Dual Stereovision Application for 3D Field Mapping and Vehicle Guidance

Abstract

This research aims to develop a stereovision-based field recognition system capable ofperforming 3D crop row detection for tractor guidance and 3D field mapping for crop growthassessment. A stereovision camera can capture a 3D field scene by means of integrating two planeimages captured using a binocular camera. This paper focuses on introducing a 3D field mappingmethod for determining basic biophysical parameters of crops, including the height, the canopy area,the volume and the leaf area index, as the quantitative measures of crop growth status. Thisdeveloped method is capable of estimating the motion of a stereo camera mounted on a tractor bytracking feature points in sequential stereo images, and creating a 3D field scene map by integratingthe recovered camera motion with scenery feature points. Results obtained from field validationtests indicated that the developed image processing method can create a 3D field map to explicitlyshow the variability of crop heights with an RMS error of 0.04 m and extract crop row positioninformation for supporting automated guidance to navigate a tractor following crop rows reliably.