摘要
基于图像拍摄成像过程中雾霾天气及相机抖动,提出了一种从单幅图像中移除未知相机抖动的算法,利用图像的形状特征、颜色特征、纹理特征及Hough变换,可以有效地识别交通信号灯、障碍物及道路。利用先近后远,先简单后复杂的原则,设计了一种基于图像去雾和图像检测的交通信息提取算法。算法首先进行图像预处理,然后对图像进行边缘检测,获得每个物体的多边形轮廓;然后根据物体特征分别利用不同算法对物体进行分类。实验结果表明,算法可以有效地对实时环境中包括道路、车、行人、盲道、斑马线、交通灯类型等物体识别,图像检测算法可以满足导盲的要求。
The key problems of traffic information are the haze weather and camera shake appeared in the imaging process. To solve these problems,an algorithm is proposed to remove the unknown camera shake from a single image.By using shape features,color features,texture features and Hough transform,the algorithm can effectively identify the traffic signal lamp,obstacles and road. Finally,by the near-first,brief-first principles,an image to the fog and traffic information extraction algorithm is designed based on image detection. The algorithm firstly makes pre-processing of image,and then detects the image edge and obtains polygon contour for each object,at last depends on each object characteristic to use different algorithms to classify objects. The experimental results show that the algorithm can process many objects effectively including road,vehicle,pedestrian,blind path,zebra crossing,the traffic lights types. In realtime environment,the image detection algorithm can satisfy the requirements.
出处
《实验室研究与探索》
CAS
北大核心
2015年第12期9-12,31,共5页
Research and Exploration In Laboratory
基金
国家自然科学基金(61304024)
中央高校基本业务经费(DX1201A)项目
关键词
图像去雾
电子导盲
边缘检测
图像检测
交通信息
image dehazing
electronic guide system
edge detection
image detection
traffic information
