Accurate parking slot detection is crucial for autonomous vehicles to navigate automatic parking. In recent years, significant progress has been made in research in this field, with some methods achieving high accuracy and efficiency on specific datasets. However, existing methods still face some challenges in practical applications, such as: when the ground reflection covers the parking-slot marking lines, the recognition accuracy of existing methods will significantly decrease; The generalization performance of existing methods is not strong enough, and the recognition accuracy will decrease after changing the place or camera model. To address these limitations, this paper introduces Se-PSD, a novel parking slot detection method utilizing image sequences. Se-PSD analyzes a series of images to predict individual marking point locations, shapes, and orientations. Finally, through geometric rules, parking-slots can be found on the last image of the image sequence. Se-PSD prioritizes generalizability without sacrificing accuracy compared to existing methods. While real-time performance may be slightly impacted, the relaxed time constraints of automatic parking applications make Se-PSD a promising solution.