Research On Crowd Abnormal Behavior Detection Algorithm

Crowd abnormal behavior detection is the use of CCTV cameras installed in public places to capture and detect abnormal events such as brawls,traffic accidents,and stomping,and then issue timely warnings.It has a wide range of applications in the field of intelligent monitoring.In recent years,the crowd abnormal behavior detection method has achieved good development,and many excellent detection algorithms have been proposed.However,many challenging problems have not been effectively solved,such as changes in lighting,crowd occlusion,and complex backgrounds,leading to greatly reduced accuracy and robustness of crowd abnormal behavior detection algorithms.Therefore,detecting abnormal behavior of the crowd is still a difficult task.In order to improve the accuracy and robustness of crowd abnormal behavior detection algorithms,based on spatial and motion feature representations and the use of deep learning tools,the main research work and innovations are as follows:(1)Aiming at the problem of low detection performance of crowd abnormal behavior conducted by complex background and occlusion as well as other factors,a synthetic optical flow feature descriptor(SOFD)and trajectory based method is developed for crowd abnormal behavior detection.The velocity,acceleration,direction and energy of crowd motion can be firstly calculated by the developed approach according to the change of crowd optical flow field,and then a new space-time feature descriptor,i.e.,SOFD,can be constructed based on the above characteristics.In what follows,the KLT(Kanade-Lucas-Tomasi)tracking algorithm is employed to obtain the single frame of crowd motion trajectory.Finally,the two stream convolution neural network(TS-CNN)is depicted with the above mentioned characteristics to detect the crowd abnormal behavior.The simulation results show that the proposed algorithm can significantly improve the accuracy and robustness of anomaly behavior detection in real complex scenarios compared with the existing mainstream algorithms.(2)Aiming at the problem of low performance of crowd abnormal behavior detection caused by complex backgrounds and occlusions,this paper proposes a method of crowd abnormal behavior detection based on inter-frame characteristics.The proposed method first calculates the instantaneous weighted energy of crowd motion based on changes in crowd optical flow fields.The acceleration value and the direction difference optical flow value are respectively mapped to the hue value,saturation and value of the HSV to form an inter-framesaliency map representing the inter-frame motion characteristics of the crowd;finally,based on the inter-frame saliency map and the single-frame original image,design TS-CNN are used to detect crowd abnormal behavior.The simulation results show that the proposed algorithm can significantly improve the accuracy and robustness of anomaly behavior detection in real complex scenarios compared with the existing mainstream algorithms.