Aiming at the shortcomings that the conventional Hough transform line detection algorithm is likely to cause false detection, a line detection algorithm based on improved Hough transform is proposed in this paper. First, the edge points in the image are identified into different classes and the image is divided into different parts according to the parameterization of the seed point, and then points in the same region will be parameterized specifically and the parameters of the line in this area will be determined. Finally, the points around the line will be set as background and region division and the line detection will be continued until there is no seed points left in the image. The improved algorithm achieves special parameterization for points in different regions. The points in the same area only vote for one straight line. The experimental results show that, compared with the conventional Hough transform and the algorithms reported in the literatures, the improved algorithm reduces the rate of false detection and has better robustness to fractured edges.