一种适用于情报图像处理的图像集分类算法

无人机情报处理系统是无人机地面控制系统的重要组成部分之一，主要负责对无人机侦察载荷下传的侦察情报进行处理，从复杂的情报中获得直观的情报产品并传递给上级和友邻单位。对于搭载光电载荷的无人机情报处理当前仍以依靠人力鉴别为主。介绍一种基于快速近似最近邻( FLANN) 搜索特征的K 近邻用分类决策，可去除背景信息对分类性能的影响；为了进一步提高算法的运行速度及减少算法的内存开销，采用特征选择的方式分别减少测试图像和训练图像集的特征数目，并尝试同时减少测试图像和训练图像集中的特征数目平衡分类正确率与分类时间之间的矛盾。该算法保留了原始NBNN 算法的优点，无需参数学习的过程，实验结果验证了算法的正确性和有效性。

An Image Set Classification Algorithm for Inteligence Image Processing

UAV intelligence processing system is one of the important parts of UAV ground control system, and it is mainly responsible for UAV reconnaissance in the load transfer. Reconnaissance intelligence processing and intuitive intelligence products are collected from complex intelligence and passed to the superior and friendly units.UAV with photoelectric payload still relies mainly on human identification for intelligence processing. A K nearest neighbor classification decision based on fast approximate nearest neighbor (FLANN) search feature is introduced which is able to remove the influence of background information on classification performance. In order to further improve the running speed of the algorithm and to reduce the memory overhead of the algorithm, the feature number of the test images and the contradiction between the classification accuracy as well as the classification time of the feature number balance between the test image and the training image set are tested at the same time. The algorithm preserves the advantages of the original NBNN algorithm and it does not need the process of parameter learning. The experimental results verify the correctness and effectiveness of the algorithm.