未系安全带智能识别中安全带定位关键方法研究

安全带定位是实现机动车未系安全带智能识别的关键。针对道路监控图像特点,提出一种基于梯度变换的安全带定位方法。该方法在对卡口图像进行预处理的基础上,采用自适应阈值边缘检测算法以及积分投影方式定位车辆位置,设立车牌检测区域,以减少运算量,降低干扰,同时利用训练得到的Haar分类器识别车牌位置,通过逼近方式切取车辆右侧图像。最后采用梯度变换算法求得车窗各边缘坐标,实现车窗精准定位,并计算得到安全带位置。试验表明,该方法可实现安全带的准确定位,具有较好的实用性,为后续安全带识别奠定基础。

Research on Key Method of Seatbelt Location in the Intelligent Recognition System for Seatbelt Unfasten

The seatbelt location is the key to realizing the intelligent recognition of seatbelt unfasten. According to characteristics of the road monitoring image, a method of seatbelt localization based on gradient transform (GF) is proposed. With the method, by preprocessing the road monitor image, the vehicle position is located by using canny edge detection based on adaptive threshold and horizontal and vertical integration projection. And then license plate detection region is designed, with purpose of reducing image data computation, eliminating the interference from other regions. Nextly, license plate location is obtained by Haar-like cascaded classifier, which can get the image of vehicle on right in the method of approximation. Finally, the edge coordinates of car window can be calculated by gradient transform (GF) that can locate the position of seatbelt. The result shows that the method proposed can locate the position of vehicle window and seatbelt with relatively high applicability and accuracy, which may meet requirements of the urban traffic management.