基于目标区域特性海面小目标检测快速算法.pdfVIP

Vol. 31, No. 3 ACTA AUTOMATICA SINICA May, 2005 Fast Small Offshore Target Detection Based on Object Region Characteristic1) ZOU Yu-Lan WANG Guo-You ZHANG Lei (Institute for Pattern Recognition and Artificial Intelligence, State Education Commission Key Laboratory for Image Processing and Intelligent Control, Huazhong University of Science Technology, Wuhan 430074) (E-mail: snow.zyl@163.com) Abstract A fast object detection method based on object region dissimilarity and 1-D AGADM (one dimensional average gray absolute difference maximum) between object and background is proposed for real-time defection of small offshore targets. Then computational complexity, antinoise performance, the signal-to-noise ratio (SNR) gain between original images and their results as a function of SNR of original images and receiver operating characteristic (ROC) curve are analyzed and compared with those existing methods of small target detection such as two dimensional average gray absolute difference maximum (2-D AGADM), median contrast filter algorithm and multi-level filter algorithm. Experimental results and theoretical analysis have shown that the proposed method has faster speed and more adaptability to small object shape and also yields improved SNR performance. Key words Small target detection, average gray absolute difference maximum, computational complexity, signal-to-noise ratio (SNR), ROC curve 1 Introduction Because the features of multiscale small targets object such as shape, texture, moment, structure, motion continuity and fractals are undeterministic and distinctnes