Abstract: Frame-to-frame differences between weak and small moving targets in infrared laser images are often minimal, rendering reliable detection and tracking challenging. Conventional target-tracking algorithms frequently encounter issues such as unstable positional estimates and chaotic motion trajectories, leading to distorted detection. To address these challenges, research on intelligent tracking and detection of small moving targets in infrared laser imagery has explored the integration of bidirectional filtering using the Camshift algorithm. In this approach, bidirectional filtering, incorporating both bidirectional diffusion and shock filtering, was employed to enhance the quality of infrared laser images. Subsequently, the Camshift algorithm, combined with the HSV color model, was applied to analyze the color probability density distribution of the enhanced images, thereby improving tracking accuracy. The positional and shape features of weak and small targets in infrared laser images were extracted, enabling contour detection of the moving targets. Based on the contour detection results, a fully convolutional Siamese network was employed for continuous tracking of moving targets in infrared laser video sequences. Experimental results demonstrated that the proposed method effectively detects small and weakly moving targets, accurately tracks their motion trajectories, and satisfies real-time tracking requirements.