Real-time Identification of Crop Plant Based on Computer Vision and Crop Signaling Technology for Weed Control

Weeding, an important aspects of vegetables growing, requires exact discrimination of crops and weeds which is a major challenge in realizing the process of precision weeding. Crop signaling technology is a novel system approach which could create a machine-readable signal in crops, which are absent in weeds, to distinguish them. Combination of computer vision and color mark sensors (CMSs) was designed to detect the crop plants growing with unique machine-readable plant tags. The effects of different factors (color and area of the plant tags, number of CMSs) on the determination precision was evaluated. In addition, the effects of different concentrations of systematically crop signaling compounds on signal expression (fluorescence intensity, crop vigor, photostability) was explored. The results showed that six CMSs in association with top part of white tags were the optimal configuration for detecting tomato plants. The system effectively detected plant tags and quickly categorized crops and weeds. A one-day treatment to lettuce with 180 ppm compound produced a machine-readable fluorescent signal up to 25 days after transplanting without affecting crop growth. This study provides new insights for accurate identification and localization of crops in weeds.