基于图像统计学习进行昼夜区分的算法研究

通过对实际路口昼夜路面状况的分析,阐述了在视频电子警察系统中进行昼夜区分以提高车辆检测精度的必要性,分析了常见的昼夜区分方法的优缺点;在此基础上,提出了一种利用路面灰度特征,对视频图像统计学习进行昼夜区分的算法。该算法通过对实际路面的平均灰度值随时间变化趋势进行了统计学习,寻找出随天气季节变化的昼夜区分时间。该算法应用于相关电子警察产品之中,取得大量工程实测图像数据,验证了此算法的可靠性、高效性以及通用性,对提高视频电子警察产品的性能有一定的实用价值。

Study on Algorithm of Distinguishing Day and Night with Image Statistics

The technique of distinguishing day and night is of importance in enhancing the detecting precision of the Video Electronic Police System.The advantage and disadvantage of common ways to distinguish day and night in electronic police equipment is analyzed and a study on this technique is carried out.An algorithm of distinguishing day and night with image statistics is proposed based on the analysis of the characteristics of the grey level of actual road surface.The time of dividing day and night in different conditions of weather and season could be calculated by the statistics of the grey level of actual road surface.The algorithm has been demonstrated in a traffic project and in cluded in correlative electronic police equipment,using the data collected by electronic police equipment,having proved to be reliable,high efficient and applicable.It will be of certain value in improving the performance of Video Electronic Police equipments.