Highway Visibility Detecting Technology Based On Video Images

Low visibility is an important cause miducing traffic accidents. Especially on the freeway, traffic accidents are prone to be piled up under extremely adverse weather conditions. Therefore, accurate detection of low visibility and warning in advance has great practical significance to ensure safe and efficient operation of highway. Along with the well-rounded development of video images processing and the improvement of video monitoring system of the expressway, cameras have become the main traffic monitoring instruments in current. This paper focuses on the problem of highway visibility detecting by using video images, and develops the visibility detection system to meet the needs of different sections. The specific research contents are as follows:(1) In order to meet the special requirements of major fog areas of the expressway where require a higher detection accuracy, an improved dual differential luminance algorithm is put forward. Aiming at the problem that some parameters are difficult to be quantified in the prime formula, this paper proposes a kind of optimization algorithm by using the modified coefficient method. Meanwhile, combined with the traffic emergency control measures under different visibility, ten detection modes are subdivided to improve the detection accuracy. In addition, the transition conditions of the convergence between the day and night algorithm are studied in two aspects: pixel-value difference and grayscale average of the object regions.(2) An experimental system is built according to the detection principle of dual differential luminance algorithm, and a set of visibility detection dedicated system is developed based on C#. Key functional modules, such as gray level extraction of object regions, detection mode selection, visibility calculation, are realized; it verifies the validity of the optimization algorithm. The test accuracy conforms to the relevant standards and satisfies the special requirements of the highway visibility detection.(3) To meet general needs of common fog areas by making full use of the existing video monitoring system, dark channel prior optimization method is applied to the field of highway visibility detection for the first time. The paper utilizes region growing and modified coefficient method to eliminate the deviation caused by the assumption of dark channel in the image patch. Then, by using multi-mode classification method in different visibility conditions, it reduces errors when extracting atmospheric optical brightness. And then, guided image filtering is adopted to eliminate the blocky effect, achieving accurate extraction of target transmittance. According to the physical relationship between transmittance and extinction coefficient, accurate visibility values are obtained.(4) Based on the existing video monitoring system, a set of general system for highway visibility detection is developed with the C# language. The system achieves the key functions, for example, video images processing, transmittance optimization and visibility calculation. The effectiveness of the dark channel prior optimization algorithm is verified by the video images of the experimental system in the scientific research base and typical sections. The test results are consistent with the actual situation, and meet the general requirements of the highway visibility detection.