摘要
受光照条件、图像噪声和复杂背景等因素的影响,在机动车驾驶员检测过程中难以获取不同卡口图像下的驾驶员特征.为了解决上述问题,文中提出基于多识别区域融合的精准驾驶员位置检测框架,用于提高驾驶员识别率.首先基于图像梯度特征算法获得车牌定位,然后使用自适应方法得到车窗区域,最后采用多识别区域融合策略得到准确的驾驶员区域.在10个图像测试库上的测试表明,文中方法可以获得较高的识别率.
Features of motor vehicle driver are difficult to be acquired in motor vehicle driver detection due to the various illumination conditions, image noise and complex backgrounds. Therefore, an accurate driver location detection framework based on multi-detected region fusion is proposed in this paper to promote the driver identification rate. In the detection process, license plate location is obtained firstly by the algorithm based on image gradient features. Then, the window area of the vehicle is determined by an adaptive method. Finally, the multi-detected region fusion strategy is introduced to obtain the accurate driver region. Experiments on the testing image library verify the high identification rate of the proposed algorithm.
作者
霍星
檀结庆
赵峰
景永俊
邵堃
HUO Xing1 , TAN Jieqing2, ZHAO Feng2, JING Yongjun3, SHAO Kun3(1. School of Mathematics, Hefei University of Technology, Hefei 230009 2. Tiansheng Science and Technology Corporation, Hefei 230094 3. School of Computer and Information, Hefei University of Technology, Hefei 23000)
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2018年第3期283-292,共10页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金项目(No.61502136
61572167)
科技部国际合作项目(No.2015DFA11450)
安徽省科技强警计划项目(No.1604d0802018)
广东省省级科技计划项目(No.2016B010108002)资助~~
