一种针对遥感图像中群目标的自动识别方法

An Automatic Teeming Target Recognition Method from Remote Sensing Images

传统群目标识别方法通常利用目标与背景的纹理差异分割目标,虽然具有简单高效的优点,但是对群目标的空间方位没有直观的认识。针对遥感图像中群目标的特性,提出了利用子目标群方位关系来实现群目标自动识别的算法。算法在边缘提取的基础上,改进了基于置信度的Hough变换子目标定位方法。将目标群方位、结构信息等与RO I区域的分离紧密结合起来从而达到后续精确判断目标群装备的目的。实验结果表明相比于传统方法,算法在检测遥感图像中群目标时具有更高的精度。

Texture difference between target and background is widely used in teeming target recognition to segment target. This method is simple but without good understanding of directional spatial relationship. An algorithm based on directional spatial relationship between groups of sub - objects for recognizing teeming target automatically is proposed. After edge detection, improved Hough transform based on reliability to locate the sub - objects is adopted. The directional spatial relationship between groups of sub - objects, structural information and separation of ROI region are combined together tightly in order to reach the goal of judging the condition of equipment accurately further. Experimental results show that in contrast to traditional methods, this proposed algorithm is effective for teeming target recognition from remote sensing images.