Sunlight reflected on the moving sea surface, often referred as sun glitter, can significantly impact the input data in many application areas like defense, security and border surveillance. The effectiveness of vision-based algorithms like target detection and tracking will decrease remarkably under the sun glitter. To deal with this drawback and give a reaction, the sun glitter should be detected in the scene firstly. In this paper, a straightforward and efficient sun glitter detection and classification method for infrared scenes is introduced. In the first step, the behavior of sun glitter in infrared scenes is analyzed. Two stepped frame buffering strategy with dilation process is used to model the spatiotemporal connection between glitter points. After some thresholding steps, the sun glitter detection decision is taken by checking the total number of glitter candidate pixels. And the classification, which means labeling the sun glitter level as low, mid or high for the scope of this paper, is made by checking the number of blobs that are close to each other. The performance of proposed algorithm is tested with real world infrared dataset, which has 30 different scenarios.
|