POSE-GUIDANCE ALGORITHM PERFORMANCE FOR DIFFERENT CROP STAGES OF DEVELOPMENT

摘要

An algorithm that treats the estimation of the vehicle guidance parameters, offset and heading angle, as a pose recognition has been developed. The pose-guidance algorithm (PGA) classified the pose (offset and heading angle) of an image by using the minimum Euclidean distance classifier after encoding it into an eigenspace built by principal component analysis. PGA performance was tested for com and soybean with five different stage of maturity. The PGA performance was measured by mean absolute error between the actual and estimated poses of 100 test images. The PGA performance was similar for binary and raw images. Each crop and stage of maturity presented different heading angle and offset average absolute errors. The heading angle average error for all stages was 0.5 and 0.6 degree for corn and soybean, respectively. The offset average error for all stages was 3.1 and 3.9 cm for corn and soybean, respectively.