中文摘要：

车身参数化模型的自动提取有助于进一步的车型分析及评价，提高海量图像数据的重用率，以及提升利用图像重建三维模型的高效性。通过图像中Canny特征点的自反对称对车身前视图对称轴的自动提取，利用给定的车身参数化模板，提出了应用于车身的利用山谷线的分水岭方法得到的一种新型图像分割技术，结合动态规划的优化思想，全自动地对车身前视图参数化模型进行提取。对背景复杂的高质量车身图像，也可以得到理想的提取效果。

英文摘要：

For the massive amount of data in images nowadays, automatic detection of automotive parametric model can help further analyze and evaluate different car types and improve the efficiency of 3D model reconstruction from images. Using the front-view parametric template, a new image segmentation method was proposed by applying the method of the watershed algorithm with valley lines. The symmetric axis was computed through the reflexive symmetry of Canny feature points in front-view images of vehicle bodies. Combined with the dynamic programming optimization method, the parametric model of the vehicle body fi-ont-view images was automatically constructed. For high precision automotive images with complex background, the desired detection effect can also be obtained.