Aiming at the problem of 3D target detection in complex environment with random false target interference and strong clutter, a multi-level track before detect algorithm based on the parallel-line-coordinate transformation was proposed. Radar measurement points were placed in the normalized radial range-time plane, azimuth-time plane and pitch angle-time plane successively, the non-coherent accumulation and points filtering were carried out in each level, and each level was processed only on the results of the previous level. Based on the prior information, track constraint and track fusion were carried out to obtain the final detection result. The performance of the algorithm was verified and analyzed through simulation. The results show that the proposed algorithm can effectively filter out false measurement points and achieve the detection of real tracks under different signal-to-interference clutter ratios, measurement errors, and clutter density.