Research And Implementation On The Algorithm Of Plant Location And Measurement Based On Binocular Vision

Inner Mongolia is rich in biodiversity,but environmental pollution and climate deterioration have led to an increase in endangered plants year by year.It is urgent to conduct accurate and effective plant protection.Using drones equipped with binocular cameras to investigate the survival of endangered plants in the wild,determine the spatial distribution of endangered plants,and then measure biological characteristics and other data,which are of great significance to understand the growth of plants and identify the endangered level of plants.This paper focuses on the two requirements of plant location and biological character measurement,and studies the plant location and measurement algorithm based on binocular vision technology.The main contents of the algorithm include binocular camera calibration,image defogging and stereo matching.In the calibration part of binocular camera,the calibration method for the internal and external parameters of the camera in this article is Zhang's calibration method,and the back projection error of the calibration results is analyzed.In the part of the image defogging,aiming at the halo phenomenon and the color distortion of the bright area in the classical dark channel prior defogging algorithm,an improved dark channel defogging algorithm is proposed,which combines the fog density coarse and fine mapping with the transmittance optimization of the bright area based on the tolerance mechanism.Compared with the classical defogging algorithm,the improved algorithm is clearer and faster in image defogging.In the stereo matching part,to solve the problem of slow search speed of traditional orb algorithm using BF algorithm,this paper uses LSH algorithm to optimize it effectively.Aiming at the lack of matching accuracy of traditional orb algorithm,this paper proposes an improved orb algorithm which combines grid motion statistics(GMS)and PROSAC two-step purification algorithm,the matching accuracy is improved.In addition,this paper also studies the local stereo matching algorithm based on disparity plane.The algorithm generates complete disparity map,and the matching speed meets the real-time requirements.Finally,this paper designs a plant location and measurement system.Based on VS2010 platform,calling opencv2.4.10 library,aiming at Ulmus pumila cv.jinye,the whole process of plant positioning and measurement system design is gradually realized.Through the experiments of small herbaceous plants positioning and measurement and Simulating UAV positioning at high altitude viewing angle,the results show that the system designed in this paper has small error,and the accuracy and real-time performance meet the actual needs.