Pedestrian spatial position estimation method and device based on weighted fusion and robot

Abstract

The invention discloses a pedestrian spatial position estimation method and device based on weighted fusion and a robot, and the method comprises the following steps: collecting a real-time scene image, carrying out the box marking of a target pedestrian, and generating a box coordinate; acquiring image coordinates respectively corresponding to at least two body parts of a target pedestrian in the real-time scene image; converting the image coordinates corresponding to the at least two body parts into space coordinates under a robot coordinate system; and fusing the at least two space coordinates by adopting a secondary weighted fusion strategy to generate an accurate position of the target pedestrian relative to the robot. According to the invention, secondary weighted fusion is carried out by using a plurality of coordinate points of the target pedestrian, such as two coordinate points of the top and the bottom of the head of the target pedestrian, so that error estimation caused by limitation of a target detector is reduced, errors of pedestrian position estimation are reduced, and the accuracy and the stability of pedestrian position estimation are improved.

Description

technical field [0001] The invention relates to the field of robots, in particular to a weighted fusion-based pedestrian spatial position estimation method, device and robot. Background technique [0002] With the development of artificial intelligence technology, the target detection algorithm based on deep learning has been popularized and implemented in the industrial field. At this stage, target detection algorithms such as FasterRCNN, YOLO, SSD, etc. calculate the pictures transmitted by the camera and mark them in the form of boxes on the pictures. For mobile robots, the spatial information of pedestrians has a great influence on the robot's mobile decision-making, but the information marked on the picture cannot be used directly. It is necessary to transform the marked information of the picture into the spatial position, that is, the conversion of pedestrians corresponding to the marked position on the picture is the position of the pedestrian relative to the robot ...

AI Technical Summary

Problems solved by technology

In practical application scenarios, pedestrians may be blocked, pedestrians nearby may not be able to see their complete legs, and there are certain errors in the target detection results, such as Figure 4a and Figure 4b As shown, the strategy of estimating the spatial position of pedestrians by simply using the bottom coordinates of pedestrians marked in the picture as a reference point is often less stable

Images