Detecting Crop Rows for Automated Rice Transplanters Based on Radon Transform

摘要

This paper describes a vision-based crop row detection algorithm for developing guidance sensor of automated rice transplanters. Saturation images show robust to enhance rice sprouts that they were not affected by shadows and light reflection. Color images were transformed into saturation images and binary images sequentially, and then they were processed with method of Radon transform to extract all possible lines. The crop line for guidance reference was extracted according to the position of camera. It was verified by 138 images taken in early rice season. Experiments showed the vision sensor could do perfect when paddy fields were prepared carefully, while required some improvement when it was used on untidy paddy fields. Embedded real-time vision sensor will be developed in near future.