Vision-based Navigation Line Extraction by Combining Crop Row Detection and RANSAC Algorithm

Abstract

Crop row detection is an important technique for the agricultural robot to spray and weed the target. A new method based on visual navigation line detection was proposed for the current phenomenon of serious seedling pressure in plant protection operations with agricultural machinery. The light-independent Cg component was constructed on the basis of the YCrCb colour model. The 2Cg-Cr-Cb feature factor was selected to greyscale the image. The Otsu algorithm was used for image segmentation to reduce the effect of illumination changes on image segmentation. Following this, morphological processing was applied to filter out the noise points in the image. Feature points were extracted according to the horizontal strip method. Straight lines were fitted by least squares after rejection of outliers in crop rows using the RANSAC algorithm. As a result of the experiments, the algorithm proposed by the research provides a reliable guidance line extraction for the weeding robot along the centre line of the maize row.