Research On Key Technologies Of Distributed Panoramic Color Imaging And Warning System

The panoramic imaging and warning systems based on visible light sensors, infrared sensors or other kinds of photoelectric sensors are proposed to solve the problem that detecting low altitude objects(such as sea skimming missiles, low flying aircrafts) is difficult for the radio detection and ranging systems. This kind of systems can help operators observe the surrounding area, and warn about the suspect objects, which makes them be a supplement of the radio detection and ranging systems and play an important role in defense of asymmetric threats. The feasibility study of optoelectronic surveillance and warning equipments started in 1970 s, and this kind of systems has been developed for over 40 years. According to the methods of acquiring the panoramic images, the panoramic situation awareness and warning systems can be divided into 3 categories:(1) the systems using sensors with narrow field of view and precise scanning mechanisms to acquire the panoramic images;(2) the systems using one single sensor to acquire the panoramic images directly;(3) the systems using several distributed sensors to acquire the panoramic images. With the development of photoelectric sensor technology, we can get the sensors with better performance and lower prices. The systems of the third category can acquire the panoramic images with high spatial resolution in real time, which makes them occupy the position in the mainstream of practical applications.The key theories and algorithms of a panoramic color imaging and warning system based on distributed photoelectric sensors are researched in this dissertation. This kind of systems can be applied in various kinds of scenarios which require being monitored. They can provide panoramic images visually, and offer the functions of object detection, tracking, and warning. Compared with current systems, the system using the technologies proposed in this dissertation is easier to be constructed, provides more natural panoramic images and uses more efficient algorithms in object detection and tracking.The method for obtaining the panoramic images, the algorithms of object detection, tracking, and warning are described from Section 2 to Section 4 in this dissertation according to the process flow in our system. The experiments for performance tests of each algorithm, and a prototype device with basic functions based on the methods discussed in this dissertation which demonstrates that our research is operable and easy to use are presented in Section 5. The proposed method does not require that all the cameras' optical centers coincide with each other while the conventional panoramic image acquisition with multi-sensors method does. We simplify Zhang's flexible technique combining the actual needs for camera calibration, and project all the images captured by each sensor to a cylindrical model to obtain the panoramic images. A novel color transfer algorithm is introduced in the dissertation for color normalization to make the color of stitched images more natural. We improve an object detection approach, and make it execute in parallel for object detection, so that the high resolution images can be processed faster. An object tracking method which uses the spatial and temporal information of the object simultaneously for tracking is proposed in the dissertation to achieve a steady tracking for the given object. Several implementation schemes for warning that can be applied in different scenarios are described in the dissertation referring to the knowledge of the intelligent video surveillance.A comprehensive conclusion is drawn in the end of this dissertation. With the knowledge of the panoramic situation awareness and warning system based on distributed photoelectric sensors gained during the study, we prospect the future of this system.