Design And Implementation Of Ship Detection System For Satellite Optical Remote Sensing Image

Using satellite optical remote sensing images for ship detection is an effective means of ocean activity detection,which is of great significance for maintaining national marine security.In recent years,artificial intelligence technology has developed rapidly,and it has irreplaceable advantages in image recognition and target detection.A research institute put forward the pre-research project of satellite optical remote sensing image ship detection system,which requires the use of deep learning method for real-time ship detection of optical remote sensing image,based on this pre-research project,this paper studies and realizes the prototype of the system principle.The ship detection system is mounted on the remote sensing satellite,which processes the optical remote sensing images captured by the satellite in real time,and separately summarizes the images with ship information detected,and finally sends them back to the ground control station through the satellite data transmission center.The whole ship detection system consists of two modules: remote sensing image acquisition module and ship detection module.First,the system uses a custom high-speed serial computer expansion bus standard(peripheral component interconnect express,PCIE)high-speed data acquisition card to receive remote sensing images,and sends them to the ship detection module through the PCIE bus..Using the TX1 embedded development board as the hardware platform of the ship detection system,the Linux operating system environment was built in the TX1 board card,and the PCIE bus driver development was completed to realize high-speed image data transmission between the board and the acquisition card.The function and performance of the module are tested.The module can complete the reading and writing of data.,and has a high data transmission rate,meeting the requirements of system function and performance.Then,the second development of the Darknet deep learning framework,combined with a single step(You Only Look Once v3,YOLOv3)algorithm,designed and implemented the training model of the ship detection system.The accelerated training test of the model is done in a PC environment with Titan XP graphics card.The optical remote sensing image provided by the project is used to make the training data set for the training of the ship detection model,so as to improve the model recognition accuracy.The empirical data set verifies that the model has a high average detection accuracy.Finally,the ship detection model and its detection system software are transplanted to the TX1 embedded development board to realize the ship detection module.The whole ship detection system is tested,the ship detection is carried out by using the optical remote sensing image data sent by satellite sensor,and the ship target in the image is identified and marked.After the system is further improved and installed in the remote sensing satellite,the ship detection efficiency can be greatly improved,the bandwidth utilization of satellite remote sensing image data link can be improved,and the labor cost can be reduced.The optical remote sensing image ship detection system designed in this paper meets the functional and performance requirements of the project.