Development of line-detection algorithms for local positioning in densely seeded crops

摘要

Herbicide applications could possibly be reduced if applications were targeted. Targeted applications require prior identification and quantification of the weed population. This task could possibly be done by a weed scout robot. The ability to position a camera over the inter-row space of densely seeded crops will improve the quality of the weed detection. Four line-detection algorithms were evaluated with simulated crop row images. Based on the results, the algorithms were revised and tested againwith field images of wheat collected at the 3-leaf, 5-leaf, 2-tiller, and 3-tiller stages. When tested with simulated crop row images, the best algorithm was able to determine the angle of parallel lines in an image with an accuracy of 0.5°, and to findthe centreline of an inter-row space within 2.1 mm, even in noisy conditions. When tested with field images containing weeds, the revised version of the same algorithm was able to determine the angle of the inter-row spaces within 11.5 ° and determinedthe location of the centreline of an inter-row space with an accuracy of 27.2 mm. Results of all algorithms improved when only the 2-tiller and 3-tiller stages were considered. This paper documents the performance of the line-detection algorithms underlab and field conditions.