A Method of Abnormal Trajectory Analysis Based on Adjoint Model

Abstract

The invention discloses an abnormal track analysis method based on an adjoint model, which comprises the following steps: face recognition is performed by a video structuring system, face video data is obtained and preprocessed, and then accompaniment analysis is performed on accompanying persons through a frequent model mining algorithm Mining to obtain accompanying relationship data and accompanying risk coefficient. The accompanying risk coefficient is greater than the risk threshold of the accompanying personnel and recorded as the risk accompanying personnel; through the unsupervised learning algorithm, the accompanying relationship data of the risk accompanying personnel is used to detect the abnormality of the accompanying track, based on the abnormality of the risk accompanying personnel Trajectories use automatic feature engineering algorithms to train supervised learning models and conduct risk-accompanied personnel abnormal trajectory analysis. The present invention is based on the face video structured data, combined with the adjoint analysis model, and on the basis of considering the accompaniment relationship of personnel, analyzes the abnormal trajectory through the abnormal trajectory detection algorithm, and overcomes the difficulty in accuracy and applicability of building a model only from the perspective of time and position. Bad confinement.

Description

technical field [0001] The invention belongs to the technical field of video data processing, and in particular relates to an adjoint model-based abnormal trajectory analysis method. Background technique [0002] The intelligent perception of public security information is based on the crime characteristics and public security situation in a certain time and space, using the theory and method of machine learning under the background of artificial intelligence, through the classification, screening, analysis, and prediction of police information, to identify possible crimes and cause crimes. The various elements of social unrest and their symptoms are closely monitored, their development trends and degree of harm are accurately predicted, warning signs are captured, early warnings are made in a timely manner, and advance prevention is formed to effectively prevent and control the occurrence of crimes and the outbreak of major vicious cases. A set of operating mechanisms. Tra...

AI Technical Summary

Problems solved by technology

[0009] The purpose of the present invention is to: solve the deficiencies in the above-mentioned prior art, provide a kind of abnormal trajectory analysis method based on adjoint model, based on adjoint analysis and abnormal trajectory comprehensive analysis, improve the detection accuracy of abnormal trajectory; in the stage of unsupervised learning abnormal detection, Use FastDTW and improved density clustering technology to reduce system overhead and improve computing performance; use automatic feature engineering technology in the supervised learning stage to solve the difficult problem of trajectory feature selection and improve model analysis efficiency and quality

Images