Vision-based robot following method and following robot

Abstract

The invention provides a vision-based robot following method and a following robot, and relates to the technical field of computer vision. The vision-based robot following method comprises the following steps: receiving target object information; using a visual camera to acquire a visual field input image, preprocessing the visual field input image to generate a detection input image, and inputingthe detection input image into a pedestrian detection neural network model for detection, wherein the detection input image is formed by splicing a plurality of images which have different resolutions and are related to the visual field input image; obtaining a pedestrian detection result, and determining a following target in the pedestrian according to the target object information; and carrying out robot following on the following target. On the basis of ensuring real-time accuracy of far and near target tracking, a requirement of target tracking on the computing power of equipment is reduced, and the power consumption is decreased.

Description

technical field [0001] The invention relates to the technical field of computer vision. Background technique [0002] Intelligent mobile robots following moving targets are widely used in the fields of home services, helping the elderly and the disabled, scene monitoring and smart cars, etc., and have broad application prospects. The tracking of target objects by mobile robots involves the fields of computer vision, motion control, and pattern recognition. Among them, for robot vision, the purpose is to simulate the human visual mechanism, calculate the importance of information in the visual scene, and extract the salient features or target object features of interest in the image. [0003] The following process of a vision-based following robot usually includes image acquisition, object detection, and object tracking. With the rapid development of artificial intelligence and deep learning technology, object detection methods based on convolutional neural network (CNN) al...

AI Technical Summary

Problems solved by technology

However, although the convolutional neural network algorithm improves the real-time accuracy of tracking, because the convolutional neural network-based target detection algorithm usually contains a large number of computationally intensive operations, the requirements for real-time detection computing power and bandwidth are very high.

Images