Mining Sophisticated Patterns and Actionable Knowledge from Massive Moving Object Data

Abstract

Video surveillance system especially for humans and vehicles, is one of the current challenging research topics in computer vision. The proposed system detects and recognizes objects by extracting object features with different activities in every frame in an images sequence. This paper presents the proposed detection and recognition algorithms. The first step is applying  proposed detection algorithm to detect the foreground pixels using the foreground detection model by using multi. The features extracted from each detected object is made by using color, texture,  size and shape information in addition to statistical features to facilitate classifying objects by using the combination of three classification techniques (Support Vector Machine, Linear regression and M5 rules). The last step in the proposed system is  using the extracted features for object recognition. The proposed system gives acceptable results and can be applied easily by the end user; it enables the user to interact with multiple video scenes and frames. The proposed method is implemented and tested on a wide variety of challenging and adapts the model in background subtraction, classification, the detection process and recognition of objects. Due to many experiments in proposed algorithm, the algorithm is achieves high performance compares with the traditional algorithms of detection and recognition of objects.