A guidance directrix approach to vision-based vehicle guidance systems

In a vision-based automatic agricultural vehicle guidance system for row-crop applications, finding guidance information from crop row structure is the key in achieving accurate control of the vehicle. This paper describes a robust procedure to obtain a guidance directrix. The procedure includes row segmentation by K-means clustering algorithm, row detection by a moment algorithm, and guidance line selection by a cost function. Auxiliary information, such as the known crop row spacing, is used to aid in the development of the guidance directrix. Two image data sets, one taken from a soybean field and the other taken from a corn field, were used to evaluate the accuracy of the proposed image processing procedure. The average RMS offset error from 30 soybean images was 1.0 cm with an average cost of 4.99. In contrast, the average RMS offset error from 15 corn images was 2.4 cm with an average cost of 7.27. The proposed image processing procedure was implemented on a vision-based guidance tractor. © 2004 Elsevier B.V. All rights reserved.