Engineering Realization Of Vegetation Recognition For UAV Low Altitude Aerial Survey Images In Gobi Area

In recent years,the research of vegetation recognition using UAV aerial survey images is in the ascendant,but it mainly concentrates on the areas with high vegetation coverage,such as forests,urban green space,and so on.The recognition of Gobi vegetation is of low concern.This paper integrated UAV low altitude aerial survey technology and image recognition method,and made a useful exploration around the engineering realization of vegetation recognition in Gobi area.In this paper,taking the Gobi area in Northwest China as the study area.Combined CW-20 UAV with Pix4 D Mapper and eCognition Developer,an experimental platform was established.The layout scheme of image control points was optimized and verified.Based on red and green coding theory,the calibration field of non-metric camera was established and verified.The low altitude aerial images of UAV in the study area were processed and the Digital Orthophoto Images were generated.The Digital Orthophoto Images were segmented and the segmentation effect was evaluated with object-oriented multiscale segmentation technology.Based on visible light vegetation indices,the objects in Gobi area were identified and the accuracy was evaluated.The experimental results showed that,the mean square error of the check points was ?X = 8cm,?Y = 12 cm,?Z=22cm,which proved the accuracy is far higher than the requirement of 1:1000 topographic map specification for aerial triangulation;in terms of recognition accuracy,the producer accuracy of vegetation identification reached more than 90%,and the kappa coefficient of overall identification of objects in Gobi area reached more than 0.8.It proved the feasibility of vegetation identification in Gobi area based on UAV low altitude aerial survey images,and achieved the purpose of obtaining the information of vegetation distribution and providing technical support for Gobi ecological protection.