Research On Environment Perception Technology Of Plant Protection Drones Based On Binocular Vision

In recent years,there has been an increasing demand for the use of drones to replace traditional manual operations in agricultural plant protection.Plant protection drones usually work at low altitudes and need to perceive complex operating environment.They can automatically avoid obstacles such as trees or telegraph poles by accurately measuring obstacles.How to provide fast and high-precision obstacle avoidance capability for plant protection drones has become one of the difficulties in promoting plant protection drones.Aiming at the phenomenon that the traditional feature matching is slow and inaccuracy,which leads to large ranging error,this thesis proposed a matching algorithm based on edge feature enhancement.Based on the binocular vision theory,some key technologies are researched and improved.An environment perception system based on binocular vision is designed to improve the accuracy of short-range measurement.Then the system has tested and verified.Firstly,three main camera calibration techniques are studied.Then this thesis has made a detailed principle and parameters error analysis for Zhang Zhengyou's camera calibration method with simple operation,high precision and low cost.This step ensures the accuracy of subsequent ranging based on the binocular vision environment perception system.After that,the thesis studied the process of image enhancement,including color map graying and filtering denoising.In addition,an adaptive median filtering algorithm is used to eliminate the mismatch between noise filtering and preserving image edge details.This adaptive median filtering algorithm can effectively filter noise and preserve image edge at the same time.The thesis research on the feature-based stereo matching algorithm.The SURF feature detection algorithm and the ORB feature detection algorithm are widely used because they both have high speed and robustness,then the thesis compared and analyzed these two algorithms.Afterwards,aiming at the working environment of plant protection drones,a feature matching optimization algorithm based on edge information enhancement of canny operator is proposed.After binarization operation,edge detection,ORB feature extraction and k-nearest neighbor feature matching on the denoised image,the algorithm uses RANSAC theory to eliminate mismatching points.Then by compared with the traditional stereo matching algorithm,it is proved that the optimization algorithm improves the accuracy and speed of stereo matching in short-range.Based on the research and optimization of the key technologies of binocular vision,an environment perception system based on binocular vision has designed and the computer implementation of software algorithm has completed.Perceptual obstacles in real-world environments using a calibrated binocular camera.The optimized image enhancement and stereo matching algorithm are used to process the images containing obstacles.The experimental results verify that the optimization matching algorithm has a certain improvement in matching accuracy and speed.At last,compared the optimization algorithm with the traditional SIFT binocular ranging,it proves the effectiveness of the method in the same environment in terms of accuracy and speed,which can provide reliable technical support for the automatic obstacle avoidance function of plant protection drones.