Study On Weed Detection Using Machine Vision

On the field of agriculture and plant protection, weed identification technology in machine vision research at home and abroad has made part of the results. Many studies have focused on the individual weed plants. Biological networks are an essential part of the system, the reality of the weed population is composed of many plant communities in the network.Taking weed community as the research object, using machine vision to identify weeds, this paper designs a weed from between-row crops elimination method based on position feature of crops. Many scientists believe that network of weed is a random network. The paper investigates weed community using the idea of the complex network, weed plants abstract for the node, allelochemics released by weed abstract for the link, build network model of weeds. We research characteristics of the model, such as the power law, growth, preferential attachment, and find the weed networks have small-world properties and scale-free properties. Based on these characteristics, we establish a weed evolution model of scale-free networks. The model uses two types of attack strategies and found that error tolerance and attack vulnerability are generic properties of weed networks. According to the characteristics of the weed network, we can remove the high price node of weed networks, so it will reduce the dosage of chemical herbicides and be good for environmental protection.Finally, the paper designs weed identification system based on machine vision. We set up the hardware environment which can achieve the collection, storage and display images of weeds. System program use Visual C++ as the development tool. This system exploits layered design using modularization thinking, and every layer may fulfill the different function.