The application discloses a subway gate
ticket evasion early warning method fusing target detection and
behavior recognition, first acquires continuous video
stream data of a subway gate area, carries out frame extraction and pretreatment, and obtains standardized frame images; then carries out target detection on the standardized frame images; carries out effectiveness screening on the detected
human body key points, and forms a purified
pedestrian target detection
result set; matches and tracks the purified
pedestrian targets, generates complete space-time trajectories and key point sequences of each
pedestrian from entering to leaving the gate area; extracts
ticket evasion behavior characteristic parameters, inputs a light space-time graph
convolution network for
behavior recognition; when the
behavior recognition result confidence is higher than a preset threshold, an early warning
signal is triggered. The method realizes full-process
automatic processing from behavior recognition to evidence preservation, can effectively distinguish complex
ticket evasion behaviors such as tailing, drilling and climbing and
jumping, significantly reduces the
false positive rate and the false negative rate, has strong real-
time response capability, high recognition precision and complete evidence chain.