Recognition of 3D compressed images and its traffic monitoring applications

Abstract

The need for improved security requires the development of better monitoring systems. These systems must manage large numbers of images sent over communication networks. For example, a high transmission load results when security cameras continuously transmit images. Operators are bored and fatigued from viewing cycled images on a few monitors. In this thesis, we present a smart camera network that reduces transmission load and provides operators with relevant images. A smart camera is capable of compressing and processing images. Our 3D compression method (based on Mizuki's 2D method) retains features (contour, color, depth, and motion) that form the basis of object detection. The compressed image is used directly in segmentation. To reduce transmission loads, we use mobile agents to filter the network. Mobile agents are dispatched to smart cameras with user defined goals that conditionally allow image transmission. For traffic monitoring, smart cameras determine traffic flow and accidents. Mobile agents send images or information based on image content. The contribution of this work is the 3D compression method designed for processing compressed image data and the design of a system to improve camera centered networks using smart cameras and mobile agents. Our system can be used in security systems, where image content can promote an efficient transmission of images.