The invention discloses a multi-camera target tracking method based on video structural description, which comprises a video structural description function and a multi-camera target relay tracking function, and used for tracking an interested target in a scene and performing structural description on characteristics of the target. The method comprises the steps of extracting a target profile from a video scene through a Gaussian mixed model, obtaining a complete target profile through Blob communication region analysis and obtaining target position and size, calculating the color histogram of the garget, realizing the target tracking in the scene according to position characteristics of the target in two frames, and contrasting coordinate information and color information of the target with information in a file when the target appears in the scene of another camera, and then performing relay tracking if conditions are satisfied. The method utilizing a plurality of cameras is beneficial to solving the target tracking problem in conditions of blocking, chaos scene, light change cancellation and the like.