基于机器视觉的行人异常行为检测方法

Abnormal Pedestrian Behavior Detection Method Based on Machine Vision

常规的行人异常行为检测方法使用编码-解码器进行记忆寻址,易受到记忆大小和稀疏度的影响,导致帧级AUC偏低,因此基于机器视觉设计一种全新的行人异常行为检测方法。结合检测图像的初始状态处理噪声、增加平滑度,提高行人异常行为检测性能,再利用机器视觉技术量化动态参数,输出异常行为轨迹特征,实现行人异常行为检测。实验结果表明,在不同场景下,本文设计方法的检测帧级AUC始终较高,获取的检测结果较清晰,说明本文设计方法具有一定的应用价值。

Conventional pedestrian abnormal behavior detection method uses encoding-decoder for memory addressing,which is easy to be affected by memory size and sparsity,resulting in low frame-level AUC.Therefore,this paper designs a new pedestrian abnormal behavior detection method based on machine vision.Combined with the initial state of the detection image,the noise is processed and the smoothness is increased to improve the abnormal behavior detection performance.The machine vision technology is used to quantify dynamic parameters,output abnormal behavior track features,and realize pedestrian abnormal behavior detection.Experimental results show that under different scenarios,the detection frame-level AUC of the design method in this paper is always higher and the detection results obtained are clear,which indicates that the design method in this paper has certain application value.