基于自反馈模板提取的车辆遥感图像识别

Remote Sensing Image Recognition for Vehicles Based on Self-Feedback Template Extraction

模板提取技术的空白使绝大部分高效的模板匹配算法建立在人工提取模板的基础上,人工提取模板的缺陷必将在识别过程中逐级传播,从而降低图像识别的准确度与稳定性,最终影响到基于遥感图像的交通状态辨识效果.文中针对模板匹配图像识别方法中的模板提取问题,提出了一种基于自反馈模板提取的车辆遥感图像识别方法,并利用数学推导论证了自反馈模板提取方法的正确性,以Matlab软件为平台完成了多个路段车辆高分辨率对地遥感图像的识别和交通流的辨识.对多个路段高分辨率图像识别结果的分析验证了文中方法的可行性和有效性.

Due to the blank of template extraction technology, most efficient template matching algorithms are con- structed on the basis of artificial template extraction. The defects of current template extraction may reduce the accuracy and stability of image recognition through their progressively spreading in recognition process, and finally affect the results of traffic state identification on the basis of remote sensing images. In order to solve this problem, a remote sensing image recognition method for vehicles is proposed on the basis of self-feedback template extrac- tion, and the correctness of self-feedback template extraction is demonstrated by mathematical derivations. Then, a high-resolution remote sensing image recognition and a traffic flow identification are carried out for several certain road sections on the platform of Matlab. Finally, the feasibility and effectiveness of the proposed method are verified through analyzing the remote sensing image recognition results of several road sections.