Crop picking system based on instance segmentation and path planning

Abstract

The invention provides a system for improving crop picking precision and controlling a mechanical arm to move to automatically find a way based on computer vision. According to the system, the work ofYolact on ResNet-101 is combined,; a traditional two-stage instance segmentation model is modified into a one-stage model, 64 robust masks are generated from the deepest layer, prediction coefficients are output according to four layers of feature pyramid networks in sequence, and mask coefficients of different layers are endowed with different weights to improve the mask precision. Not only is aplant positioned, but also a mask is generated to tightly cover the crop pixel by pixel, and the central point and contour coordinates of the crop are output. Then an A * algorithm and a heuristic function of the A * algorithm are used for sequentially utilizing the coordinates to carry out piecewise shortest path planning and generate corresponding G code statements to control a mechanical arm to move between plants and on the plants, and finally a mechanical claw is combined with crop center points and contour coordinates to pick crops.

Description

technical field [0001] The invention relates to intelligent picking in intelligent agriculture, in particular to a grasping system for controlling the automatic pathfinding of a manipulator and identifying crops. Background technique [0002] Currently, there are mainly two types of picking machinery that are most used in the world: one is the fruit picking machine that separates the fruit from the fruit tree by shaking or impacting. The shaking and catching methods usually cause damage to a degree that cannot be achieved by fresh fruits. Accepted; the second approach is selective crop harvesting using picking robotics. Since the 1980s, experts and engineers in related fields have started to research and develop crop picking robots. The typical approach employed combines computer vision systems with robotic arms and end effectors to selectively pick individual ripe crops. [0003] At present, the deep neural network algorithms commonly used in automatic crop picking system...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problems of low detection accuracy, poor shape recognition effect, long path search time, and stagnation after a long search, this invention proposes a heuristic automatic pathfinding and grabbing system

Images