A bandpass filter-based approach to crop row location and tracking

A method of locating crop rows in image sequences is described. Unlike several previously reported algorithms, the method does not rely upon the segmentation of plant material from the background on the basis of absolute brightness or colour. Rather, the periodic amplitude variation due to parallel crop rows is exploited. Given the geometry of the camera arrangement and the crop row spacing, a filter is derived which allows the crop rows to be extracted whilst attenuating the effects of partial shadowing and spurious features such as weeds. The position and orientation of the rows are tracked using an extended Kalman filter. The method has been used to guide a mechanical hoe in winter wheat with an RMS positional error of 15.6 mm at a speed of 1.6 ms-(1), despite the presence of complex shadows cast by the tractor in the imaged area. (C) 2000 Elsevier Science Ltd. All rights reserved.