为去除图像中与直线检测无关的边缘点，获取完整的直线边缘，准确提取直线段特征，提出一种方向预测搜索与边缘曲率约束的直线段提取方法。利用双阈值方法检测目标边缘点以跟踪边缘轮廓;利用最近跟踪产生的边缘点中含有的方向信息预测下一边缘点的搜索方向，在边缘断裂处采用深度优先搜索方式在预测方向上跨越断裂点继续搜索边缘点，有效避免分岔路径错误选择，保证获得更加完整的边缘轮廓；利用边缘曲率约束和长度约束在边缘轮廓中提取直线段边缘点；利用直线段边缘点拟合直线，并从全局角度合并离散直线段得到最终的直线段。实验结果表明，该方法能有效处理边缘的断裂和分叉，准确提取出更完整的直线段和端点信息，并且具有较快的运算速度，性能优于现有的典型直线段提取方法。

In order to get rid of the edge points which are not related to line detection,obtain complete line edge points and extract line segments accurately,a new line segment extraction method based on predicted direction search and edge curvature constraint is proposed.Firstly,the dual-threshold method is employed to detect the edge points of the target to track its edge contour.Meanwhile,the direction information contained in the edge points generated by the latest tracking of edge points is used to predict the search direction of the next edge point.The depth-first search method is used to cross the break points along the predicted direction and continue to search the edge points,which effectively avoids wrong selection of bifurcation paths and ensures a more complete edge contour.Secondly,the curvature and length constraints of the edge are employed to extract the edge points of the line segments in the edge contours.Finally,the line segments are fitted by using the edge points of the line segments,and the final line segments are obtained by merging the discrete line segments.The experimental results show that the proposed method can effectively deal with edge fracture and bifurcation,extract complete line segments and end-point information accurately and quickly.The performance of the proposed method is superior to that of the existing line segment extraction methods.