基于邻域特征的红外低慢小目标检测

Infrared low, slow and small target detection based on neighborhood characteristics

红外搜索系统具有不发射电磁波、抗电磁干扰能力强、目标指示精度高等优势,在低慢小目标探测领域有很好的应用前景。目前国内外基于红外搜索系统的目标检测算法通常利用当前图像与背景图像配准、差分的手段提取疑似目标,往往需要较大的存储空间保存周视背景图像,高精度实时图像配准算法的工程化应用也有较大难度。针对以上问题,设计了一种基于邻域特征的红外低慢小目标检测方法。通过高通滤波、边缘检测法提取疑似目标、目标邻域特征值统计法剔除背景干扰、多帧图像目标信息相关等处理过程,可在有效排除地物、云层及飞鸟等多种干扰的同时,准确地检测图像中的无人机目标。试验结果表明,该方法相比传统LCM算法目标检测概率更高、虚警率更低,且不涉及图像差分,具有对硬件资源要求低、实时性好等优点,有较高的工程应用价值。

The infrared search system has the advantages of not emitting electromagnetic wave,strong anti-electromagnetic interference ability and high precision of target indication,and has a good application prospect in the field of low,slow and small target detection.At present at home and abroad,the target detection algorithm of infrared search system usually extracts the suspected target by using the current image and the background image registration and difference,which is often necessary to save the panoramic background images in the large storage space,and the engineering application of high precision real-time image registration algorithm is also difficult.For the above questions,a infrared low,slow and small target detection method was designed for the infrared search system.Through the processes of morphological filtering,extracting suspected targets with edge detection method,eliminating background interference with target neighborhood eigenvalue statistical method and correlating targetinformation based on multi-frame image,it can effectively eliminate the interference of ground objects,clouds and birds,while accurately detecting the UAV′s target in the image.Experimental results show that this method has higher target detection probability and lower false alarm rate than the traditional LCM algorithm.Moreover,it does not involve image difference,and has the advantages of low demand for hardware resources and good real-time performance,etc.,and has high engineering application value.