Research On Automatic Spraying Method Of Fruit Trees

Automatic fruit tree pesticide sprayer,i.e.mobile machine carries out spraying work for plants depending on ambient environment to replace labor work and lower labor load.This process can be divided into four parts: power drive,motion control,ambient environment identification and decision making as well as pesticide spraying,among which ambient environment identification and decision making can be included into autonomous navigation.Simultaneous localization and mapping(SLAM)is the foundation and core of autonomous navigation and provides theory foundation and solution for it.SLAM i.e.processing information obtained with sensor based on corresponding algorithm to complete its own status location as well as the establishment of environment for mission target and using known map feedback to estimate its own position to adapt to complicated environment.This paper studied binocular vision-based SLAM in unknown environment.Pesticide spraying needs to determine the location and distance of the target first.Binocular distance measurement can better measure the distance to the target in image processing while high distance precision requires calculation based on camera-identified focal length parameter.Another key for binocular distance measurement is obtaining a stable point n the target.This paper adopted SIFT algorithm for obtaining stable point on target since the SIFT character point has stable scale,determined rotation and stable precision.However,not all SIFT character points are on the plants.Character point selection range should be restricted.This paper made improvement based on threshold segmentation algorithm and proposed a threshold segmentation algorithm based on largest component analysis so as to ensure all character points are on the plant and lower quantity of character points and thus improve calculation efficiency.Binocular distance measurement also requires image position offset of the same point in left and right camera.The same character point in two images can be identified based on character point matching.For those with many wrong-matched points,this paper made improvement in ultimate geometric constraint relationship and proposed a parallel restriction relationship to eliminate wrong-matched points by combing RANSAC algorithm.At last,center matching point is selected for distance measurement so as to use the distance between character points as input of SLAM system to use Kalman filter to estimate related status and make correction and then establish map,which laid a foundation for future study.In terms of experiment inspection,character points in related images were obtained and binocular distance measurement and matching experiment were carried out with good result.