원문정보
초록
영어
Intelligent analysis is the development direction of the research for Geo-Video. The key is how to extract automatically the implicit spatial-temporal correlation between in the various targets. For this purpose, this paper presents a dynamic graphs framework to simulate the complex moving target’s behavior patterns in Geo-Video and uses this model to detect abnormal events. Firstly, compute the moving target’s orientation matrix by Optic Flows and use the window clustering algorithm to classify different targets; and then, creating a dynamic graph whose vertex is the cluster center and the weight of edge is the Euclidean distances between two centers; finally, simulate different targets by defining the evolution rule of dynamic graphs in specific geographical scene. Experiments show that the proposed method can directly express the moving target’s behavior patterns and detect abnormal events in Geo-Video. In addition, this proposed method provides a new way of thinking about the intelligent analysis for Geo-Video.
목차
1. Introduction
2. Related Works
2.1. Target Behavior Analysis
2.2. Abnormal Detection
3. Preprocessing
3.1. Correction in the x Direction
3.2. Correction in the y Direction
4. Targets Clustering based on Directions
4.1. Calculate Speed Matrix
4.2. Calculate Orientation Matrix
4.3. Direction classification
4.4. Window-based Clustering
5. Generate Dynamic Graph
5.1. Vertex
5.2. Edge
5.3. Evolution Rules
6. Experiment and Analysis
6.1. Use Dynamic Graph Method to Detect Anomalies
6.2. Experimental Comparison
7. Conclusions
References