Classification of small moving objects in atmospherically-degraded video

Acquisition and classification of moving objects in imaging through long-distance atmospheric path (more than 1-2 km) may be affected by distortions such as blur and spatiotemporal movements caused by air turbulence. These distortions are more meaningful when the size of the objects is relatively small (for instance, few pixels width). This work aims to study and quantify the effects of these distortions on the ability to classify small moving objects in atmosphericallydegraded video signals. For this purpose, moving objects were extracted from real video signals recorded through longdistance atmospheric path. Then, various geometrical and textural object features were extracted, and reduced to two principle components using principle component analysis (PCA). The effect of the atmospheric distortion on object classification was examined using support vector machine (SVM) classifier. Furthermore, the influence of image restoration on the classification performances was examined for the real-degraded videos. Results show how classification performances are decreasing when the images are degraded by the atmospheric path compared to the case where successful image restoration is performed.