A target tracking and location robot system based on omnistereo vision

Abstract

Purpose Because of their large field of view, omnistereo vision systems have been widely used as primary vision sensors in autonomous mobile robot tasks. The purpose of this article is to achieve real-time and accurate tracking by the omnidirectional vision robot system. Design/methodology/approach The authors provide in this study the key techniques required to obtain an accurate omnistereo target tracking and location robot system, including stereo rectification and target tracking in complex environment. A simple rectification model is proposed, and a local image processing method is used to reduce the computation time in the localization process. A target tracking method is improved to make it suitable for omnidirectional vision system. Using the proposed methods and some existing methods, an omnistereo target tracking and location system is established. Findings The experiments are conducted with all the necessary stages involved in obtaining a high-performance omnistereo vision system. The proposed correction algorithm can process the image in real time. The experimental results of the improved tracking algorithm are better than the original algorithm. The statistical analysis of the experimental results demonstrates the effectiveness of the system. Originality/value A simple rectification model is proposed, and a local image processing method is used to reduce the computation time in the localization process. A target tracking method is improved to make it suitable for omnidirectional vision system. Using the proposed methods and some existing methods, an omnistereo target tracking and location system is established.

Motion Planning of Autonomous Mobile Robot Using Recurrent Fuzzy Neural Network Trained by Extended Kalman Filter

“…So, the vision system based on the catadioptric omnidirectional camera is carried on the mobile robot to percept surroundings in this paper. Detailed imaging theory is introduced in our previous work [29, 30].…”

Automatic Rectification of the Hybrid Stereo Vision System

“…The relation of the essential matrix and the fundamental matrix is: where and are the intrinsic parameters of the two images. The decomposition of an essential matrix is: where and differ by a scale factor which can be calculated using two 3D points offline [ 35 ]. According to (9) and (10), (8) can be written as: …”

Target Tracking in Complex Scenes Based on Computer Vision

“…1 By detecting and identifying video content, people can quickly access key information in video content and use this key information to help people solve many of the problems they face today. 2 The main research work is to use the deep learning network model to identify the key content in the video by identifying the content in the video. It mainly introduces from the aspects of video data preprocessing, data augmentation, feature extraction, feature aggregation and multimodal feature fusion.…”