【作者】 王树文； 张长利； 房俊龙；

【Author】 Wang Shuwen, Zhang Changli, Fang Junlong (Engineering College, Northeast Agricultural University, Harbin 150030 China)

【机构】 东北农业大学工程学院；

【摘要】 为了有效地减轻农业生产的强度,提高劳动生产率,保护环境,本文综合运用计算机视觉技术、图像处理技术、人工智能等,建立了一个杂草自动识别系统,实现田间杂草的自动识别。田间杂草的识别是准确使用除草剂有效的依据,对提高农作物产量、降低成本、生产出满足消费者需求的无害产品是非常重要的。首先,通过计算机视觉系统获取带有复杂的背景物的杂草图像,利用图像预处理的滤波去噪、图像分割、图像增强等多种基本图像处理的方法对杂草图像进行了处理。其次,利用彩色图像的颜色特征和形态特征从复杂的土壤背景环境下分离出植物(作物和杂草),应用形态学中的膨胀和腐蚀方法对形状上有明显差别的植物和杂草进行分离,并对杂草区进行标识。本文中可变阈值分割方法和“栅格”技术的采用,使该计算机视觉系统节省了时间,提高了精度和识别处理的速度。实验证明从复杂的土壤背景下识别出杂草,成功率达到96%。更多还原

【Abstract】 To effectively alleviating agricultural production intensity and improving work productivity and protection environment, Computer vision, image processing and artificial intelligent technology were synthetically applied to setting up automatic identification system of field weed for identifying field weed. Accurate identification of field weed will provide effective standard for using of herbicides, it is very important to increase crop production, reduce cost and produce harmless product of satisfying for consumer requests. First, the images of weed under complex soil background were captured through computer vision system, then to process weed image applying basic methods that include filtering noise and dividing of image and highlighting of image Second, applying color and shape feature separated the plants from complex background and applying erosion and dilation of morphological algorithm feature separated weed from plants, at the same time to show weed district. In this paper changing threshold segmentation and grid technology is applied so that computer vision system is to save time, to improve accuracy and processing speed. The experiments show that the rate of identification precision for weed from complex soil background was not less than 96%.更多还原