A5008323685- Remote-Sensing Image Classification
- Face and Expression Recognition
- Sparse and Compressive Sensing Techniques
- Advanced Optical Network Technologies
- UAV Applications and Optimization
- Adaptive Control of Nonlinear Systems
- Advanced Image Fusion Techniques
- Blind Source Separation Techniques
- Indoor and Outdoor Localization Technologies
- Remote Sensing and Land Use
- Distributed Control Multi-Agent Systems
- Interconnection Networks and Systems
- Network Traffic and Congestion Control
- Speech and Audio Processing
- Advanced SAR Imaging Techniques
- Robotic Path Planning Algorithms
- Control and Dynamics of Mobile Robots
- Opportunistic and Delay-Tolerant Networks
- Software-Defined Networks and 5G
- Power Line Communications and Noise
Aselsan (Turkey)
2022-2024
TUBITAK BILGEM
2017
Middle East Technical University
2016
Bilkent University
2010
We propose a lightweight sparsity-based algorithm, namely, the basic thresholding classifier (BTC), for hyperspectral image (HSI) classification. BTC is pixelwise which uses only spectral features of given test pixel. It performs classification using predetermined dictionary consisting labeled training pixels. then produces class label and residual vector Since incorporating spatial information in HSI quite an effective way improving accuracy, we extend our proposal to three-step...
We propose a decentralized hybrid swarm control mechanism for quadrotor helicopters. The includes formation, rotation, tracking, and inter-agent collision avoidance capabilities. provide stability analysis of the proposed rule assuming that each has single-integrator dynamics on formation tracking level. Then we integrate level inputs to obtain desired position quadrotors design low individual controllers track these positions. Since output controller is location member, are required have in...
We propose a nonlinear kernel version of recently introduced basic thresholding classifier (BTC) for hyperspectral image (HSI) classification. BTC is sparsity-based linear classifier, which utilizes inner product similarity measure based on the fact that data are linearly separable in feature space. In practice, pixels different classes given HSI not always distinct and may overlap. This known as inseparability problem, reduces performance classifier. order to solve this problem space, we...
The mechanisms behind the sparsity-based techniques for hyperspectral target detection and classifications applications are quite similar except construction methods of dictionaries used by algorithms. In image classification, formed using some known labeled training samples each class. contrast, only a priori spectrum information is available applications. addition, most time, background materials unknown in an arbitrary scene. Although practical approaches such as sliding window exist,...
The authors present a sparsity‐based algorithm, basic thresholding classifier (BTC), for classification applications which is capable of identifying test samples extremely rapidly and performing high accuracy. They introduce sufficient identification condition (SIC) under BTC can identify any sample in the range space given dictionary. By using SIC, they develop procedure provides guidance selection threshold parameter. exploiting rapid capability, propose fusion scheme individual...
In this paper, the dynamic model and validation of agile micro drones, capable formation flight solo attack, are presented. The hardware architecture their features, ground systems required for operation introduced. Developed software modules, route plan attack generation features explained. results from real-flight tests covered.
This letter focuses on the problem of online and instantaneous generation smooth trajectories their analytical time derivatives, where future reference signals provided by a trajectory planner or user are a-priori unknown. For this purpose, we propose modified cubic spline polynomial which transforms piecewise references into continuous functions for digital control systems. The resulting not only in but also analytically differentiable. inferred derivative values could then be used variety...
In this letter, we propose target detector version of recently introduced basic thresholding classifier for hyperspectral images. The proposed technique is a sparsity-based low complexity which achieves high detection rates with very false alarm and performs extremely rapidly. We also new decision metric, background model, spatial smoothing procedure in order to increase the probability further. Experiments show that presented method outperforms other state-of-the-art approaches.
In this thesis, we propose a light-weight sparsity-based algorithm, basic thresholding classifier (BTC), for classification applications (such as face identification, hyper-spectral image classification, etc.) which is capable of identifying test samples extremely rapidly and performing high accuracy. Originally BTC linear works based on the assumption that classes given dataset are linearly separable. However, in practice those may not be context, also another algorithm namely kernel (KBTC)...