基于FPGA的自适应阈值运动目标检测

帧间差分法是运动目标检测最为常用的方法之一，具有算法简单、数据处理量小、易于实现等优点，因而得到广泛应用。而帧间差分法通常是根据单一的光照条件设定的固定阈值以判断是否有运动物体，因此限制了其在不同光照条件下的应用。针对上述情况，以色度、饱和度、亮度(HSL)颜色空间中的亮度值为依据，建立光强-亮度-阈值表，从而以动态的阈值调整拓展了帧间差分法在不同光照条件下的应用。通过仿真和FPGA实现了该目标检测系统，并搭建了验证测试系统。实验结果表明，当光照强度为90 W/m2时，在3.0 m/s和0.5 m/s速度下，自适应阈值的检测正确率分别达到93%和95%，与固定阈值设定较大时的检测结果相近；在0.59 W/m2的低光照条件下针对上述运动速度，检测率分别达到84%和92%，与小固定阈值检测结果相近。

Moving object detection with adaptive threshold based on FPGA

Frame difference is one of the most widely used methods in moving object detection, for its advantages such as simple algorithm, less data to be processed, easy to be implemented and so on; however, frame difference method usually detects a moving object by using a specific threshold value measured under an invariable lightness condition, which limits its application in variable illumination situations. According to the above problem, the illumination-lightness-threshold table was created based on the lightness in the hue, saturation, lightness (HSL) color space to dynamically adjust the threshold value for different illumination conditions. After simulation, a detection system was implemented by field-programmable gate array (FPGA), and an experimental setup was built to verify the system. Experimental results show that, the recognition rate reaches 93% and 95% for a moving target with 3.0 m/s and 0.5 m/s velocity under 90 W/m2 illumination by the adaptive method, which is similar to the method with a large fixed threshold; under 0.59 W/m2 illumination, the recognition rate reaches 84% and 92% respectively for the two moving speeds above, which is similar to the fixed threshold method with a small threshold value.