用于成像探测目标识别的块特征提取方法

Block feature extraction method for target detection in imaging fuze

针对成像引信图像处理的特殊性,研究了目标图像块特征提取方法。该方法是由细化算法改进而来。利用图像的连通性和形体变化趋势将图像分成几个基本块。每块有4个元素,即长度特征、合特征、分特征、锁定标志。提取各块上出现的形体特征及其与其他块之间的拓扑关系作为图像的特征,设计了表示块结构和特征的数据结构以便于系统处理。该算法便于硬件实现,从而保证了图像识别的实时性。实验结果表明:该方法可以有效实现目标识别。

According to the special need of image processing in imaging fuze,a block feature extraction method improved from thinning algorithm was developed.An image can be divided into several blocks according to their connectivity and changing trends.Each block has four elements,namely its length,converge tendency,diverge tendency and lock flag of the block.The structure features of each block and the relationship with other blocks were extracted.The data structure which can express the blocks′ shape and characteristics was designed for system processing.The algorithm was easily implemented in hardware to ensure real-time image recognition.Experimental results confirm that the algorithm can accomplish target recognition effectively.