RTD-SEPs: Real-time detection of stem emerging points and classification of crop-weed for robotic weed control in producing tomato

摘要

A novel technique for enabling robotic weed control in a commercial processing tomato field having densely populated weeds is described. It is necessary to accurately locate the stem emerging points (SEPs) of crop plants for the successful application of a mechanical weeding actuator to remove weeds during automated weeding. However, it is a difficult and challenging task to locate the SEPs in complex natural scenarios such as when the main stem is occluded by weeds or crop foliage, the crop plants are lying on the soil surface, there are non-uniform planting bed conditions, or there is leaf damage due to insects etc. To overcome these challenges a novel crop signalling concept has been proposed to mark the crop plants at planting to make them machine-readable. Plants lacking this crop signal were classified as weeds and removed by the robotic weed knife actuator. A machine-vision algorithm was developed to analyse the seven views of the crop plants taken by camera with help of a specially designed imaging chamber and locate the SEPs of tomato plants, which was passed to the robotic weed knife control algorithm to remove weeds. The algorithm was successfully detected and located the main stems of tomato plants in outdoor environment with success rate of 99.19% while traveling at a speed of 3.2 km h(-1) with a processing time for all views of 30 ms f(-1) (C) 2020 IAgrE. Published by Elsevier Ltd. All rights reserved.