多尺度的小目标遥感网络传输图像分类检测

发布时间：2018-06-29 12:05

本文选题：遥感网络 + 遥感图像　；参考：《控制工程》2017年05期

【摘要】：遥感网络所传输的小目标图像在遥感图像环境中容易与背景图像的中央周边差产生视觉误差,造成遥感图像的视觉显著性检测效果不好。考虑局部像素的特征与相邻的像素特征之间的差异,对中央周边差的局部显著性进行变邻域搜索,提出一种面向遥感网络的遥感图像小目标图像显著性亮点检测算法,进行随机场邻域系统设计,构建小目标图像显著亮点NSDFB频域分解结构,基于多尺度的DoG滤波器叠加,求解目标特征在图像中的稀有度且评定其显著性,采用变邻域搜索方法实现对显著亮点检测模型的改进。实验结果表明,该算法进行遥感应用场景的小目标显著亮点检测分类结果性能较好,在遥感通信领域的小目标识别和多媒体信息特征检测检索等领域应用价值高。
[Abstract]:In the remote sensing image environment, the small target image transmitted by the remote sensing network can easily produce visual error with the central and peripheral difference of the background image, resulting in poor visual significance detection effect of the remote sensing image. Considering the difference between the local pixel feature and the adjacent pixel feature, the local saliency of central and peripheral difference is searched by variable neighborhood search, and a new algorithm for detecting saliency of small target image in remote sensing image is proposed, which is oriented to remote sensing network. The random field neighborhood system is designed to construct the NSDFB frequency domain decomposition structure of the small target image. Based on the DoG filter superposition of multi-scale, the rarity of the target feature in the image is solved and its significance is evaluated. The variable neighborhood search method is used to improve the model of bright spot detection. The experimental results show that the algorithm has good performance in detecting and classifying small target bright spots in remote sensing application scene, and has high application value in the field of remote sensing communication field, such as small target recognition and multimedia information feature detection and retrieval.
【作者单位】：钦州学院资环学院;钦州学院人文学院;广西科技大学软件学院;
【基金】：广西高校社科项目(SK13LX443)
【分类号】：TP751

【参考文献】