| Automatic target recognition (ATR) system under complex maritime scenes is a research hot and difficult topic that the scholars from home and abroad have been engaging. The module of Infrared (IR) small weak target detection is in the front of ATR system whose performance mostly depends on the target detection. This dissertation deeply studies the detection of IR small weak target in the dynamic sea background. The main contributions of this thesis are given below.1.Because publicly available dataset--the base of algorithm test and evaluation does not exist, we firstly introduce an IR objects image dataset under maritime background, then, label targets bounding boxes manually in the database which are treated as the ground truth of judging different methods.2.This paper presents a novel fast detection approach of IR small weak objects under maritime dynamic scenes. The main idea is firstly background estimation based on the fast local minimum filter, secondly, fast detection and location of objects in the background subtraction maps. Extensive experiments on the datasets which has been established give the testing result of the qualitative and quantitative comparisons with other methods. The result demonstrates that our algorithm shows a better performance with higher precision and more suitable for real-time application.3.This paper presents a detection approach of small weak moving objects based on sequential images in low signal-noise-ratio condition. The main idea roots in Hough transformation, which votes the potential moving trajectory of an object in the sequential images. Therefore, the weight of true trajectory will increase continually with the continuous input of moving information. The object is finally confirmed according to its weight. Extensive experiments show that our algorithm can effectively remove false alarms and confirm true objects. Moreover, it can handle miss frame which is much common in the object detection. |
PDF Full Text Request