Road-aware System Based On Multi-camera And Multi-parameters Fusion

With the continuous development of social progress and transport, the importance of traffic safety becomes higher and higher which is always threatened by bad weather conditions, so the requirements for the monitoring of bad weather conditions along the road becomes more and more important. However, ice, especially dark ice, which is the most dangerous situation to traffic safety, can’t be detected effectively yet, which comes difficult on traffic monitoring.Nowadays monitoring for weather conditions along roads covers a small area and is quite expensive, this system detects the parameters on roads with multi-eye vision fusion technology, double-eye polarization image fusion for water and ice on roads, infra/visible light image integration for visible temperature parameters on image, temperature/double-eye result fusion for the precise division of the road state, results in wide-range, low-cost, high reliability online detection of weather condition on roads, and with some information appeared on visible light image, the situation of roads is intuitive and the detection can be automated operated. The main research content and contributions are as follows:A new method to detect road surface meteorological parameters by polarization image fusion and is designed, so is correlated algorithm and a system program.The theory and workflow of infrared temperature measurement are studied, automatic calibration and temperature measurement of the infrared camera are realized, so is the image fusion of infrared and visible image.The binocular3D vision technology is studied to realize calibration of cameras, distortion correction of images, dual-stereo calibration and3D reconstruction, along with which we construct camera models and derivate a lot of formulas. As a result we get the depth information which help the remote temperature detection of infra camera, and the basic matrix which help the binocular polarization image fusion.At last complete software and hardware constructions are designed, we can realize polarization images and the infrared image acquisition, the binocular image fusion, visible/infrared image fusion and decision level fusion processing.