and AprilTag has a very small data payload (4 to 12-bit encoding). Compared with the identification, AprilTag can be automatically detected and positioned at a farther distance, with very low resolution, uneven illumination, strange rotation, or when hidden in the corner of another cluttered image. Therefore, multiple tags in a single image can be detected through AprilTag.

[0054] Optionally, edge detection is performed on the captured image according to the gradient to obtain the gradient value of each pixel in the image, and the edge information in the image is determined according to the gradient value of each pixel, wherein the gradient is a vector operator that represents the spatial rate of change, which can be used to measure the degree of grayscale change of the image at different positions. The edge is the place where the grayscale value changes greatly in the image. Therefore, the edge information in the image can be found by calculating the gradient value of each pixel in the image. The algorithm for edge detection can be a local binary pattern algorithm, a Canny operator algorithm, etc. The embodiment of the present application preferentially performs edge detection through the Canny operator algorithm. Although the detected edge information will be reduced, the noise in the image is correspondingly reduced.

[0055] Taking the Canny operator algorithm as an example, the specific steps are as follows: Convert the color image to a grayscale image for subsequent edge detection. Common grayscale conversion formulas can be used, such as weighted average of the pixel values ​​of the three RGB channels. Smooth the grayscale image to reduce the impact of noise on edge detection. Common smoothing filters include Gaussian filters and median filters. Use gradient operators (such as Sobel, Prewitt, etc.) to calculate the gradient value of each pixel in the image. The gradient represents the degree of grayscale change in the image and is often used to detect edges. Perform non-maximum suppression on the gradient image to retain the local maximum value in the gradient direction and suppress non-edge points. For each pixel, compare its gradient value with the gradient values ​​of the two neighboring pixels along the gradient direction, and set the non-maximum point to 0. According to the set high threshold and low threshold, the gradient image is divided into three parts: strong edge, weak edge and non-edge. The gradient value of the strong edge pixel is higher than the high threshold, the gradient value of the non-edge pixel is lower than the low threshold, and the gradient value of the weak edge pixel is between the two. Weak edges are associated with strong edges by connecting pixels to form complete edge lines. Common edge connection methods include connectivity-based methods (such as 8-connectivity or 4-connectivity) and edge tracking-based methods.

[0056] Optionally, the required quadrilateral pattern is found in the edge image and screened, thereby eliminating the non-straight edge information in each edge information and obtaining the target edge information in each edge information. The non-straight edge information is the edge in the image that is not a straight line shape, and the target edge information is the straight line edge information. The adjacent edge information is searched in the straight line edge information. If a closed loop is finally formed, a quadrilateral is detected.

[0057] Optionally, for the obtained edge image, polygon analysis needs to be performed on it. First, use edge information structure analysis to find polygons, that is, determine the surrounding relationship of edge information in the edge image, that is, determine the outer edge information, inner edge information, and the nested relationship between the outer edge information and the inner edge information. The determined edge information all has a corresponding relationship with the original image. Therefore, the original image can be completely represented by the edge information. In the embodiment of the present application, a coding method is used to determine multiple polygon images, that is, encode the target edge information and assign different coding values to different edge information to facilitate confirming the hierarchical relationship of the polygons and the nested relationship between the obtained target edge information. Specifically: starting from the starting point, edit the pixels of the edge information, find the edge information points of the same type as the starting point. When scanning to the starting point again, the polygon closed loop is formed, switch to the next starting point and repeat this operation until all edge information points are traversed, and exclude polygon images with less than 4 sides to obtain multiple polygon images.

[0058] Optionally, in the embodiment of the present application, a polygon convex hull search algorithm is used to calculate the convex hull area of the convex hull in each polygon image and the polygon area of the polygon image itself. Optionally, scan according to the convex hull area and polygon area of each polygon image until the target identification image is scanned.

[0059] In the embodiment of the present application, AprilTag is used for identification image detection, so as to quickly and accurately determine whether there is an identification image in the collected image.

[0060] Optionally, scanning based on the convex hull area and polygon area of multiple said polygon images until the target identification image is scanned includes:

[0061] Eliminate the first target polygon images in multiple said polygon images whose polygon area is greater than the convex hull area to obtain second target polygon images;

[0062] Eliminate the third target polygon images in the second target polygon images whose ratio of the polygon area to the convex hull area is less than a preset value to obtain the final polygon images;

[0063] If a quadrilateral image is obtained after quadrilateral approximation of the final polygon image, it is determined that the target identification image is scanned.

[0064] Optionally, compare the polygon area and convex hull area of each polygon image to obtain a comparison result. Optionally, determine the polygon images with a polygon area greater than the convex hull area as the first target polygon images in multiple polygon images, and eliminate the first target polygon images from the polygon images to obtain the remaining second target polygon images.

[0065] Optionally, calculate the ratio of the polygon area to the convex hull area of each polygon image in the second target polygon image, and compare the ratio with a preset value to obtain a comparison result. The preset value is set according to the actual situation. In one embodiment, the preset value can be 0.8, 0.75, etc.

[0066] Optionally, determine the polygon images with ratios less than the preset value as the third target polygon images in the second target polygon images, and remove the third target polygon images from the second target polygon images to obtain the final polygon image.

[0067] Optionally, perform quadrilateral approximation on the final polygon image using the Douglas-Peucker algorithm. Therefore, if a quadrilateral image is obtained after performing quadrilateral approximation on the final polygon image, it is determined that the target identification image has been scanned. If a quadrilateral image is not obtained after performing quadrilateral approximation on the final polygon image, it is determined that the target identification image has not been scanned. In one embodiment, the process of quadrilateral approximation is as follows: Step 1: Find the two endpoints A and B of the curve, and draw the chord between the curves, that is, the line segment AB. Step 2: Find the point C on the curve that is farthest from the line segment AB, and find the distance d between them. Step 3: Compare the distance d with a preset threshold in advance. If it is less than the threshold, the line can be directly approximated as the curve, and this section of the curve is processed. Step 4: If the distance d is greater than the given threshold, use the point C to divide the curve into AC and BC, and perform steps 1 to 3 on each of the two segments of the line. When all the curves are processed, connect the broken lines formed by the segmentation points in sequence, so that the curve before processing can be approximated as a straight line, excluding polygons with a vertex number not equal to 4 in this case.

[0068] In the embodiment of the present application, the identification image is detected by the convex hull area and the polygon area of the polygon image, so that it is possible to quickly and accurately determine whether there is an identification image in the collected image.

[0069] Optionally, perform decoding processing on the target identification image to obtain target pose data corresponding to the target identification image, including:

[0070] Step 1021, obtain the target vertex coordinates of the target identification image;

[0071] Step 1022, determine the inlier coordinates of the dot matrix in the target identification image based on the target vertex coordinates, and match the original identification image based on the inlier coordinates;

[0072] Step 1023, determine the single linear transformation matrix of the target identification image based on the target vertex coordinates and the original vertex coordinates of the original identification image;

[0073] Step 1024: Decompose the single linear transformation matrix into a rotation transformation vector and a translation transformation vector to obtain the target pose data.

[0074] It should be noted that the encoding methods of the identification codes are divided into three types, and the lengths of their black-edge color blocks are 8, 7, and 6 color blocks respectively. For the decoded content, a point array needs to be generated within the detected quadrilateral to calculate the value of each color block, and then a simple classifier is constructed according to the local binary pattern to classify the color blocks within the quadrilateral. Encoding the positive example color blocks as 1 and the negative example color blocks as 0, the code of this identification can be obtained. After obtaining the code, it is matched with the codes in the known library to determine whether the decoded identification is correct.

[0075] The quadrilateral detected in the previous step does not necessarily meet the requirements, so it is necessary to encode, match, and check the quadrilateral obtained in the previous step. For the obtained quadrilateral, first, its internal dot matrix needs to be determined. The quadrilateral in the previous step is stored in the form of four vertices. Based on these four points, the specific coordinates of the dot matrix within the quadrilateral can be determined. The number of dot matrices is different for different encoding methods (8*8, 7*7, 6*6 according to the selected encoding method), and its encoding method is determined by the user when calibrating the object. Different encoding methods generate different internal point coordinates. After the above screening and verification, the ID of the observed code and its rotation can be determined at this point, and other parameters of this identification can be calculated, including the rotation orientation of the identification relative to the original identification and the similarity degree between the observed identification and the matching identification.

[0076] Therefore, the embodiments of this application can be understood as: obtaining the target vertex coordinates of the target identification image, and determining the internal point coordinates of the dot matrix in the target identification image according to the target vertex coordinates. Optionally, the original identification image is matched according to the internal point coordinates, and the single linear transformation matrix of the target identification image is determined according to the target vertex coordinates and the original vertex coordinates of the original identification image.

[0077] The single linear transformation matrix is a 3*3 matrix, which refers to the transformation of mapping a point in an image to the corresponding point in another image. For the first set of corresponding points in two images, the homography transformation is to solve a coordinate transformation between the corresponding points, and the single linear transformation matrix is mapped in the following way.

[0078]

[0079] Among them, (x 1 , y 1 , 1) is a point in the image, (x 2 , y 2 , 1) is the corresponding point mapped to another image, and H is the single linear transformation matrix.

[0080] It should be noted that under the pinhole camera model, the same image has correlations in different spaces. A linear transformation maps this image from one space to another corresponding space. The linear transformation matrix is this mapping matrix and can be decomposed into a rotation matrix and a translation matrix. According to the coordinates of the four vertices obtained in the previous step, the linear transformation matrix relative to the original coordinates can be calculated. By combining the information of multiple cameras with the linear transformation matrix, the relevant pose information of the identifier can be obtained. By decomposing this matrix, the rotation transformation vector rvec and the translation transformation vector tvec can be calculated. For the obtained rotation vector, the correct pose of the identifier can be obtained based on this vector. Therefore, it can be understood that by decomposing the rotation transformation vector and the translation transformation vector of the linear transformation matrix, the target pose data is obtained.

[0081] Therefore, the parsing process of the target pose data can be understood as: identifier image, identifier detection, identifier decoding, linear transformation, PnP conversion to world coordinates, identifier world coordinate pose.

[0082] In the embodiment of the present application, highly accurate target pose data is parsed from the identifier image, and then the point cloud data is updated based on the target pose data to obtain highly accurate point cloud data, so that highly accurate navigation map information can be determined through the highly accurate point cloud data.

[0083] Optionally, updating the second point cloud data based on the target pose data to obtain the target point cloud data at the second moment includes:

[0084] Step 10311: Obtain the decoding time information for decoding the target identifier image and the delay time information for receiving the target pose data;

[0085] Step 10312: Determine the target time information based on the second moment, the decoding time information, and the delay time information;

[0086] Step 10313: Align the time information of the target pose data and the time information of the second point cloud data with the target time information to obtain the aligned pose data and the aligned point cloud data;

[0087] Step 10314: Replace the aligned point cloud data with the aligned pose data to obtain the target point cloud data at the second moment.

[0088] It should be noted that the loop detection of Google's Cartographer enables the SLAM intelligent vehicle to generate more accurate navigation map information in large indoor scenarios. Therefore, the algorithm used in the embodiments of this application is the Cartographer algorithm. The message passing and interaction between the identification algorithm process and the Cartographer algorithm process are as follows: Cartographer reads Landmark data and passes it into Cartographer ROS through ROS, and calculates it in the Cartographer core algorithm through a real-time callback function from Cartographer ROS. The Cartographer algorithm is divided into two parts: the front end and the back end. The real-time positioning data that needs to be returned for positioning is obtained from the information fed back by the front end.

[0089] Therefore, the identification algorithm publishes identification data at a data frequency of 30hz, publishes the calculated identification data when the SLAM intelligent vehicle passes by the identification, and publishes Landmark messages at 30hz. When the SLAM intelligent vehicle does not pass by the identification, empty Landmark data is published. Cartographer ROS obtains Landmark data at 30hz and overrides the predicted pose at the Cartographer front end.

[0090] When ROS receives empty Landmark data, it does not process the message.

[0091] Optionally, refer to Figure 2 , Figure 2 which is a schematic diagram of the time information alignment process provided by the embodiments of this application. When Cartographer ROS obtains Landmark, it stores the data in the Landmark queue, subtracts the Apriltag processing identification algorithm time and the ROS publishing and receiving message time from the current time to obtain the time when the camera captures the image, and then aligns it according to the relationship between this time and the calibration time information of the radar and the camera, so as to obtain the pose of the identification obtained at the current radar time. The Landmark pose value obtained at this time overrides the pose of the vehicle at the front end, and then matching processing and subsequent Cartographer algorithms are performed.

[0092] Therefore, the embodiments of the present application can be understood as follows: obtaining the decoding time information for decoding the target identification image and the delay time information for receiving the target pose data. Further, subtracting the decoding time information and the delay time information from the second moment to determine the target time information, that is, target time information = second moment - decoding time information - delay time information. Further, aligning the time information of the target pose data and the time information of the second point cloud data with the target time information to obtain the aligned pose data and the aligned point cloud data. Optionally, replacing the aligned point cloud data with the aligned pose data to obtain the target point cloud data at the second moment.

[0093] The embodiments of the present application update the point cloud data through the target pose data to obtain high-precision point cloud data, so as to determine high-precision navigation map information through the high-precision point cloud data.

[0094] Optionally, determining the target navigation map information based on the local sub-map information at all moments includes:

[0095] Step 10331: Obtain the correlation information between the local sub-map information at different moments;

[0096] Step 10332: Determine the error term between the associated local sub-map information based on the similarity or difference degree between the correlation information;

[0097] Step 10333: Minimize the error term between the associated local sub-map information to obtain the optimized associated local sub-map information;

[0098] Step 10334: Determine the target navigation map information based on the optimized associated local sub-map information.

[0099] It should be noted that the loop detection of Google's Cartographer enables the SLAM intelligent vehicle to generate more accurate navigation map information in a large indoor scene. Therefore, the mapping algorithm used in the embodiments of the present application is the Cartographer algorithm.

[0100] Optionally, referring to Figure 3 , Figure 3It is a schematic diagram of the overall framework of Cartographer provided by an embodiment of the present application. Cartographer establishes local submap information submap based on the point cloud data laser scan obtained by lidar. When a local submap information submap is completed, no new laser scan will be inserted into this local submap information submap, and this local submap information submap will be added to the candidate queue for loop detection. When the current laser scan arrives, it will first be inserted into the corresponding local submap information submap in the optimal pose. Then, a window is drawn near the current laser scan, and the most suitable match of a certain scan in a previous local submap information submap is searched within the window. Once this match is found, it is added to the optimization problem of loop constraint for a rectification process of map building.

[0101] The Cartographer system combines local and global methods, and both methods optimize the pose. In the local method, a continuous scan is matched with a part of the navigation map information by using a non - linear optimization method, that is, it is matched with the local submap information submap, which is also scan matching. The accumulated error in this process will be eliminated during loop detection in global slam. The determination of the local submap information submap is to continuously calibrate the scan point set and the coordinates of the local submap information submap, that is, to transform the scan into the coordinate system of the local submap information submap. A certain number of laser scans are used to determine a local submap information submap, and the local submap information submap is represented by a probability grid, which can also be said to be grid - based navigation map information. A navigation map information is represented by discrete grid points, and the values of these grids can represent the probability that this grid is an obstacle. When a new laser scan is inserted into the local submap information submap, its grid probability value is updated according to a ratio.

[0102] Optionally, before inserting the scan into the local submap information submap, it is necessary to optimize the scan relative to the current local submap to make the probability of this scan in the local submap information subamap the largest.

[0103] Optionally, in the embodiments of the present application, loop detection is performed through the SPA optimization algorithm. SPA optimization is a pose graph optimization technology, and the object of its optimization is the world poses of all nodes. There are two types of nodes in Cartographe, namely key frame nodes and submap nodes, and a submap is a sub-navigation map information formed by splicing a continuous number of laser key frames together. These two types of nodes are both objects of optimization, and all constraints (edges) are determined between these two types of nodes, and no constraints are determined between nodes of the same type. The error term measures the error between the observed value and the "true value". A constraint uniquely determines an error term. Taking the i-th local submap information submap and the j-th key frame node node as an example, the constraint between them is the relative pose observation value, which is represented in the coordinate system of the local submap information submap, and the corresponding error term is expressed as:

[0104] e ij = z ij - h(pose i , pose j )

[0105] where e is a vector rather than a scalar, and usually the 2-norm of the vector is taken. e ij is the error term, pose i and pose j are the world poses of submap and node respectively, which are variables, and h(pose i , pose j ) represents the coordinates of the j-th node in the i-th submap, z ij is a constant parameter, and h(·) is the coordinate transformation relationship.

[0106] Considering the weights of different elements in the vector, this weight can be reflected by the information matrix. Therefore, the magnitude of the error is expressed as

[0107] In one embodiment, there are k error term optimization objectives to minimize the overall error of all error terms.

[0108] According to the classical L-M optimization method, for the above-formulated problem, the x increment is calculated. When solving the increment equation, the matrix represented by the left parenthesis is usually not directly inverted. One of the common methods is to perform its triangular decomposition and use the characteristics of the triangular matrix to solve quickly. Since there are no constraints between all nodes, the Hessian matrix H usually has obvious sparsity. The Hessian Matrix itself represents a square matrix composed of the second-order partial derivatives of a multivariate function. In L-M, the square of the first derivative (Jacobian matrix) is used to approximately represent the Hessian matrix. The error term is only related to the corresponding two nodes:

[0109] Hessian matrix

[0110] submap1 ... submap i ... submap N node1 ... node j ... node N <![CDATA[J k > 0 <![CDATA[J i > 0 0 <![CDATA[J j > 0

[0111] The Hessian matrix is the sum of all error terms, as follows Figure 4 , Figure 4 is the error pose schematic diagram provided by the embodiments of the present application.

[0112] Since H has good sparsity, the Cholesky decomposition can be performed on the sparse matrix, which decomposes a symmetric positive definite matrix into the product of an upper triangular matrix U and its transpose.

[0113] Therefore, it can be understood that the loop detection is used to establish the correlation information between the local sub-graph information at different times, that is, to obtain the correlation information between the local sub-graph information at different times.

[0114] Furthermore, determine the similarity degree or difference degree between the correlation information, and determine the error terms between the correlated local sub-graph information according to the similarity degree or difference degree between the correlation information. Specifically: use the similarity measurement method to calculate the similarity or difference degree between the correlated local sub-graph information. Common similarity measurement methods include Euclidean distance, cosine similarity, structural similarity index (SSIM), etc. The calculated similarity or difference degree is used to construct an error term to measure the error between the local sub-graph information. The specific form of the error term can be adjusted according to different tasks. For example, the Euclidean distance can be used as the error term, or a custom loss function can be defined.

[0115] Optionally, minimize all the error terms between the correlated local sub-graph information to obtain the optimized correlated local sub-graph information. Optionally, determine the target navigation map information according to the optimized correlated local sub-graph information.

[0116] It should be noted that map construction parameters need to be optimized during the map construction process, specifically: optimizing the local SLAM parameters and global SLAM parameters of Cartographer. The adjustment of local SLAM parameters mainly targets the band-pass filter parameters, motion filter, voxel filter parameters, and scan matcher parameters. In global SLAM, corresponding weights are mainly set according to the source of residuals. The band-pass filter parameters determine that the point cloud data between min_range and max_range is regarded as valid data. Reducing max_range can reduce the computational load in positioning without affecting the map construction effect. The motion filter affects the number of nodes in the submap. The radar scan data is filtered by setting a threshold. The smaller the threshold, the more nodes are generated when moving the same distance or turning the same angle, and the smaller the submap established. Although a small threshold can make the navigation map information more refined, when positioning, the robot needs to establish constraints with all submaps in the navigation map information, and the computational load will increase significantly. The voxel filter downsamples the point cloud data, reducing the data volume while maintaining the general shape of the point cloud. As the parameter increases, the point cloud data gradually becomes sparse, which may cause some features in the navigation map information to disappear. The scan matcher inserts the point cloud data obtained after lidar scanning into the position with the highest probability in the navigation map information. The difference between the Ceres Scan Matcher and the Real Time Correlative Scan Matcher is that the latter can generate stable navigation map information without relying on sensors in an environment with rich feature navigation map information, but it requires relatively high corresponding computing resources. Subsequently, the corresponding strategy is adjusted according to the specific hardware resources. In addition, according to the degree of trust in the residuals of each part, the weights of relevant terms in the cost equation can be appropriately adjusted to provide guidance for the optimizer during submap optimization.

[0117] Furthermore, laser point cloud filtering needs to be performed. To reduce the impact of clutter generated by noise and other reasons in the navigation map information on positioning, multi-channel median filtering is added during the map construction process. By adjusting the minimum and maximum vertical angles, the number of consecutive measurements of the internal angle, and the removal of outliers, the map construction effect is affected. After adding the filtering, it can be clearly seen that the clutter in the navigation map information is reduced, and the obstacle contour is clearer.

[0118] Optionally, after determining the target navigation map information based on the local submap information at all times, it further includes:

[0119] Receiving the running path sent by the dispatching terminal;

[0120] If a marker image is scanned during the process of running according to the running path, obtaining the current position information based on the marker image and displaying the current position information in the target navigation map information; or,

[0121] If no identification image is scanned during the operation along the described operation path, the current position information is obtained based on the odometer data and displayed in the target navigation map information.

[0122] Optionally, receive the operation path sent by the dispatching terminal. Further, the SLAM intelligent vehicle operates in the target navigation map information according to the operation path. If an identification image is scanned during the operation along the operation path, the current position information is obtained according to the identification image and displayed in the target navigation map information. If no identification image is scanned during the operation along the operation path, the current position information is obtained based on the odometer data and displayed in the target navigation map information.

[0123] Optionally, refer to Figure 5 , Figure 5 is the dispatching schematic diagram provided by the embodiment of the present application. In a factory scenario, the dispatching system plans the operation path of the vehicle through the path planning module according to the determined factory scenario navigation map information, the real-time position of the vehicle, and the requirements of the service task, and sends a computer program to the control module to control the vehicle to move along the planned trajectory in the environment. The SLAM intelligent vehicle starts the real-time positioning algorithm according to the sensor data obtained during the operation to realize the real-time and accurate estimation of the current position of the vehicle.

[0124] It should be noted that for the optimization of positioning function parameters: These parameters also adjust the local slam and global slam parameters at the same time. In addition to parameters such as the motion filter and the scan matcher, they also include the downsampling rate and the global sampling rate constrained in the navigation map information. The point cloud data obtained through the motion filter is matched and optimized with the existing navigation map information and the generated submaps using the scan matcher to obtain the pose of the vehicle in the navigation map information. The purpose of constrained downsampling is to accelerate the optimization rate, but too large downsampling will lead to the lack of constraints and cause positioning failure, and too small downsampling will lead to a slow running speed and loss of real-time performance. Reducing the global sampling rate can also improve the speed of cartographer positioning.

[0125] The embodiment of the present application can quickly and accurately dispatch through the determined navigation map information, and quickly and accurately locate the current position information.

[0126] Optionally, refer to Figure 6 , Figure 6 is the overall scheme flow schematic diagram provided by the embodiment of the present application. The identification image data obtains the vehicle pose through the identification pose estimation algorithm, and is published through the Landmark message to perform message passing and interaction between the identification algorithm process and the Cartographer algorithm process.

[0127] Cartographer reads Landmark data and passes it into Cartographer ROS through ROS, and calculates it in the Cartographer core algorithm through real-time callback functions. The Cartographer algorithm is divided into two parts: the front end and the back end. The real-time positioning data to be returned for positioning is obtained from the information fed back by the front end. In the front-end part of Cartographer, the predicted relative pose of the trolley is obtained through the kinematic model of odometry data and the pose interpolation of the previous frame. Here, the obtained pose is read into the pose information through Landmark data, and the corresponding pose replacement is performed, so as to feedback the real-time pose information. It is necessary to align the landmark world coordinates with the camera and then convert them into the coordinate system of the trolley for replacement. In addition, the pose of the matching process, the sub-navigation map information of the Cartographer back end, and the key point constraints can also be correspondingly replaced with the pose information obtained by the landmark.

[0128] The data processing device for intelligent navigation provided by the embodiments of the present application will be described below. The data processing device for intelligent navigation described below can be correspondingly referred to the data processing method for intelligent navigation described above. Refer to Figure 7 as shown, Figure 7 is a schematic structural diagram of the data processing device for intelligent navigation provided by the embodiments of the present application. The data processing device for intelligent navigation may include:

[0129] An acquisition module 701, configured to acquire first point cloud data at a first moment and second point cloud data at a second moment;

[0130] A decoding module 702, configured to decode the target identification image if the target identification image is scanned at the second moment, so as to obtain target pose data corresponding to the target identification image;

[0131] A navigation map information determination module 703, configured to process the first point cloud data and the second point cloud data based on the target pose data, and determine target navigation map information.

[0132] In the embodiments of the present application, high-precision target pose data is parsed from the identification image, and then the point cloud data is updated through the target pose data to obtain high-precision point cloud data, so that high-precision navigation map information is determined through the high-precision point cloud data.

[0133] In an optional example, the navigation map information determination module 703 is further configured to:

[0134] Update the second point cloud data based on the target pose data to obtain the target point cloud data at the second moment;

[0135] Perform a matching process based on the target point cloud data and the first point cloud data to determine the local sub-map information at the second moment;

[0136] Determine the target navigation map information based on the local sub-map information at all moments.

[0137] In an optional example, the navigation map information determination module 703 is further configured to:

[0138] Obtain the decoding time information for decoding the target identification image and the delay time information for receiving the target pose data;

[0139] Determine the target time information based on the second moment, the decoding time information, and the delay time information;

[0140] Align the time information of the target pose data and the time information of the second point cloud data with the target time information to obtain the aligned pose data and the aligned point cloud data;

[0141] Replace the aligned point cloud data with the aligned pose data to obtain the target point cloud data at the second moment.

[0142] In an optional example, the navigation map information determination module 703 is further configured to:

[0143] Obtain the correlation information between the local sub-map information at different moments;

[0144] Determine the error term between the associated local sub-map information based on the similarity or difference degree between the correlation information;

[0145] Minimize the error term between the associated local sub-map information to obtain the optimized associated local sub-map information;

[0146] Determine the target navigation map information based on the optimized associated local sub-map information.

[0147] In an optional example, the data processing device for intelligent navigation is further configured to:

[0148] Perform edge detection on the collected image according to the gradient to obtain the gradient value of each pixel point in the image, and determine each edge information in the image according to the gradient value of each pixel point;

[0149] Eliminate the non-linear edge information in each edge information to obtain the target edge information in each edge information;

[0150] Encode the target edge information, and determine a plurality of polygon images based on the nesting relationship between the encoded target edge information.

[0151] Scan based on the convex hull area and polygon area of the plurality of polygon images until a target identification image is scanned.

[0152] In an optional example, the data processing device for intelligent navigation is further configured to:

[0153] Eliminate a first target polygon image in the plurality of polygon images whose polygon area is greater than the convex hull area, and obtain a second target polygon image.

[0154] Eliminate a third target polygon image in the second target polygon images whose ratio of the polygon area to the convex hull area is less than a preset value, and obtain a final polygon image.

[0155] If a quadrilateral image is obtained after quadrilateral approximation of the final polygon image, it is determined that the target identification image is scanned.

[0156] In an optional example, the decoding module 702 is further configured to:

[0157] Obtain the target vertex coordinates of the target identification image.

[0158] Determine the inlier coordinates of the dot matrix in the target identification image based on the target vertex coordinates, and match the original identification image based on the inlier coordinates.

[0159] Determine the single linear transformation matrix of the target identification image based on the target vertex coordinates and the original vertex coordinates of the original identification image.

[0160] Perform rotation transformation vector decomposition and translation transformation vector decomposition on the single linear transformation matrix to obtain the target pose data.

[0161] In an optional example, the data processing device for intelligent navigation is further configured to:

[0162] Receive the running path sent by the dispatching terminal.

[0163] If an identification image is scanned during the running according to the running path, obtain the current position information based on the identification image, and display the current position information in the target navigation map information; or,

[0164] If no identification image is scanned during the running according to the running path, obtain the current position information based on the odometer data, and display the current position information in the target navigation map information.

[0165] The specific embodiments of the data processing device for intelligent navigation provided in this application are basically the same as those of the data processing method for intelligent navigation, and will not be elaborated here.

[0166] Optionally, as Figure 8 shown, Figure 8 is a schematic structural diagram of an electronic device provided in an embodiment of the present application. The electronic device may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communication interface 820, and the memory 830 complete communication with each other through the communication bus 840. The processor 810 may call a computer program in the memory 830 to execute the steps of the data processing method for intelligent navigation, for example, including:

[0167] Obtain the first point cloud data at the first moment and the second point cloud data at the second moment;

[0168] If a target identification image is scanned at the second moment, decode the target identification image to obtain target pose data corresponding to the target identification image;

[0169] Process the first point cloud data and the second point cloud data based on the target pose data to determine target navigation map information.

[0170] In addition, when the logical computer program in the above-mentioned memory 830 can be implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several computer programs for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present application. The foregoing storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk, or an optical disc that can store program codes.

[0171] On the other hand, an embodiment of the present application further provides a non-transitory computer-readable storage medium. The non-transitory computer-readable storage medium includes a computer program. The computer program can be stored on the non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the steps of the data processing method for intelligent navigation provided in the above embodiments, for example, including:

[0172] Obtain the first point cloud data at the first moment and the second point cloud data at the second moment;

[0173] If a target identification image is scanned at the second moment, perform decoding processing on the target identification image to obtain the target pose data corresponding to the target identification image;

[0174] Process the first point cloud data and the second point cloud data based on the target pose data to determine the target navigation map information.

[0175] In another aspect, an embodiment of the present application further provides a computer product. The computer product includes a computer program. The computer program can be stored on the computer product. When the computer program is executed by a processor, the computer can execute the steps of the data processing method for intelligent navigation provided in the above embodiments, for example, including:

[0176] Obtain the first point cloud data at the first moment and the second point cloud data at the second moment;

[0177] If a target identification image is scanned at the second moment, perform decoding processing on the target identification image to obtain the target pose data corresponding to the target identification image;

[0178] Process the first point cloud data and the second point cloud data based on the target pose data to determine the target navigation map information.

[0179] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0180] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the essence of the above technical solution, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several computer programs to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0181] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of each embodiment of the present application.

Claims

1. A data processing method for intelligent navigation, characterized in that, it includes: Obtain the first point cloud data at the first moment and the second point cloud data at the second moment; If a target identification image is scanned at the second moment, perform decoding processing on the target identification image to obtain the target pose data corresponding to the target identification image; Based on the target pose data, process the first point cloud data and the second point cloud data to determine the target navigation map information.

2. The data processing method for intelligent navigation according to claim 1, characterized in that, The processing the first point cloud data and the second point cloud data based on the target pose data to determine the target navigation map information includes: Based on the target pose data, perform update processing on the second point cloud data to obtain the target point cloud data at the second moment; Based on the target point cloud data and the first point cloud data, perform matching processing to determine the local sub-map information at the second moment; Based on the local sub-map information at all moments, determine the target navigation map information.

3. The data processing method for intelligent navigation according to claim 2, characterized in that, The processing the second point cloud data based on the target pose data to obtain the target point cloud data at the second moment includes: Obtain the decoding time information for decoding the target identification image and the delay time information for receiving the target pose data; Based on the second moment, the decoding time information, and the delay time information, determine the target time information; Align the time information of the target pose data and the time information of the second point cloud data with the target time information to obtain the aligned pose data and the aligned point cloud data; Replace the aligned point cloud data with the aligned pose data to obtain the target point cloud data at the second moment.

4. The data processing method for intelligent navigation according to claim 2, characterized in that, The determining the target navigation map information based on the local sub-map information at all moments includes: Obtain the correlation information between the local sub-map information at different moments; Based on the similarity or difference degree between the correlation information, determine the error term between the correlated local sub-map information; Minimize the error term between the correlated local sub-map information to obtain the optimized correlated local sub-map information; Based on the optimized correlated local sub-map information, determine the target navigation map information.

5. The data processing method for intelligent navigation according to claim 1, characterized in that, Before the step of if a target identification image is scanned at the second moment, perform decoding processing on the target identification image to obtain the target pose data corresponding to the target identification image, it further includes: Perform edge detection on the collected image according to the gradient to obtain the gradient value of each pixel point in the image, and determine each edge information in the image according to the gradient value of each pixel point; Eliminate the non-linear edge information in each edge information to obtain the target edge information in each edge information; Encode the target edge information, and determine a plurality of polygon images based on the nesting relationship between the encoded target edge information. Scan based on the convex hull area and polygon area of the plurality of polygon images until a target identification image is scanned.

6. The data processing method for intelligent navigation according to claim 5, wherein, the scanning based on the convex hull area and polygon area of the plurality of polygon images until a target identification image is scanned includes: Eliminate the first target polygon images in the plurality of polygon images whose polygon area is greater than the convex hull area to obtain second target polygon images; Eliminate the third target polygon images in the second target polygon images whose ratio of the polygon area to the convex hull area is less than a preset value to obtain the final polygon images; If a quadrilateral image is obtained after quadrilateral approximation of the final polygon images, it is determined that the target identification image is scanned.

7. The data processing method for intelligent navigation according to any one of claims 1 to 6, wherein, the decoding process of the target identification image to obtain the target pose data corresponding to the target identification image includes: Obtain the target vertex coordinates of the target identification image; Determine the inlier coordinates of the dot matrix in the target identification image based on the target vertex coordinates, and match the original identification image based on the inlier coordinates; Determine the single linear transformation matrix of the target identification image based on the target vertex coordinates and the original vertex coordinates of the original identification image; Perform rotation transformation vector decomposition and translation transformation vector decomposition on the single linear transformation matrix to obtain the target pose data.

8. A data processing device for intelligent navigation, wherein, it includes: An acquisition module for acquiring the first point cloud data at the first moment and the second point cloud data at the second moment; A decoding module for decoding the target identification image to obtain the target pose data corresponding to the target identification image if the target identification image is scanned at the second moment; A navigation map information determination module for processing the first point cloud data and the second point cloud data based on the target pose data to determine the target navigation map information; Preferably, the navigation map information determination module processes the first point cloud data and the second point cloud data based on the target pose data to determine the target navigation map information, including: Performing update processing on the second point cloud data based on the target pose data to obtain the target point cloud data at the second moment; Performing matching processing on the target point cloud data and the first point cloud data to determine the local sub-map information at the second moment; Determining the target navigation map information based on the local sub-map information at all moments; Preferably, the navigation map information determination module performs update processing on the second point cloud data based on the target pose data to obtain the target point cloud data at the second moment, including: Obtaining the decoding time information for decoding the target identification image and the delay time information for receiving the target pose data; Determine target time information based on the second moment, the decoding time information, and the delay time information; Align the time information of the target pose data and the time information of the second point cloud data with the target time information to obtain the aligned pose data and the aligned point cloud data; Replace the aligned point cloud data with the aligned pose data to obtain the target point cloud data at the second moment; Preferably, the navigation map information determination module determines the target navigation map information based on the local sub-map information at all moments, including: Obtain the association information between the local sub-map information at different moments; Determine the error term between the associated local sub-map information based on the similarity or difference degree between the association information; Minimize the error term between the associated local sub-map information to obtain the optimized associated local sub-map information; Determine the target navigation map information based on the optimized associated local sub-map information; Preferably, before the decoding module decodes the target identification image to obtain the target pose data corresponding to the target identification image if the target identification image is scanned at the second moment, further include: Perform edge detection on the collected image according to the gradient to obtain the gradient value of each pixel point in the image, and determine each edge information in the image according to the gradient value of each pixel point; Eliminate the non-linear edge information in each edge information to obtain the target edge information in each edge information; Encode the target edge information, and determine a plurality of polygon images based on the nested relationship between the encoded target edge information; Scan based on the convex hull area and the polygon area of the plurality of polygon images until the target identification image is scanned; Preferably, the decoding module scans based on the convex hull area and the polygon area of the plurality of polygon images until the target identification image is scanned, including: Eliminate the first target polygon images in the plurality of polygon images whose polygon area is greater than the convex hull area to obtain the second target polygon images; Eliminate the third target polygon images in the second target polygon images whose ratio of the polygon area to the convex hull area is less than a preset value to obtain the final polygon images; If a quadrilateral image is obtained after quadrilateral approximation of the final polygon image, it is determined that the target identification image is scanned; Preferably, the decoding module decodes the target identification image to obtain the target pose data corresponding to the target identification image, including: Obtain the target vertex coordinates of the target identification image; Determine the inlier coordinates of the dot matrix in the target identification image based on the target vertex coordinates, and match the original identification image based on the inlier coordinates; Determine the single linear transformation matrix of the target identification image based on the target vertex coordinates and the original vertex coordinates of the original identification image; Perform rotation transformation vector decomposition and translation transformation vector decomposition on the single linear transformation matrix to obtain the target pose data.

9. An electronic device, Characterized in that, It includes a processor and a memory, and the memory stores multiple computer programs; the processor loads the computer programs from the memory to execute the data processing method for intelligent navigation according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores multiple computer programs, and the computer programs are suitable for being loaded by a processor to execute the data processing method for intelligent navigation according to any one of claims 1 to 7.

Images