Design And Application Of Sea Target Recognition System Based On UAV Infrared Remote Sensing Image

In a range test, it is very important to identify and keep monitoring the focus targets at sea in order to know the process and results of the test in real-time. Under traditional test conditions, videos for test are often collected through test video ships, video helicopters or land cameras. There are usually great risks of test using video ships and helicopters; while it is also hard to get ideal filming effect with land telephoto lens restrained by weather factors or filming distance. Therefore, UAV is used to film target ship and the infrared target images at sea are obtained through the infrared image sensor on UAV platform. The focus target recognition is executed to achieve effective detection and identification. Then, the acquired target ship images are transferred to the command center in real-time for the commander to fully know the accurate test state and make decisions efficiently.This thesis makes in-depth analysis on the imaging characteristics of infrared ship target images. To preprocess the infrared images, this thesis employs denoising, gray-scale modifying and sharpening processing. The original images are first processed using median filter and gray-scale modified using linear transformation function. Aiming at the edge fuzzy problem after denoising and gray-scale modifying, this thesis makes further sharpening process with Laplace Operator which is proved to be more accurate between all spatial domain methods.In this thesis, wavelet transform method is introduced to process target segmentation and detection for pretreated infrared images. Targets on sea are always complicated and fuzzy. Two kinds of algorithms based on wavelet transform are adopted to segment targets. First, feeble infrared target detection method based on edge feature wavelet algorithm is used to detect targets. If the target is not segmented by the first step, a target segment and detection algorithm based on wavelet multi-scale energy combinations and their mutual-across filtering method is introduced in this thesis to ensure successful target detection.A deep research on traditional pattern recognition technologies and artificial neural network theories is carried out in this thesis. The Pattern recognition technique based on BP-ANN is utilized for target recognition. The BP-ANN target recognition software is programmed and used in practical application. The target recognition accuracy rate is increased by the ANN method combined with the analysis of information such as torpedo situation, warship AIS and information acquired by other measurement and control devices.Finally, this thesis constructs the uav infrared remote sensing image video platform. This thesis focuses on the design of the information acquisition and recognition processing system, the system mainly consists of infrared remote sensing image acquisition module, the remote sensing image target segmentation and detection module and the remote sensing image target recognition module the system can be used to achieve the target ship image preprocessing, target segmentation and detection, and the final target recognition.