基于形态学的农田景物区域检测技术

研究了形态学在农田景物区域检测中的应用，通过形态学分析对农田景物区域进行了形态特征提取，在此基础上利用先验知识对各区域进行了识别分类。试验表明，本文提出的方法可以取得良好的效果。

Detection of Areas of Cropland Scenery Using Morphology

Farm crop image analysis is the basic part of information process technology of visual navigation for the autonomous agricultural robot. Automatic detection of the cropland areas under the complicated background is dealt with in this paper. Some knowledge based approaches were introduced to implement the area detection. In order to detect areas, a lowpass filter was used to preprocess the original images at first, then binary image was obtained by means of grad transformation and the morphological features of farm crop image were extracted. On this basis, cropland areas were classified. After the thin texture was moved with mathematical morphology thresholding dilation, the complete areas of cropland could be obtained. All of the algorithms proposed herein are knowledge based. Even the binary images of cropland are occluded or fragmentized, the areas can be detected efficiently and effectively. Experiments show that this method is reliable.