提出了一种目标分割预检测方法来提高检测红外弱小目标的准确性和实时性。针对红外图像的特点, 利用改进的自适应背景感知算法抑制目标图像的背景以提高目标检测概率; 根据已有的先验知识构造属性集, 把灰度直方图限定在感兴趣区域, 减少背景的影响; 然后, 利用属性直方图的最大熵进行图像分割以检测目标。为了提高分割算法运算速度, 应用了快速递推算法。实验结果表明, 本文提出的背景抑制算法能更好地抑制背景, 提高图像的整体信噪比; 分割算法具有更好的分割检测效果, 候选目标点分割准确、虚警目标点较少, 运算速度提高了91%。对分割图像进行后续处理, 剔除了大部分虚警目标点, 为后续目标准确检测提供了有力保障。

A segmentation and detection method for small infrared targets is proposed to improve the accuracy of target detection. Aiming at the characters of an infrared image, an improved background perception algorithm is used to suppress backgrounds for increasing the target detection probability, and the bound set is constructed to limit the gray level histogram into a Region of Interest (ROI) to reduce the interference of background. Then, target detection is achieved through image segmentation using the maximum entropy of a 2D bound histogram. Furthermore, the fast recurring algorithm is applied to the proposed algorithm for accelerating the running speed of the segmentation algorithm. Experiment results show that proposed background suppression algorithm has the better performance in background suppression and can improve the signal to Noise Ratio (SNR) of the image. Moreover, the segmentation algorithm shows a better effectiveness in target segmentation and detection, and its candidate target is separated more accurate with less false alarm points and running speed has improved by 91%. Through post-process for the image, most of the false alarm points are eliminated, which provides powerful guarantee for the subsequent accurate detection.