基于机器视觉的路面交通标志识别的应用研究

发布时间：2018-04-20 06:30

本文选题：神经网络 + Hu矩　；参考：《广东工业大学》2015年硕士论文

【摘要】：本文对路面交通标志包括方向指示标线和车道线进行了识别研究,为驾驶员驾驶车辆提供必要的交通信息,从而可以提高行车安全。本文首先对驾驶员的感知特性进行了研究分析,得到驾驶员的感知特性。同时对不同年龄段不同性别的驾驶员进行了交通行为的实验调查,认为交通标志识别对安全驾驶有很大的帮助。路面交通标志识别系统主要包括三个部分,图像获取、图像检测以及图像识别。为了克服在图像获取过程中可能出现的各种干扰、噪声,本文先对图像进行了预处理,包括灰度化、图像滤波以及图像增强。其中图像滤波采用了中值滤波的方法对图像进行降噪处理,效果较好。为了能够满足实时性以及准确性的要求,必须把图像变得更加的简单,信息尽量少。为此,本文对图像进行了二值化处理,通过实验比较,最小误差法二值化方法的效果最佳。通过分析路面方向指示标线的形状特征,提取出直行、左转弯、右转弯等7类图像的不变距特征作为识别的根据。在不变距特征的提取中,本文分别分析了Hu矩和Zernike矩,为识别路面交通标志提供可靠的依据。在车道线识别中,本文采用了Steerable Filter方法,通过对滤波方向进行控制实现边缘检测,并运用几何计算方式判断汽车是否偏离车道。在不同的光照强度下均可得到良好的效果,具有良好的抗干扰能力。本文还通过对驾驶员换道意图的判断,减少车道偏离的警报误判。本文运用采用Viola-Jones(VJ)方法进行人脸中心横坐标检测,然后使用Kanade-Lucas-Tomasi(KLT)方法对人脸进行跟踪。根据人脸横坐标的移动为依据来判断驾驶员是否有换道意图。最后,本文通过BP神经网络识别系统对路面方向指示标线进行了实验验证,BP神经网络系统识别能够满足实时性及准确性的要求。对人脸横坐标的偏移判断出驾驶员的换道意图,结合车辆偏离车道线实验,设计出结合驾驶员换道意图的车道偏离预
[Abstract]:In this paper, the identification of road traffic signs, including direction indication and lane line, is carried out to provide necessary traffic information for drivers to drive vehicles, thus improving the driving safety. In this paper, the driver's perceptual characteristics are studied and analyzed, and the driver's perceptual characteristics are obtained. At the same time, the traffic behavior of drivers of different ages and different genders were investigated, and it was concluded that traffic sign recognition was of great help to safe driving. The road traffic sign recognition system includes three parts: image acquisition, image detection and image recognition. In order to overcome the interference and noise that may occur in the process of image acquisition, the image is preprocessed first, including grayscale, image filtering and image enhancement. The median filter is used to reduce the noise of the image, and the effect is good. In order to meet the requirements of real-time and accuracy, it is necessary to make the image simpler and less information. In this paper, the binarization of the image is carried out, and the results of experiments show that the method of minimum error binarization is the best. Based on the analysis of the shape features of the road direction indicating lines, the invariant distance features of 7 kinds of images, such as straight line, left turn and right turn, are extracted as the basis of recognition. In the extraction of invariant distance feature, Hu moment and Zernike moment are analyzed respectively, which provides a reliable basis for the identification of road traffic signs. In lane line recognition, Steerable Filter method is used to realize edge detection by controlling the direction of filter, and geometric calculation is used to judge whether the vehicle deviates from the lane. Under different illumination intensity, good effect can be obtained and good anti-interference ability can be obtained. By judging the driver's change intention, this paper reduces the false alarm of lane deviation. In this paper, Viola-Jones-VJ (Viola-Jones-VJ) method is used to detect the center of human face and then Kanade-Lucas-Tomasi-KLT-based method is used to track the face. According to the movement of the face coordinate, the driver is judged whether he has the intention of changing the lane. Finally, the BP neural network recognition system is used to identify the road direction indication line. The experimental results show that the BP neural network system can meet the requirements of real-time and accuracy. According to the deviation of face coordinate, the change intention of the driver is judged. Combining with the vehicle deviation from the lane line experiment, the driveway deviation predetermination combining the driver's change intention is designed.
【学位授予单位】：广东工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U463.6;U495;TP391.41

【相似文献】