A coherency-based algorithm to build new tracks from local Lagrangian information of particles is introduced. In Lagrangian Coherent Track Initialisation (LCTI), we mainly concentrated on how to initialise first four-time steps, new entries and lost tracks, by a physics-based iterative forward-backwards four frame technique as a complementary function of Kernelized Lagrangian Particle Tracking (KLPT) and Shake The Box (STB). Methods like as KLPT and STB require initial tracks to start predicting and optimising trajectories. More accurate initialisation technique could prevent the algorithm from failing or improve the convergence speed. Moreover, particles are continuously entering into the domain, which means, no matter how effective the algorithm can track a specific number of particles, new trajectories must be fed into the tracked pool. Otherwise, all tracked particles would leave the domain and at some point, there would be no new tracks. In the case of having complex flow motion, some particles lose their trajectories at the end of each time step. It is vital to return those lost particles into the tracked pool since an increasing number of lost tracks will increase the chance of being ignored by the algorithm as ghost or noise. The idea is that new entry and lost particles follow the same behaviour of their surrounding neighbours which means a new track is built if the trajectory is coherent with other neighbour tracks.