When a super wide-field infrared staring system is used to detect a target, it may show complex background, more noise jamming and little target information. Therefore, this paper proposes a spatial-temporal fusion algorithm. In spatial processing, the single image was filtered by an improved Robinson operator, and the primary detection of the target was carried out. Then, the image auto-partition and spatial false-alarm suppression were presented to eliminate the suspected target from a non-target detection region. In temporal processing, the temporal characteristics of the target was considered, the neighbor judgment method was improved to confirm the real target. A detection experiment for the space target was carried out in a moon light, and this proposed algorithm was vilified. It shows that the background interference of original image is suppressed greatly, the local Signal to Noise Ratio(SNR) can be improved more than 1.3 times, the false target number decreases by 70% after spatial processing. Moreover, the infrared weak target is detected successfully with the detection probability of higher than 95%, the false-alarm probability of less than 1.5% and the minimum detection SNR of 2.86 after spatial processing. It concludes that this algorithm is suitable for the infrared weak target detection of super wide-field image and can detect the moving point targets effectively.It has a good suppression effect for background interference, isolated point sources and instantaneous noise.