中文摘要：

为解决复杂自然背景下交通标志难检测的问题，选取警告与禁令两大类交通标志进行研究，分别针对警告和禁令标志建立圆形、三角形模板，利用模板对图库中图片进行匹配，并提出了可变形的模板匹配算法。该算法不同于传统交通标志检测方法，针对采用单个颜色空间对图像进行粗定位时出现漏选区域的弊端，采用叠加后的RGB、HSV颜色空间对复杂背景下的交通标志进行粗定位。为避免使用固定模板所带来的误检与漏检的情况，进一步利用仿射变换使固定模板变为可变模板并利用可变模板对结果进行模板匹配。仿真结果表明，利用可变模板对交通标志进行检测，检测准确率达85％以上，对比其他方法，准确率较高。由此可见，基于AT—SBTM的交通标志检测算法能够准确地检测到交通标志，适于交通标志的识别，可实现辅助驾驶。

英文摘要：

To solve the problem of difficult detection of traffic signs under the complex background of nature, the two kinds of warning and ban traffic signs are selected to be studied, according to which the templates of circle and triangle are established which are used to match the pictures of gallery. Then the template matching algorithm based on the variable shape is proposed, which is different from the traditional traffic sign detection method. According to the disadvantages of selection leakage when the single color space is applied in coarse location of image, the coarse location is conducted in traffic signs with complex background by RGB and HSV after superposition. In order to avoid error and leak inspection using fixed template, the affine transformation is employed further to make fixed template vari- able which matches the results. Simulation shows that by using the variable template for traffic sign detection, its accuracy has been more than 85 %, higher than other methods. Therefore, it could be able to accurately detect the traffic signs, which is suitable for recognition on traffic signs and implemented in the auxiliary driving.