EFFICIENT ANOMALY DETECTION ON UAV IMAGES FOR SEARCH AND RESCUE
Abstract
Search and rescue activities include finding and rescuing people and vehicles trapped in difficult. In recent times, unmanned aerial vehicles (UAV) have been used in both military and civilian applications. It is a huge resource for the search and rescue mission. Because this device can carry high-resolution image sensors, a wide range of activities, diverse terrain without too many cores force and cost for the search process. However, the large number of images obtained and combined with high resolution in a large area of a scene is a great barrier to detect with the naked eyes. Therefore, automatic target detection is the right solution. To avoid missed targets, increasing the detection efficiency of the algorithms is necessary. In this study, we propose a method to increase the efficiency anomaly detection of the decision rule based on the ratio test of the ability to use a non-parametric model to estimate the probability density function of the background data by combining with techniques: noise cancellation; SIFT, SURF feature extraction. Test results on the sample data set showed noticeable differences, especially in the case of image noise.