针对工业零部件序列图像量测与重建中的线特征提取问题，提出了一种以CAD设计数据为引导，高精度检测图像中直线的数学模型。以三维计算机辅助设计(CAD)数据为基础，通过求解CAD三维直角坐标系与图像坐标系间的参数转换关系，将物方线特征投影至图像上，得到图像线特征初值范围。利用自适应最小二乘模板匹配提取出边缘点序列，对生成的上述点序列采用最小二乘法进行拟合，得到图像线特征的直线方程。通过求解图像相交直线的交点，高精度地获得表达工业零部件轮廓的直线段。采用上述算法对真实工业零部件图像进行了线特征提取，重建的直线长度与游标卡尺量测的直线长度之差的中误差为0.232 mm。实验结果表明，文中所给算法自动化程度高、稳定性强，可用于规则工业零部件高精度图像量测与重建。

For feature extraction from image measurements and reconstruction from industrial component images,a mathematical model guided by Computer Aided Design(CAD) data is put forward to detect image lines with high accuracy. Firstly, using CAD data, the line initial values are obtained by projecting object line features through the conversion of 3D CAD coordinate and images. Then a sequence of edge points are extracted by adaptive least squares template matching (LSTM) and leads to image line equations from these points based on the principle of least squares fitting. Finally the contours of industrial components are obtained accurately by solving the intersections of image intersecting lines. By using the algorithm on real images of industrial components, we perform a line feature extraction experiment with an average standard deviation of 0.232 mm between the rebuilt line and the actual line length.This proves that the algorithm has high automation extraction and strong stability, and can be used to the image measurement and reconstruction of industrial components with high accuracy.