Pedestrian re-identification video stream complex target detection method based on event-related potential

The invention discloses a pedestrian re-identification video stream complex target detection method based on event-related potential. According to the method, video streams of a complex target appearing in different scenes are collected and made into an experimental normal form, and electroencephalogram data of a subject when the subject watches video content are collected. Secondly, analyzing the electroencephalogram topographic map, and marking ERP features including P300 and P300-D which correspond to target appearance and target disappearance respectively; the method comprises the following steps: constructing a positive and negative sample pair, obtaining inter-class essential features based on a contrast representation learning method and a time-space feature attention extraction method, solving the problem of extreme class imbalance of samples in a video normal form and the problem of distinguishing two similar classes, carrying out back propagation by using circloss, updating model parameters, and ext