摘要
针对目前泊车辅助系统中的车位线识别问题,建立了基于360°全景鸟瞰图像的全自动车位线检测与识别模型,考虑到光照对图像处理结果的影响,先对图像进行了预处理,然后采用一种基于中值的自适应Canny边缘检测技术,并通过Hough变换,再根据车位线特征的先验知识对Hough变换结果进行限制和优化,实现了车位线的识别。同时对实际采集到的图像进行验证,结果识别率达到94.2%,证明了该方法的有效性和鲁棒性。
This paper describes a fully-automatic method for recognizing parking lines based on a 360-degree panorama bird＇s eye view image which is gaining popularity in parking assist system. With the effect of illumination on the outcome of image processing in consideration, the pre-processing is operated on the image firstly. Then an a- daptive Canny edge detection technology based on the image median and Hough Transform is used. Finally, utilizing the priori knowledge about the characteristics of marking lines to constrain the outcome of Hough Transform, we real- ize the recognition of marking lines. Experiments on the actual image show that proposed method successfully recog- nizes the most parking lines, and demonstrated a detection rate of 94. 2%, which proves the effectiveness and ro- bustness of the proposed method.
出处
《电子科技》
2014年第4期146-150,共5页
Electronic Science and Technology
基金
河南省教育厅科技基金资助项目(12A520030)
郑州市科技计划基金资助项目(121PPTGG358-10)
