The invention discloses a video-based traffic information acquisition method. The method comprises the following steps: based on the image processing and the mode identification theory, setting a detection region and carrying out geometric correction, extracting the foreground object by Gaussian Mixture Model, identifying a target by a separator, matching the identification result with the foreground object, eliminating the detected false-alarm, fusing the particle filter algorithm and the vehicle detection algorithm to achieve the self-adaptive track of vehicles, achieving the effective target increase and target fusion on basis of the detection result, and finally achieving the calculation of parameters, such as flow statistics, speed, intensity, time occupancy ratio and the like. Furthermore, according to the detection track result; the microscopic traffic analysis on the target, such as the detection on illegally parked vehicles, retrogradation, and overspeed can be also carried out. At the same time, the traffic status, such as traffic congestion detection can be judged; and the queue length calculation under the congestion condition can be calculated. The method has the advantages of high accuracy and low computing complexity and can meet the requirement of the actual application.