融合分数阶微分边缘特征的自适应跟踪

为提高跟踪方法对背景信息、光照变化的抗干扰能力，提出融合分数阶微分边缘特征信息的改进跟踪方法。将R-L分数阶微分边缘检测算子与Laplacian边缘检测算子进行融合，构造出对高、中频信息提升，而对低频信息能够非线性保留的混合边缘检测算子，并利用其实现目标模板及场景图像的边缘特征信息检测。基于目标色度特征和边缘特征两种直方图模型，分别建立场景图像的反向概率投影值，根据背景信息的动态变化，以抑制背景干扰信息为目的，建立自适应融合的反向概率投影图，提高算法对不确定环境变化信息的鲁棒性。实验结果表明，混合边缘检测算子能够提高边缘图像的信噪比，改善边缘提取的效果。边缘、色度特征信息自适应融合的跟踪方法单帧跟踪时间小于20 ms，满足实时性要求。该方法能够在光照变化明显和与目标颜色相似背景等情况下有效实现目标识别与跟踪功能。

Adaptive tracking method with fractional differential edge feature

To enhance the anti-jamming ability with respect to the background and the illumination variation, an improved tracking method based on the fractional differential edge feature was proposed. The R-L fractional differential edge detection operator and Laplacian edge detection operator were fused to construct a hybrid edge detection operator that could enhance the high and middle frequency information and hold the low frequency information nonlinearly. The hybrid edge detection operator was used to detect the edge information of the target template and the tracking scene. The hue histogram and edge histogram of the target were separately modeled. According to the dynamic information of the background, the back-projection image could be modeled by adaptively fusing the back-projection probability values, which can be separately obtained based on the two histograms to suppress the interfering information from the background. Experiments show that the hybrid edge detection operator can enhance the signal-to-noise ratio of the edge image and improve the quality of the edge image. Furthermore, the tracking time of the tracking method based on the adaptive fusion of the edge feature and hue feature is less than 20 ms, which can meet the real-time requirements of tracking systems. The tracking method can be applied to effectively track the target in an environment with varying illumination in a background similar to the target.