一种基于改进剩余谱的红外舰船目标显著性检测方法

Infrared Ship Detection Technology Based on Improved Residual SpectralSignificance Detection

为了解决传统显著性检测方法在红外舰船目标检测中精度不高的问题,提出了一种基于改进剩余谱的显著性检测方法。该方法首先利用自适应六边形中值滤波和小波图像融合对红外图像去除噪声;然后在剩余谱显著性检测算法的基础上,引入超像素分割处理舰船目标区域以减少计算量,同时保留舰船目标的边缘信息;此外,引入目标掩码排除背景干扰以提高算法的准确性;最后,通过提取图像的显著图实现舰船目标的检测。实验证明,该方法相较于改进之前检测舰船目标的准确性提高了约3%,在岛屿等复杂背景下能够准确检测出舰船目标。

In order to solve the problem that the traditional significance detection method is not accurate in infrared ship target detection,a significance detection method based on improved residual spectrum is proposed.Firstly,adaptive hexagon median filtering and wavelet image fusion are used to remove noise from infrared images.And then,based on the residual spectral significance detection algorithm,superpixel segmentation is introduced to process the target area of the ship to reduce the amount of calculation,while retaining the edge information of the ship target.In addition,the target mask is introduced to eliminate background interference to improve the accuracy of the algorithm.Finally,the detection of ship targets is realized by extracting a salient map of the image.Experiments show that this approach can achieve an accuracy about 3%higher than that of the previous,and therefore can help to accurately detect ship targets in complex backgrounds such as islands.