遗传算法在运动目标检测中的应用

Application of Genetic Algorithms in Motion Object Detection

运动目标的检测在智能交通监控、车辆导航、辅助驾驭、高速公路上的自动收费等都有广泛的应用。运动目标的正确检测十分困难,当观测采样数据受到异常污染时,检测的性能对模型的微小变化不敏感。这就要求系统检测具有在其特性或参数发生摄动时仍可使品质指标保持不变的性能即鲁棒性。该文提出了在动态图像序列中基于Otsu法和遗传算法相结合的方法动态确定差分图像二值化的阈值以代替人工确定阈值的方法,从而鲁棒地检测出运动目标。

Motion object detection is widely used in the areas of smart traffic monitoring,vehicle navigation,assistant driving and automatic toll collection on freeways.Satisfactory result of motion object detection is hard to achieve due to the lack of sensitivity of detected capability to minor changes of the modal when the sampled data are polluted.Thus,the quality index stability is required for system detection even when its characteristics and parameters undergo changes,which can be realized through robustness.This paper uses an adaptive thresholding method combined Otsu method with genetic algorithms to automatically choose the threshold value instead of determining the threshold value manually,therefore,motion object are detected robustly.