Researches On Weed Recognition And Vision Navigation Of Weeding Robot

Automation and intelligence of agricultural machinery is the trend of China's agriculture in 21 century. In 2007, China's central government proposes that Developing the modern agriculture would be the primary task of construction of"New Countryside"and investment on the agriculture technology would be increased. Precision Agriculture is a new concept, and it will optimize the use of fertilizer, herbicide and water, etc., and at most reduce the destruction on the environment. This dissertation is to have a research on weed recognition and navigation of weeding robot based on machine vision with a view to improve the intelligence and automation of agricultural devices.In this paper, it elaborates the domestic and international research on weed recognition and navigation based on machine vision.This research uses USB interface camera to acquire the images,and carry out image segmentation using color factor in RGB color space directly.After image segmentation,the binary image could be filtered to delete the noise by mathmatics morphology.Dividing the image by 25 regions,we can get the position of weed according to region area threshold,and finish the job of weed recognition based on machine vision.This dissertation is also to have a research on vision navigation of weeding robot.The research uses Model0776 manufactured by 3Com company in America to acquire image, carry out image segmentation using color factor in RGB color space directly. After image segmentation,the binary image could be filtered to delete the noise by mathmatics morphology,detect central ine of crop rows by optimized Hough transformation. At last, compute the pose of weeding robot according to the pose of navigation camera and perspective principle of camera.Indoor weeding robot prototype was developed in this research, which consist of navigation camera, weed recognition camera, computer, body, a robotic arm and etc. Test shows that weeding robot can recognize the weed by the rate of 100%,and algorithm is good real time and also the precision of pose of weeding robot is high.The algorithm of weed recognition and vision navigation is insensitive to outside condition change.