偏振光实时定位的标识线导航路径图像识别

Image Recognition of Marking Line Navigation Path Based on Polarized Light Real-Time Positioning

针对传统方法存在图像噪声免疫性弱、识别偏差大、导航路径实时性较差的问题,提出基于偏振光实时定位的标识线导航路径图像识别方法。通过选取四个偏振光的偏振方向向量、时间信息、太阳方向向量判定方式与地磁信息,获取当地的地理坐标,完成偏振光实时定位。采用中值滤波技术实现图像的降噪与边缘处理,通过目标、背景区域与整体路径图像的关系,利用Otsu方法在直方图灰度变化区间内选取一个阈值,令该阈值满足类内方差极小、类间方差极大的条件,从而获得最佳阈值。再使用架构的判定函数验证阈值,根据抽行扫描法获得标识线的中心像素方位,并对其拟合,最后依据数学形态学实现对标识线的有效识别。实验结果表明,所提方法可以有效去除图像中存在的噪声点,在路径识别中能有效识别出各类标识物,识别效果较好。

Aiming at the problems of traditional methods such as weak immunity to image noise, large recognition deviation and poor real-time performance of navigation path, an image recognition method based on polarized light real-time positioning is proposed. In this regard, we reported a real-time image recognition method based on polarized light. The polarization direction vector, time information, sun direction vector determination method and geomagnetic information of the four polarized lights were selected to obtain the local geographic coordinates for completing the real-time positioning of polarized light. Based on median filtering technology, image denoising and edge processing were achieved. According to the relationship between the target, background region and the whole path image, Otsu method was used to select a threshold in the gray change interval of the histogram, so that the threshold can meet the conditions of minimum variance within the class and maximum variance between classes, and the best threshold was obtained. The decision function of the architecture was used to verify the threshold. Eventually, based on mathematical morphology, the effective identification of the identification line was completed. The results show that the method with excellent recognition effect, can effectively remove the noise points in the image, and identify all kinds of markers in the path recognition.