基于神经网络的杂草识别试验研究

作者：袁海波1,史 岩2,梁安波3,葛艳艳4 日期：2008-04-23/span> 浏览：3904 查看PDF文档

基于神经网络的杂草识别试验研究
袁海波1,史 岩2,梁安波3,葛艳艳4
（1.青岛农业大学 机电工程学院，山东 青岛 266109;
2.青岛农业大学 建筑工程学院，山东 青岛 266109;
3.青岛农业大学 国际合作交流处，山东 青岛 266109;
4.东南大学 机械工程学院，江苏 南京 211189）〖JZ)〗

摘 要：以苗期玉米和冬小麦为例，对比土壤、小麦、玉米、玉米田间杂草及小麦田间杂草的两种绿色增强因子统计图，利用过绿特征（2GRB）作为参数，结合BP神经网络，运用计算机图像处理技术，采用像素位置直方图法，识别出了杂草，并确定出了杂草位置、面积。实验结果表明，该方法可以准确识别出杂草，误差正确识别率高；小麦和玉米田间杂草识别时间短，可以满足实时性要求。
关键词：变量施药；杂草识别；图像处理

Analysis and testing of weed identification based on neural network
YUAN Haibo1, SHI Yan2, LIANG Anbo3, GE Yanyan
 (1.College of Mechanical & Electrical Engineering, Qingdao Agricultural University, Qingdao 266109, China;
2.College of Construction Engineering, Qingdao Agricultural University, Qingdao 266109, China;
3.International Intercommunion School, Qingdao Agricultural University, Qingdao 266109, China;
4.College of Mechanical Engineering, Southeast University, Nanjing 211189, China)

Abstract: Contrasting the two green strength genes of the soil, wheat, corn, and the weed, a system to identify the weed from the crop was designed. It used the 2GRB and BP neural network, with the help of pixelpositionhistogram diagram, to calculate the area and position of weed. The results show that the method could identify the weed from the field and crop, and its correct rate is high. The program gives the result that running time of identifying weed in wheat field is short. In a word, the system can satisfy the request of real time.
Key words: variable spray; weed identification; picture disposal