摘要
对常见的图像识别研究方法进行了总结,主要包括以下5个方面:侧重于农作物病虫害监测的颜色特征识别方法;应用于作物大小形状分级、叶面积、叶倾角等领域测算方面的形态特征识别方法,包括统计分析法、结构分析法和频谱分析法的纹理特征识别方法;利用高光谱分辨的遥感技术、光谱仪和近红外光谱进行植被、杂草和作物内部品质监测的光谱识别方法;最后,是集各种识别方法特长的综合识别方法。对这5种方法的适用范围和优缺点进行了分析比较及综合,构建出基于图像识别技术的农情信息提取技术体系系统框架,并提出了新的技术应用思考,从中发现目前图像识别技术中存在的问题,为今后的研究指明方向。
This paper aims to summarize the common image recognition research method, mainly from the following five aspects： the color feature recognition method for supervising crops diseases and insect pests; the form feature recognition method for classifying the crops＇ size and measuring leaf area and leaf dip ; the texture feature recognition method including statistics analysis method , structural analysis method, and spectrum analysis method; the spectral recognition method for monitoring vegetation, weeds and the crops＇ inside quality by means of using hyper spectral remote sensing technique, spectrograph and near-infrared spectral; the integrative recognition method combing all the advantages of the methods were mentioned above. The application ranges as well as the disadvantages and advantages of the five methods were analyzed, compared and integrated, a systematic frame of collecting agricultural condition information can be constructed based on image recognition technique and then the reflections on new application area can be put forward, from which the existing problems of image recognition can be discovered and the future research will undergo in the right direction.
出处
《自然灾害学报》
CSCD
北大核心
2011年第3期18-24,共7页
Journal of Natural Disasters
基金
国家863计划"星陆双基遥感农田信息协同反演技术"(2009AA12Z143)
关键词
农作物
图像识别
图像特征
crops
image recognition
image feature
