Search

This paper presents a three phase approach to crime prediction based on video analysis, neuro-fuzzy inference and density mapping. In the first phase, crime indicator concepts are modeled and used in building classifiers for crime indicator events. Both indicator concept modeling and indicator event classification are performed using Generalized Maximum Clique Problem (GMCP) method. In the second phase, a neuro-fuzzy inference system modeled from training data is used to make predictions about classified crime indicator events obtained from the first phase. Finally in the third phase, kernel density estimation (KDE) is used to fit a spatial probability density function to the predicted crime indicator events across the study area. The major advantages of this method include the potential to predict crime in real time due to the use of video based events, the ability to generate fuzzy rules from data, the ability to optimize fuzzy rule-base by learning and the ability of weighting different crime variables. The proposed framework has prospects for developing a police field decision support system. The feasibility of the framework has been tested in a simulated experiment using sampled clips from violent scene detection (VSD) 2014, Hollywood Human Action (HOHA) and HMDB datasets and the results are quite promising for real life implementation.

This paper presents a new approach to crime hotspot detection and monitoring. The approach consists of three phases’ namely: video analysis, crime prediction and crime mapping. In video analysis, crime indicator events are modelled using statistical distribution of semantic concepts. In crime prediction, a neuro-fuzzy method is used to model indicator events. In crime mapping, kernel density estimation is used to detect crime hotspots. This approach is tested in a simulated platform using violent scene detection (VSD) 2014 dataset.