Special scene lane line detection method and device using video time sequence information

The invention provides a special scene lane line detection method and device using video time sequence information, equipment and a storage medium, and the method comprises the steps: obtaining road video data, segmenting the road video data into video clips, and carrying out the frame extraction of each video clip; performing lane line marking on each frame of video image data; carrying out random erasing processing and augmentation processing on the annotated image data; inputting the training image data into a pre-constructed lane line detection model according to a time sequence, and predicting to obtain a lane line prediction result of a current frame by using the training image data of a preset frame number as historical information by using an LSTM neural network; calculating a loss function value based on a comparison result of a predicted value and a truth value so as to train the lane line detection model; and performing lane line identification by using the trained lane detection model to obtain a