Advances in signaling techniques for crop detection in robotic weeding: A review

Abstract

Weed is an undesirable plant that grows with the main crop and compete with the main crop for water, nutrients, light and space. Timely removal of weed is essential for good crop yield. The manual weed control is a labour and time intensive task. The mechanical weed control facilitate only inter row weeding. Machine vision technology uses a series of image-processing methods to extract the shallow features of weeds or crops such as texture, shape, color, or spectral features of images and sends them to a classifier to identify crop/weed and communicate the results to a real time controller to activate the weed removal mechanism which is either a mechanical knife or spraying system. Still intra-row weeding is one of the most challenging field operations because of lacking technology to accurately distinguish between weed from crop in real field condition. Crop signaling is a new concept defined as any communication process that governs basic interactions between sensors and exogenous fluorescent signals applied to crops. It reduces the difficulties in conventional machine learning and deep learning technologies in terms of precise crop location, less graphical processing time and adaptability to in cooperate in actual field condition in a moving platform. The different crop signaling techniques such as systemic markers, green fluorescent protein, plant labels and topical markers are cost effective as well as good in locating the crop position with high accuracy. The robotic weeding system integrated with topical marker in tomato plant reported an overall crop detection accuracy of 99.1% at a forward speed of 3.2 kmh-1 at 30 ms-1f-1.