Research On Detection Algorithm Of Criminal Abnormal Trajectory Based On Pedestrian Moving Position Data

With the wide popularity of the mobile terminal,location information has become the important basis of the public security organ to crack down on all kinds of crime.At present,there are many researches based on trajectory data,but there are relatively few researches on abnormal trajectory detection of suspect.In the era of big data,how to detect abnormal behavior of suspect in physical space by using trajectory data has become a difficult point in the field of public safety.To solve this problem,this paper proposes a suspect anomaly detection algorithm based on machine learning.First of all,the suspect trajectory pattern is defined by the experience of life and police service;Then,according to the characteristics of mobile terminal location data,the pedestrian trajectory preprocessing method and the method of trajectory stop area division are designed;Finally,the abnormal model detection method is designed according to the characteristics of the crime anomaly.Specifically,the research content of this paper mainly includes the following three aspects:(1)The trajectory data pretreatment algorithm based on heuristic for GPS trajectory.Aiming at the problems of instability,data redundancy and outliers in GPS trajectory data collection,this paper proposed a pre-processing method of denoising,filtering and deletions of the original data step by step in order to obtain trajectory data with practical significance.The experimental results show that this method has a good cleaning effect and retains the user’s activity information effectively.(2)Algorithm of trajectory stay region partition based on density clustering.In order to distinguish different stay areas in the trajectory data,the partitioning algorithm was divided into three stages for processing,and the data is gradually simplified and the process mapping table was established.By improving the grower and noise cluster conditions of the density clustering algorithm,the repeated data in the same staying area of the trajectory can only occupy the weight once in the growth process,so as to avoid the excessive length of a few tracks leading to the weight increase and the excessive extension of the clustering range,and then divide the region more finely.The experimental results show that the method presented in this paper is more intuitive and accurate than the direct use of density clustering for the same stop point d(3)Abnormal trajectory detection algorithm based on the longest common subsequence.Abnormal trajectory mining needs to combine period information.Since a single region contains little period information,it needs to combine period information on a long-time scale.On the basis of dividing the stay area of the trajectory,the trajectory information was converted into a time series with the stay area information.After dividing the sequences,the parts with abnormal activity frequency were screened out by comparing the similarity between the sequences.Finally,whether the part contains a specific place type was verified.This method effectively improved the efficiency of anomaly detection,reduced the amount of data in the process of operation,and finally realized the purpose of criminal anomaly detection.