The embodiment of the invention discloses a lane line detection method and system based on an improved fitting algorithm. The method comprises the following steps: calculating a probability graph corresponding to target image data based on a semantic segmentation algorithm, the target image data comprising first area data and second area data, and the probability graph comprising a first probability graph corresponding to the first area data and a second probability graph corresponding to the second area data; carrying out image fitting on a target sample point in the first probability graph;verifying a target fitting model of image fitting by adopting the sample points in the second probability graph and the remaining sample points in the first probability graph; detecting lane lines inthe target image data according to the target fitting model; wherein the sample points are sample points screened out from all pixel points of the first probability graph and all pixel points of the second probability graph. By adopting the method and the device, the accuracy of a detection result can be ensured and the robustness can be improved through a semantic segmentation algorithm in combination with an improved data fitting algorithm by utilizing the priori structure information of the lane line.