Target behavior pattern online classification method based on inductive consistency multiclass classification

Abstract

The invention discloses a target behavior pattern online classification method based on inductive consistency multiclass classification. The method comprises steps: 1, related variables are defined; 2, initialization is carried out; 3, a training track data set and a compressing track data set of one class are used to calculate the p value of the current track point in the tested track; 4, the step 3 is repeated to calculate the p value of the current track point corresponding to other classes; 5, the sizes of the p values corresponding to all classes are compared to determine a target behavior pattern class corresponding to the current track point; 6, after the target behavior pattern corresponding to each track point on the current tested track is classified, the training track data set is updated; 7, a multi-factor oriented Hausdorff distance matrix is updated; and 8, target behavior patterns corresponding to a next tested track are classified. The method can realize online classification and discrimination on the target behavior patterns in a condition with a reduced calculation amount, and the application prospect is wide.

Description

technical field [0001] The invention relates to an online classification technology in data mining and a high-level fusion technology in information fusion, and belongs to the field of pattern recognition and intelligent information processing. Background technique [0002] In the field of early warning and surveillance, with the continuous improvement of target detection technology and information fusion technology, various targets are detected, tracked and identified, forming a constantly updated target track. A large amount of historical track data is stored and accumulated in various target intelligence processing systems in the field of early warning and surveillance. Using data mining and cluster analysis technology in trajectory data mining, the target track can be divided into different categories, so as to dig out the behavior rules of the target. The behavior pattern of the target refers to the category of the target behavior law that the current observation targe...

AI Technical Summary

Problems solved by technology

Many scholars at home and abroad have studied the problem of trajectory classification, but these methods mainly consider the position and shape characteristics of the target, and do not make full use of the position, speed and heading characteristics of the target, and the existing methods are mainly used for offline classification. It is not applicable to the field of early warning and monitoring with high real-time requirements

The online classification method of target behavior patterns based on multi-dimensional features can make full use of the target's position, speed and heading characteristics, and realize the online classification and discrimination of target behavior patterns through online learning and sequential classification of multi-dimensional track data. All historical track data need to be recalculated during classification and discrimination, and there is a problem of large amount of calculation

Images