Abstract: Autonomous underwater vehicles (AUVs) usually operate in unknown marine environments. To enhance navigational safety, we present an underwater obstacle target detection method based on a background cancellation constant false alarm rate detector to enable autonomous collision avoidance. Aiming at the non-stationary and non-Gaussian characteristics of underwater acoustic signals, the method introduces a background cancellation constant false alarm rate detector to perform threshold-based detection processing on sonar beam data. Then, combined with the motion characteristics of the target, multi-target tracking is conducted via data association and the Kalman filter, producing obstacle target tracks and estimated track lifetimes. The method is verified through numerical simulation and pool experiments. The results of the pool experiment show that the target detection rate of the background cancellation constant false alarm rate detector is 93.9%, which is 3.6% higher than that of the traditional detector. In addition, it can establish target tracking trajectories more quickly, demonstrating good application prospects.