Infrared Planar Radar Timing Filter Change Detection

Change detection,widely used in intelligent monitoring area,can recognize and judge the behavior of moving objects in the scene,as well as generate alarms automatically under appropriate conditions.However,the existing video-based change detection method is not suitable for high density and complex situations,illumination variations and shadows in the scene lead to errors in change detection.Therefore,a change detection method from a perspective of time series analysis is proposed in this paper,using infrared rotary plane radar to detect and convert the distance of the shot object in order to generate depth information for change detection is feasible.In addition,to detect scene changes is in the form of "touch",whi ch can realize all-weather and all-round monitoring.When detecting,only one cross-section is generated by transverse cutting of the scene,which can also simplify the change detection.The time series data collected by infrared rotary plane radar is a type of periodic data.In consistent with the basic principle of conventional change detection based on video stream,for change detection of periodic time series data,generally,the period should be estimated first,then the background model of time series is established,and the abnormal degree is calculated subsequently,finally the change points decision is made,and t he algorithm is based on the most classical Gauss mixture background subtraction method.Firstly,the time series data are preprocessed,such as interpolation and filtering.Then,the background model is built using the improved Mixture Gauss Model with the background being updated in real time,and time series change detection is followed.In the end,the density adaptive K_Means algorithm is used to cluster the change points,and the least squares linear fitting is applied to each cluster to make the change dynamic more intuitive.The experiment discussed the effect of time series filtering change detection,and compared it with the existing Mixture Gauss model and Vibe background subtraction method.In the process of experimental analysis,the performance of each algorithm is evaluated by the indexes of Precision,Recall and F.As the experimental results show,the proposed algorithm has some advantages in the adaptability of background and the accuracy of change detection,and it can be concluded that the method meets the requirements of real-time change detection by calculating the delay rate.