Determination of crop rows by image analysis without segmentation

Abstract

A method based on computer vision for detection and localisation of crop rows, especially of small-grain crops, is described. The method is intended for use in a system for automatic guidance of agricultural implements in selective treatment of rows and/or inter-row spaces, e.g. with an inter-row cultivator. The computer vision system consists of a colour video camera and a computer. The camera is focussed on the field surface from an inclined angle to obtain images that cover up to about five rows simultaneously. New images are continuously transferred to the computer, which processes them and calculates the necessary lateral movements of the implement. The processing method does not include a segmentation step, which is found in most other methods for plant detection. The segmentation step has been replaced by computation of centres of gravity for row segments in the image. This approach has proven to reduce the computational burden of the image processing software. The estimation of the orientation and the lateral position of the centre lines of the rows is accomplished by weighted linear regression. The accuracy of the estimation was determined by comparing the calculated row centre line with the position of a reference string, which was placed parallel to the row along the centre line of an adjacent inter-row space.