Computer vision guidance of field cultivation for organic row crop production

Abstract

<abstract> <bold>Abstract. </bold>Today, many agricultural vehicles are equipped with instrumentation such as global navigation satellite system (GNSS) receivers and LIDAR to assist operators in guiding agricultural implements. For organic row crops, in-season weeding is required and cultivators can be equipped with mechanical sensors to assure proper positioning of the working tools. However, mechanical sensors perform poorly in the early stages of crop growth. Alternative computer vision systems are available commercially but are expensive. Therefore, the objective of this study was to develop a webcam-based system which is capable of supplementing the mechanical guidance system for row crop cultivation during the early stages of crop growth. A computer vision guidance system was developed for a 700 MHz ARM nanocomputer to control a legacy mechanical guidance system. A low-cost CCD camera was mounted to the cultivator frame in-line with a crop row to obtain a video stream of the plants passing beneath the equipment. The Python OpenCV platform was used to develop an application to identify the lateral offset of the plant rows and to adjust the hydraulic steering accordingly. The system was tested successfully for travel speeds up to 6 km/h in several corn and soybean fields under varying ambient light and crop conditions.