The intelligent parking lock (IPL) can monitor parking spaces with its built-in magnetic sensor, and the parking information can be used to realize automatic payment, parking reservation, parking sharing, and so on. However, in the operation of locking and unlocking, the IPL generates magnetic interference signals, which will result in false detections of traditional parking detection algorithms. To solve this problem, a robust parking detection algorithm was investigated. The vehicle magnetic signal is modeled as a variance sequence, and a filtering method of interference signal is applied to remove noisy signal. The parking detection algorithm is implemented based on the state machine. In the experiment, fifty IPLs were deployed in the parking spaces. The results show that the proposed algorithm is more robust and can improve the vehicle detection accuracy.