摘要
自动从大量网络图片中识别出地图是地图监控中一个非常重要的任务。从人类视觉认知上,地图图片明显不同于一般的图片,因此需要研究地图图片的特征及其在自动识别中的应用。通过分析地图图片与一般图片的差异,选择了地图图片的颜色特征、局部二值模式特征和灰度共生矩阵特征作为地图图片识别的基础特征,然后针对不同特征找出最优的单核支持向量机,通过线性加权组合生成多核支持向量机,对地图图片进行自动识别。实验证明:这些特征对地图图片的自动识别是有效的,结合多核支持向量机,地图图片识别准确率高。
Automatically identifying a map from a large number of network images is a very important task in map monitoring. From the perspective of human visual cognition, the map is obviously different from the general picture. Therefore, it is necessary to study t-he characteristics of the map image and its application in automatic recognition. After anal-yzing the difference between the map image and the general image, the color features of the map image, the local binary pattern feature and the gray level co-occurrence matrix feature are selected as the basic features of the map image recognition, and then the opt-imal single-kernel support vector is found for different features. The machine generates a multi-kernel support vector machine through linear weighted combination to automatically id-entify the map image. The experiment proves that these features are effective for automatic recognition of map images. Combined with multi-kernel support vector machine, the accuracy of map image recognition is improved.
作者
王雪冰
郭庆胜
王勇
柳其志
魏智威
WANG Xuebing;GUO Qingsheng;WANG Yong;LIU Qizhi;WEI Zhiwei(School of Resource and Environment Sciences,Wuhan University,Wuhan 430079,China;Chinese Academy of Surveying and Mapping,Beijing 100830,China)
出处
《测绘与空间地理信息》
2019年第9期28-32,共5页
Geomatics & Spatial Information Technology
基金
国家自然科学基金项目——解决地图要素空间冲突的智能化协同模型和算法(41871378,41471384)资助
关键词
地图图片
自动识别
特征提取
颜色直方图
map image
auto-recognition
feature extraction
color histogram
