The invention relates to a line-structure light strip center line extraction method based on improved Laplacian edge detection, and belongs to the technical field of robots or unmanned vehicles. The method comprises the steps of decomposing 8-adjacency convolution template in Laplacian into four convolution templates based on an improved Laplacian edge detection method according to the edge characteristic of the line-structure light, and quickly and effectively acquiring light strip edge pixel points and gradient direction thereof via the four convolution templates; acquiring the accurate and complete line-structure light strip edge according to the continuity of adjacent light strip space positions, the similarity of edge pixel point gradient values and the similarity of edge gradient directions, and filtering out false light strip edges; and at last completing the extraction of the line-structure light strip center line according to the gray level distribution of the light strip pixels along the edge gradient direction. According to the method, quick and effective extraction of the line-structure center line can be completed, and thereby the effective acquisition of three-dimensional information of an obstacle in the obstacle recognition based on the line-structure light is achieved.