摘要
由于微波辐射的复杂特性 ,从合成孔径雷达图象上提取类似于居民地复杂结构的目标物的研究仍处于探索中 .通过研究居民地对合成孔径雷达 (SAR)的微波散射特性 ,分析居民地在 SAR图象上的纹理特征 ,综合利用纹理分析、模式识别和颜色空间变换技术 ,提出了一种新的提取雷达图象上居民地的方法 .该方法在共生矩阵纹理分析的基础上 ,选取 3个合适的特征分量合成彩色纹理特征图象 ,再通过 HIS变换获得亮度分量 ,使用亮度阈值分割图象来提取出居民地 .此方法的特点是 ,其受雷达系统影响较小 ,适应性较强 ,以二值图象的形式记录居民地的提取结果 .试验表明 ,利用此方法在 SAR图象上提取居民地具有 70
Because of the complexity of the microwave radiation characteristics, the recognition for objects with comprehensive structure such as habitations from Synthetic Aperture Radar(SAR) image is still in investigation. A new method for extraction of habitation areas from SAR images is presented. In accordance with the mechanism of the microwave dispersion, the texture feature of habitation on SAR image was analyzed. Based on the technology of texture analysis , pattern recognition and color space transformation, the main steps of the new method to extract the habitation areas from SAR image are: firstly, many texture images are generated by analyzing the Gray Level Co occurrence Matrix(GLCM); Secondly, 3 suitable texture images, in which the habitation areas contrast to background clearly, are selected from all of the texture images to combine an artifical color texture image; Thirdly the color texture image is transformed into the intensity image by the HIS color space transformation technique; Finally, a threshold was used to decide whether each pixel on the intensity image belongs to habitation or not. Experiments illustrate that this method is adaptable and its right recognition rate is beyond 70 percentages.
出处
《中国图象图形学报（A辑）》
CSCD
北大核心
2003年第6期616-619,共4页
Journal of Image and Graphics
基金
国家自然科学基金 ( 698962 5 0 -4 )
国家高技术研究发展 863计划 ( 2 0 0 1AA13 2 0 11)
关键词
纹理图象亮度阈值法
微波辐射
合成孔径雷达
SAR
HIS变换
遥感图象
Computer image processing, Co occurrence matrix, Texture image, HIS transformation, Intensity threshold Separation
