Greenhouse Flowers Automatic Loading System Based On Visual Orientat

With the rapid development of sensor technology,computer technology,artificial intelligence and other related disciplines,the robot forward with intelligent direction of self-organization,self-learning,self-adaptive ability.For mobile robot,autonomous navigation ability is an important embodiment of the mobile robot intelligent level.There are several ways in robot navigation,in which,vision navigation is an important direction of research on mobile robot navigation.In the research of robot behavior control and visual navigation,how to improve the accuracy of robot's behavior and the real-time and veracity of visual image recognition system in an unknown environment is one of the hotspots in the research.To solve these problems,this paper presents a design scheme of high-speed image acquisition and processing system,basing on DSP-FPGA,which is applied in loading potted flowers in greenhouse.In this scheme,the system processes the captured images with edge extraction,sharpening,as well as the simulation experiment.The results of simulation experiment shows that using an improved median filtering algorithm in image processing of FPGA is not only good for the denoising of the captured image,but also has very fast computational speed.In the moving and working process of a vision robot,how to realize autonomous precise localization is a key technology for autonomous navigation of robot.This paper studies the self-localization method of robot vision based on image features,and systematic research on the process and effect of visual robot localization from the camera calibration,background modeling,and foreground object extraction to feature point extraction,motion estimation and Kalman filtering.After analyzing and comparing several kinds of feature points extraction methods in robotic vision localization,we propose an improvement on the existing robot vision positioning method.We propose an improved method for precise localization of robot based on Harris feature points,relying on the traditional Harris feature point detection,firstly,this method exclude some non-corner points and false corner points,and then process the other pixel,using the relationship of the gradient change of gray value of the pixel around the corner point,and finally draw the correct feature point by simple calculation and analysis.In the analysis of the experimental design and results,compared with other algorithm results,the results show that the algorithm in this paper obtains more accuracy in motion estimation of potted flowers and accurate tracking in shorter computing time,which shows that this algorithm has practical prospect.