Research On Filtering Visual Tracking Algorithms And Application In Intelligent Robots

Intelligent Robots combine the characteristics of the interdisciplinary integration technologies such as mechanism,material science,computer image science,intelligent control,instrumentation,pattern recognition,etc.,which makes intelligent robot has always been a research hotspot in the field of science and technology.This paper takes industrial application environment as the research background.The intelligent robot based on visual perception needs to face the interference of various factors in complex industrial environment.It is divided into three levels according to the frequency of interference factors in industrial application.The interference factors of each level put forward the requirements of accuracy or robustness for the visual tracking algorithm of intelligent robot.Therefore,the ability and efficiency of intelligent robots to cope with complex environmental challenges have become an important index to evaluate the performance of intelligent robots,which also puts higher requirements on the robustness,accuracy and real-time performance of the visual tracking algorithms.In order to improve the stability,accuracy and speed of the target tracking algorithm in complex industrial environment,this paper studies the intelligent robot vision tracking algorithm in depth.The main research contents of this paper are as follows:1.Aiming at the complex and computational complexity of target tracking algorithm,a correlation filtering tracking algorithm based on spatiotemporal context information is studied.The basic theory of spatio-temporal context information and correlation filtering method is studied,and a fast target tracking method is constructed.To solve the problem of large scale change of target,a multi-scale estimation correlation filtering algorithm based on feature pyramid is introduced on spatio-temporal context information tracking algorithm to achieve accurate target scale estimation and improve the adaptability of the tracking algorithm to the first level interference factors such as illumination change,scale change and complex background in industrial environment.2.The target tracking algorithm based on interference discrimination and position prediction is studied.In order to improve the accuracy of target feature description in order to cope with the influence of the second level interference factors,such as the interference of similar objects and fast motion in industrial environment,a method of target appearance model modeling based on the fusion of Histogram of Oriented Gradient(HOG)features and spatiotemporal context information is proposed.On the basis of making full use of the background information of the target and its surroundings,the method of describing the target features based on gradient histogram is integrated to enhance the accuracy of the target model construction.Aiming at the problem of target interference and partial occlusion caused by similar objects,a target interference information discrimination method based on average peak correlation energy is designed,and the target template update strategy is formulated according to the degree of target interference.Aiming at the problem of fast target motion,a tracking algorithm based on fusion of target motion position prediction is studied.By fusing the above three methods with the correlation filtering tracking algorithm based on spatiotemporal context information,the accurate target tracking is achieved.3.The multi-feature fusion and cascade detection target tracking algorithm is studied.Aiming at the influence of the third level interference factors such as severe occlusion and short disappearance of targets in industrial environment and the robustness requirement of tracking algorithm for industrial intelligent robots,a target description method combining color statistical features,spatiotemporal context information and HOG features is proposed.The color feature is insensitive to the fast deformation of the target,the background information around the target and the HOG feature are insensitive to the change of illumination,which form the complementary performance of the feature and improve the robustness of the tracking algorithm.Aiming at the problem of target tracking failure caused by long time occlusion,a target detection algorithm based on cascade classifier is studied.When the target tracker error reaches the preset threshold,the detection algorithm retrieves and captures the target from the image,and corrects the tracking result to avoid the contamination of the target template,so that the target tracking algorithm has strong robustness and can adapt to the long-term tracking task.4.In order to further improve the practicability and applicability of intelligent robots in industrial environment and improve the fast performance of target tracking algorithm,a fast target tracking algorithm based on data dimension reduction method is studied.Principal Component Analysis(PCA)and Matrix QR decomposition are used to reduce the dimension of the target tracking algorithm adaptively based on the correlation filtering framework,reduce the computational complexity of the algorithm and improve the speed performance.5.An intelligent robot platform based on visual tracking is designed to verify the performance of target tracking algorithm in motion control of intelligent robot platform.Intelligent robot platform can detect and track specific targets in the traveling path,providing real-time,rich and effective information for robot motion control.Intelligent robot platform mainly includes vision-based obstacle avoidance module,motion control module,communication module and positioning module and other functional modules.The effectiveness of the tracking algorithm in practical application is verified by designing obstacle avoidance test and follow-up test.In summary,this paper focuses on the requirements of intelligent robot target tracking algorithm in industrial application field,and deeply studies the target tracking algorithm based on correlation filtering to improve the robustness,accuracy and real-time performance of the target tracking algorithm under complex conditions.It also provide the theoretical and data support for the popularization and application of intelligent robots in industrial field.