基于自适应阈值与双特征的ViBe运动车辆检测算法

ViBe Moving Vehicle Detection Algorithm Based on Adaptive Threshold and Dual-feature

ViBe算法在运动车辆检测中容易产生鬼影和空洞现象,为此,提出一种融合双特征建模和自适应阈值的背景减除算法。用尺度不变局部三值模式特征和颜色特征对首帧视频序列构建背景模型后,结合2种特征的变化改进ViBe算法的判别方法。当某像素被判别为前景时,使用Otsu算法计算图像的最佳分割阈值并根据该阈值对前景像素点进行第二次判断,采用改进的ViBe更新策略更新背景模型。实验结果表明,相对ViBe算法,该算法可以抑制鬼影和空洞现象,提高运动车辆检测的精确度,且在光照变化时具有较好的鲁棒性。

ViBe algorithm is easy to produce ghost and void phenomena in moving vehicle detection.To solve this problem,a background subtraction algorithm based on dual-feature modeling and adaptive threshold is proposed.Firstly,the background model of the first video sequence is constructed by Scale Invariant Local Ternary Pattern(SILTP)feature and color feature,and then the discriminant method of ViBe algorithm is improved by combining the changes of the two features.When a pixel is identified as a foreground,Otsu algorithm is used to calculate the optimal segmentation threshold of the image and the foreground pixels are judged for the second time according to the threshold.Finally,the improved ViBe update strategy is used to update the background model.Experimental results show that,compared with ViBe algorithm,the proposed algorithm can suppress ghost and void phenomena,improve the accuracy of moving vehicle detection,and has better robustness when illumination changes.