Abstract: With the development of target detection algorithms, the intrusion detection based on surveillance video has attracted more attention. Due to the complexity of the traditional target detection algorithm and the difficulty in detecting in real time in the scene of limited computing power and storage space, this paper proposes a lightweight intrusion detection algorithm. Firstly, the preliminary screening target is extracted through the adaptive update rate of the mixed Gaussian foreground extraction algorithm. And then, the preliminary screening target is identified based on the improved residual squeeze network (R-SqueezeNet) classification. It is shown via experimental results that, without reducing the detection accuracy, the proposed algorithm can increase the detection speed by an average of 30 times compared with the traditional algorithm, and the model size is reduced to 1/40 of YOLOv3-tiny.