Generalized Expression for TP-CFAR Model in Point Target Detection of Subpixel Imaging

The two-parameter constant false alarm rate (TP-CFAR) detector is an important research direction for infrared (IR) point target detection. The traditional pixel-level TP-CFAR has great shortcomings in keeping a high detection probability for subpixel imaging targets. We establish a discretized imaging model for the point target and show that the subpixel imaging diffuses the target energy to the neighboring pixels and makes the pixel-level energy almost useless. To eliminate the diffusing effect, we propose an energy sum procedure and illustrate how this can be incorporated into TP-CFAR to derive a generalized model, which can obtain a higher energy concentration than the pixel-level model. Further, an optimization algorithm based on the maximum energy concentration measure (ECM) and a maximum likelihood estimation (MLE) method is proposed to determine the best region scale and estimate the target parameters in the energy sum procedure. The experiments on the real IR target with different projected positions verify that the proposed method is not affected by the energy diffusion and solves the problem of miss detection for point targets with subpixel imaging.