The target signal is broadened due to the cross-pixel effect of the point target and the insufficient readout bandwidth of the detector when using the remote sensing instrument with large-scale infrared detectors to image the point target dominated by stars. After the equal interval sampling and data arrangement, the target image presents a “splitting point” state, which reduces the positioning accuracy and target recognition ability of spatial infrared remote sensing instruments. Aiming at this problem, a “splitting point” targets degradation model is constructed in this paper. According to the strong correlation between splitting points, a target detection process of pre-detection - energy dimension iteration - model updating – post-detection is designed. Through morphological filtering and the multiple hypotheses tracking technique, track the observations of the ' split point ' target independently and obtain the statistical information of “split point”. Then correct the constructed model with known star centroid and map it into the current detection image to correct the target centroid. The test results show that the method can effectively distinguish the false target caused by ' split point ' and multi-target, and the positioning accuracy is improved from more than 0.5 pixels to less than 0.15 pixels.