基于视觉对比度机制的红外双极性小目标检测方法

Detection Method of Infrared bi-Polar Small Targets Based on Visual Contrast Mechanism

红外小目标在实际场景中常呈现"双极性"的特点。局部对比度检测算法,如(Local Contrast Method, LCM)可以检测亮目标,但在检测灰暗目标方面适应性不足。在LCM算法基础上研究了一种利用双层滑动窗口结构计算局部对比度的检测算法。中值滤波器预处理图像后,采用双层滑动窗口遍历整幅图像;依据双重局部对比度计算获得图像的显著图,可以在一次图像遍历中检测出不同尺度的小目标;最后,采用自适应阈值分割方法获得目标位置。经实验验证,能克服LCM算法不能同时检测双极性目标的不足,在检测双极性目标时降低了40%的处理时长,且在相同虚警率条件下,正确检测概率提升了12.4%。

Infrared small targets often show the characteristics of"bipolarity"in the actual scene. Local contrast detection algorithms, such as(Local Contrast Method, LCM)algorithm, can detect bright targets, but have insufficient adaptability to detect dark targets. On the basis of LCM algorithm, a detection algorithm that uses double-layer sliding window structure to calculate local contrast was studied. After processing the image by median filter, a double-layer sliding window was used to traversal the whole image. The salient image of the image was obtained by calculating the double-local contrast, it can detect different scales of small targets in a single traversal of the image. Finally, an adaptive threshold segmentation method was used to obtain the target position. Experimental verification shows that it can overcome the shortcoming that LCM algorithm cannot simultaneously detect bipolar targets. When detecting bipolar targets, the processing time was reduced by 40% compared with LCM algorithm, and the probability of correct detection was increased by 12.4% under the same false alarm rate.