Abstract
This paper investigates the problem of deriving vehicle guidance information from images of crops grown in rows. Advantage is taken of the approximately known crop geometry and camera calibration to derive a Hough transform which will detect parallel crop rows. Because the transform integrates information over a number of rows as well as within rows it can operate in a robust fashion with relatively few image features (plant blobs). The algorithm is implemented on one transputer along with some dedicated hardware. The method has been used on six sequences of 30 images and operates at a rate sufficient to give 10 values of vehicle offset and heading angle per second. Typical errors are 12.5 mm of offset and 1.0° of heading angle.
Users
Please
log in to take part in the discussion (add own reviews or comments).