Intelligent multiframe detection aided by Doppler information and a deep neural network.

• False tracks are excluded tremendously with the Doppler information considered. • The model uncertainty in fusion detection is bypassed by a data-driven method. • Track features on different attributes are sequentially fused by a single DNN. Multiframe detection can be effective for the detection of targets with low signal-to-noise ratios by integrating the target information in several consecutive frames. Traditional methods usually suffer from high computational complexity and insufficient use of available information. In this paper, we propose solving the multiframe detection problem through two sequential steps: Doppler-aided track formation (DTF) and data-driven track detection (DTD). By using the DTF, which considers the coupling relationship between the location measurements and Doppler measurements, the potential tracks can be extracted without introducing too many false tracks. Using DTD, which fuses multiple features along the extracted tracks through a deep neural network, the detection can be sequentially declared. Simulation results show that the detection performance of the proposed method is better than that of the traditional methods for weak targets, and that false alarms can be handled well. Both theoretical analysis and experimental studies corroborate the computational efficiency of the proposed method in real-time implementation. [ABSTRACT FROM AUTHOR]