基于图像识别的作物分类种植面积车载测算系统设计

获取作物种植面积是农情监测的主要工作。车载系统进行作物面积测算是作物面积抽样调查的方法之一。文章提出了一种基于图像识别的作物分类种植面积车载测算系统CAOES(Crop Acreage Onboard Estimation System)，分析了其原理与组成，并进行了系统面积测算的视觉测距模型的推导和试验验证。试验验证结果表明：1)镜头焦距定为8 mm能够兼顾纵向视野和图像细节两方面的要求；2)光圈和曝光系数通过软件自适应调节可以满足图像分析需求；3)经验测距模型较之二次多项式回归模型相关系数更大，预测更稳定；4)该系统用于作物面积抽样调查可进一步降低野外调查结束后的工作量，并最大限度地减少人为因素的影响，提高调查的客观性和时效性。

Design of categorized crop acreage onboard estimation system based on image identification

Crops acreage estimating is the fundamental work for government's agricultural information monitoring on a national scale. Measuring with onboard estimation system is one of the approaches to perform ground truthing. In this article the authors propose a system, namely Categorized Crops Acreage Onboard Estimation System(CAOES) Based on Image Identification System, CAOES for short, and analyze its principle and configuration. Then the authors give the vision measuring model for acreage estimating and its precision evaluating. The test results indicate: 1)FA lens with 8mm fixed focus reach a balance between the two demands of vertical visual field and image detail;. 2) with the software, the aperture and exposure index can be self-adapting to the natural light to meet the demands of image analysis ; 3) the experiential model has bigger correlation coefficient and more steady predicted value than quadratic polynomial model; 4) this system may be more laborsaving, impersonal and in time.