Target Localization Based On The Trajectory Of Unmanned Aerial Vehicles

IIVisual measurements of the target can be transformed into bearing measurementsusing the vehicle position and orientation,the camera orientation with respect to the vehicle,and the pixel location of the target.For the purpose of this project,image processing for target recognition is assumed and the pixel location corresponding to the centroid of the target is available.The task of target localization presents several challenges.Most modern small UAVs are equipped with GPS and inertial sensors,but are designed to be inexpensive and therefore include low quality sensors and poor navigation algorithms,leading to inaccurate vehicle state estimation.Small unmanned aerial vehicles(UAVs)equipped with navigation and video capabilities can be used to perform target localization.Combining UAV state estimates with image data leads to bearing measurements of the target that can be processed to determine its position.This 3-D bearings-only estimation problem is nonlinear and traditional filtering methods are prone to biases,noisy estimates,and filter instabilities.The performance of the target localization is highly dependent on the vehicle trajectory,motivating the development of optimal UAV trajectories.When different sensors perform target tracking on the system,they will have different effects on the tracking results.For example,the target recognition rate and tracking accuracy of UAVs are relatively low,the range of the image sensor is relatively short,and objects are easily not recognized.In practical applications,the tracking of moving targets or stationary targets becomes a very challenging subject due to the interruption of the fast moving camera of the UAV.In this paper an algorithm is outlined in three stages,segmentation,feature extraction and classification,in order to be able to assist a user in visually tracking small UAVsSegmentation of image sequences into differently moving objects(or,more correctly,their projections in the image)is one of the main focus in this thesis.The segmentation methods lie in(or between)two groups;those detecting flow discontinuities(local operations)and those detecting patches of self-consistent motion according to set criteria(global measurements).It is usually pointed out by proponents of the latter approach that the former is very sensitive to noisy flow measurements.Therefore,we used Kalman filters,i.e.,Kalman filter,extended kalman filters and unscented Kalman filter,for noise removal.The methods of statistical regularization and image transformation fall somewhere in between these groups,attempting to achieve global minimization of locally defined cost functions,or finding best fits to image transformations respectivelyFeature extraction is a main component in our proposed method,we have used GrayLevel Co-occurrence matrix(GLCM)features for object tracking in the UAV.The GLCM texture analysis is based on assumption that the texture information of an image is an average spatial relationship between the gray tones in the image.Thus,this similarity in spatial resolution properties makes GLCM features suitable for object tracking in our proposed method.KNN classifier is used to differentiate the target and the background,in the classification process,the unlabeled query point is simply assigned to the label of its k nearest neighbors.Here we use KNN to recognize the object.We give trained image features and test features as an input to KNN.