Robot-Based Weed Identification And Control System

Abstract

India loses agricultural produce worth over $11 billion annually due to weeds that is more than the central government’s budget allocation for agriculture in 2017–2018. There is on average 10% yield loss is reported due to weeds in agriculture. Therefore, weed control is an important aspect for the sustainable agriculture and increases in the productivity of the crop. The weeds can be controlled by mainly five methods, that is, manual, chemical, mechanical, electrical discharging, and flame methods. The most efficient method for weeding is manual weeding but it has its own disadvantages. However, the most preferred method for weeding is chemical spraying, that is, spraying of herbicides. The uniform herbicides application has proven its effectiveness in weed control. Contrary to this, it invites harmful effects on human health, environmental pollution, and wastage of input. Because of these negative consequences, agriculture is being promoted by farmers and government for limited use of herbicides. This problem can be solved by utilizing precision farming methods to address individual plants or tiny weed clusters. However, human-based precision technology requires efficient and labor-intensive human resources, 170which cannot be justified by the cost-saving from herbicide reductions. Therefore, there is huge scope of smart robotic weeding system to solve the problem of labor scarcity, labor cost, and save the resources with precise application of herbicides. The robotic-based weeding includes the following process, that is, firstly, weed and crop must be identified precisely. Detection or identification system should comprise of weed or plant detection; plant or weed classification along with their location. Second, weed control mechanisms should be designed in the perspective of weed management approach that suits the production system and cultivation practices. Third, weed control mechanism must be operated on the basis of information perceived by the perception system. Finally, all technologies should be combined to form one weeding system. There are many robotic-based technologies developed to identify the weeds and after identification removal of weed, like the AVO weeding robot developed by Eco-robotics works by identifying weeds that are not similar to the selected crop. This robot mainly targets the weed with a small dose of herbicide, allowing farmers to use 90% less herbicides. Similarly, three robotic weeders, that is, Oz. Ted and Dino are developed by Naio Technologies. Similarly, the MF-Scamp Robots designed by Blackmore identify and locate the crop in rows and columns by using vision sensors and steer accurately to a larger extend for precise application of herbicide. Another robot “BoniRob” developed by Deepfield robotics navigates autonomously along the plant rows to control weeds by using environmental sensors like satellite and GPS (row detection and Navigation), LiDAR, inertial sensors, and wheel odometry. Some of the complete autonomous weed control systems have witnessed the effective working of this technology in the field. Still, there is a need of more precise research and development in the robotics for weed perception and control to fully realize the scope.