Robot, method, device and readable storage medium for creating map points

Abstract

A method for creating a map point comprises: traversing key points of a dominant frame in a pre-processed image frame; matching the key points in a reference area of a current update frame with the key points of the dominant frame, wherein the reference area of the current update frame is an area with the same geometric attribute as a preset range to which the key points of the dominant frame belong; if the key points in the reference area of the current update frame are successfully matched with the key points of the dominant frame, then continue to match the key points in the reference area of an update frame transmitted after the current update frame with the key points of the dominant frame; combining the dominant frame and the update frame transmitted after the current update frame and in which the key points are successfully matched with the key points of the dominant frame for the last time, and calculating the depth value of the key points of the dominant frame according to a triangulation measurement method. The method can ensure the baseline length, the calculation accuracy of the depth value, the subsequent mapping accuracy, and the data association accuracy, thereby ensuring that the mapping reaches the expected accuracy.

Claims

1. A robot, characterized in that, The robot includes: Memory and processor; The memory stores executable program code; The processor coupled to the memory calls executable program code stored in the memory to execute a method for creating map points, the method comprising: Traverse the key points of the dominant frame in the preprocessed image frames; Using the preset range to which the key points of the dominant frame belong as a reference region, the key points within the reference region of the current update frame are matched with the key points of the dominant frame. The reference region of the current update frame is a region with the same geometric attributes as the preset range. Wherein, the same geometric attributes mean that the shape, size, and orientation of the reference region of the current update frame are the same as the shape, size, and orientation of the reference region of the dominant frame, respectively, and the position of the reference region of the current update frame in the current update frame corresponds to the position of the reference region of the dominant frame in the dominant frame. If a key point in the reference region of the current updated frame matches a key point in the dominant frame, then the key points in the reference region of the next updated frame are matched with the key points in the dominant frame. A successful match means that the similarity between the descriptor of the key point in the reference region of the current updated frame and the descriptor of the key point in the dominant frame is greater than a preset threshold. The depth value of the key point of the dominant frame is calculated by combining the dominant frame and the last update frame that successfully matches the key point of the dominant frame in the update frames passed after the current update frame, and according to the triangulation measurement method.

2. The robot as described in claim 1, characterized in that, The step of the processor calling executable program code stored in the memory to match key points within the reference region of the currently updated frame with key points of the dominant frame includes: Obtain the descriptor D1 of the key points within the reference region of the current updated frame and the descriptor D2 of the key points of the dominant frame, respectively; Calculate the similarity between the descriptor D1 and the descriptor D2; If the similarity between descriptor D1 and descriptor D2 is greater than a preset threshold, then it is determined that the key points in the reference region of the current updated frame are successfully matched with the key points of the dominant frame.

3. The robot as described in claim 1, characterized in that, After the processor calls the executable program code stored in the memory to match the key points within the reference region of the current updated frame with the key points of the dominant frame, the process further includes: If the key points in the reference region of the current updated frame fail to match the key points of the dominant frame, the range of the reference region of the current updated frame is expanded to obtain the expanded reference region of the current updated frame. Match the key points within the expanded reference region of the current updated frame with the key points of the dominant frame.

4. The robot as described in claim 1, characterized in that, After the processor calls the executable program code stored in the memory to match the key points within the reference region of the current updated frame with the key points of the dominant frame, the process further includes: If the key points in the reference region of the current updated frame fail to match the key points of the dominant frame, then the key points in the reference region of the adjacent frame are matched with the key points of the dominant frame. The adjacent frame is the updated frame immediately following the current updated frame. If the key points in the reference region of the adjacent frame fail to match the key points of the dominant frame, the range of the reference region of the current updated frame is expanded to obtain the expanded reference region of the current updated frame. Match the key points within the expanded reference region of the current updated frame with the key points of the dominant frame.

5. The robot as described in claim 1, characterized in that, After the processor calls the executable program code stored in the memory to match the key points within the reference region of the current updated frame with the key points of the dominant frame, the process further includes: If the key points in the reference region of the current updated frame fail to match the key points of the dominant frame, then the key points in the reference region of the adjacent frame are matched with the key points of the dominant frame. The adjacent frame is the updated frame immediately following the current updated frame. If the key points in the reference region of the adjacent frame fail to match the key points of the dominant frame, the range of the reference region of the adjacent frame is expanded to obtain the expanded reference region of the adjacent frame. Match key points within the expanded reference region of the adjacent frame with key points of the dominant frame.

6. The robot as described in any one of claims 1 to 5, characterized in that, In the method for creating map points executed by the processor calling the executable program code stored in the memory, the preset range of the key point of the dominant frame is a rectangular window centered on the key point of the dominant frame.

7. The robot as described in any one of claims 1 to 5, characterized in that, The processor calls the executable program code stored in the memory, executes the last update frame that successfully matched the keypoints of the dominant frame among the update frames passed after the current update frame and the dominant frame, and calculates the depth value of the keypoints of the dominant frame according to the triangulation measurement method. Following this step, the processor further includes: Based on the depth value and two-dimensional pixel coordinates of the key points in the dominant frame, calculate the three-dimensional coordinates of the map points corresponding to the key points in the dominant frame in the world coordinate system.

8. An apparatus for creating map points, characterized in that, The device includes: The traversal module is used to traverse the key points of the dominant frame in the preprocessed image frame; The first matching module is used to match key points in the reference area of ​​the current update frame with key points in the dominant frame, using a preset range to which the key points of the dominant frame belong as a reference area. The reference area of ​​the current update frame is a region with the same geometric properties as the preset range. The same geometric properties mean that the shape, size, and orientation of the reference area of ​​the current update frame are the same as the shape, size, and orientation of the reference area of ​​the dominant frame, respectively. The position of the reference area of ​​the current update frame in the current update frame corresponds to the position of the reference area of ​​the dominant frame in the dominant frame. The second matching module is used to match key points in the reference region of the current updated frame with key points in the dominant frame if the key points in the reference region of the current updated frame match successfully with the key points in the dominant frame. The matching success means that the similarity between the descriptor of the key points in the reference region of the current updated frame and the descriptor of the key points in the dominant frame is greater than a preset threshold. The first calculation module is used to combine the dominant frame and the last update frame that successfully matched the key points of the dominant frame in the update frames passed after the current update frame, and calculate the depth value of the key points of the dominant frame according to the triangulation measurement method.

9. A method for creating map points, characterized in that, The method includes: Traverse the key points of the dominant frame in the preprocessed image frames; Using the preset range to which the key points of the dominant frame belong as a reference region, the key points within the reference region of the current update frame are matched with the key points of the dominant frame. The reference region of the current update frame is a region with the same geometric attributes as the preset range. Wherein, the same geometric attributes mean that the shape, size, and orientation of the reference region of the current update frame are the same as the shape, size, and orientation of the reference region of the dominant frame, respectively, and the position of the reference region of the current update frame in the current update frame corresponds to the position of the reference region of the dominant frame in the dominant frame. If a key point in the reference region of the current updated frame matches a key point in the dominant frame, then the key points in the reference region of the next updated frame are matched with the key points in the dominant frame. A successful match means that the similarity between the descriptor of the key point in the reference region of the current updated frame and the descriptor of the key point in the dominant frame is greater than a preset threshold. The depth value of the key point of the dominant frame is calculated by combining the dominant frame and the last update frame that successfully matches the key point of the dominant frame in the update frames passed after the current update frame, and according to the triangulation measurement method.

10. A readable storage medium having a computer program stored thereon, characterized in that, The computer program is used to implement a method for creating map points when executed by a processor, wherein the method for creating map points is the method for creating map points implemented by the robot according to any one of claims 1 to 7.

Images