Traffic Video Classification using edge detection techniques

Classification of Videos based on their content is becoming more and more essential everyday because of the vast amount of video data becoming available. Various Feature Extraction and data mining techniques can be used to perform Video Classification. This paper uses edge detection techniques such as Object Extraction and Canny Edge Detection (using Sobel, Prewitt and Robert's operator) to extract features from the key frames. After extraction, the features are pre-processed using Discretization, PKIDiscretization, Fuzzification, Binarization, Normalization techniques and analysed using Correlation Feature Selection technique before being used by Naive Bayesian Classifier for training and testing purpose. The experimental results show a high accuracy of classification for a set of traffic surveillance videos can be achieved with the proposed combination.