A real-time omnidirectional target detection system based on FPGA

The Field Programmable Gate Array (FPGA) has been widely used in time-critical tasks such as object detection and tracking. To better accomplish the above tasks, the detection range of the system becomes more and more important. Therefore, a real-time omnidirectional target detection system was carried out. In this system, a dim and small target detection method based on cascaded guided image filter (CGIF) is proposed. Our FPGA-based system shows excellent performance in dim and small target detection with the high detection rate (94%) and the low false alarm rate (5%). First, a target detection algorithm based on cascaded guided image filter is introduced, where the details of large objects can be well suppressed, and the targets are extracted through background suppression and clustering. Second, we design an FPGA-based system, which contains multichannel videos as input through Peripheral Component Interconnect Express (PCIE) interfaces, based on mother-daughter architecture for real-time implementation. Moreover, we also propose a one-line cache display method to exhibit the detection results, which can save much storage space with the same display effect compared with the traditional cache-based scheme. The experimental results show that the proposed system has promising performance and efficiency.