Streamline is an important technique in computational fluid dynamics for vector visualization. This algorithm is based on tetrahedron cells. Firstly, hexahedron cells in data field of point cloud were divided into tetrahedron cells. Tetrahedron relations between the adjacent tetrahedron were stored, and then the point of streamline in data filed was located into the tetrahedron cell gradually by looking up the tabulation of adjacent tetrahedron relations. Then according to the deflexion angle of the velocity vector on streamline and the radius of inscribed sphere in the tetrahedron cell, integration step was adaptively determined to improve the precision and efficiency of streamline-tracking. Finally, the points of streamline were obtained by Runge-Kutta numerical integration method, and three-dimension streamline was constructed with the point location and integration step. As a result of avoiding Jacobian matrix inverse transform that is used in the traditional streamline-tracking algorithm, this method reduces the error of matrix conversion and improves efficiency and precision of streamline construction.