P1: Professional papers 1
- Design and Development of a Low-Cost Hen Eggs Incubator
Ante Palac; Ana Kuzmanić Skelin; Mirjana Bonkovic (University of Split, Croatia)
Modern trends are oriented toward appropriate solutions, which improve the food yields without disturbing natural balance among plant and animal species. This paper presents design and development of a low cost hen eggs incubator realized using cheap technology based on Arduino development environment. Applied control logic imitates the natural principles of eggs hatching and supports the natural process of meat production. It is suitable for small farms oriented toward production of healthy food and is appropriate even for DIY work.
- Classification of driving style using car’s on-board sensors and neural network
Matej Matisic; Josip Music (University of Split, Croatia)
Reckless and aggressive driving in modern cities/states cause significant material cost as well as loss of life. This in turn has significant impact on city/state function as well as health related issues. Thus in the paper a simple but promising approach to car driving style estimation is proposed and tested. It is envisaged to be a part of a larger warning (prevention) system. Three general driving styles (standing, normal driving, aggressive driving) are defined and used. The approach is based on neural network and car/engine data readily available in modern cars through OBDII interface. The approach can also be used to estimated different car parameters. Proposed approach demonstrated high accuracy with 94.7\% and MSE of 0.0353 on the test dataset acquired during 40 minute test drive.
- Face and Nose Detection in Digital Images using Local Binary Patterns
Stanko Kruzic and Vladan Papic (University of Split, Croatia)
This paper describes an approach to object detection based on Viola-Jones algorithm and LBP histogram. LBP is used on sub-windows of the input image to obtain LBP feature histogram, from which small number of key visual features is extracted using machine learning algorithm based on AdaBoost. The final decision if image contains the object is training cascade of classifiers which rejects the negative sub-windows quickly and use processor time for those sub-windows which are have higher probability for containing the object in question.
- Performance comparison for NVIDIA CUDA and Intel Xeon Phi
Petra Loncar; Sven Gotovac; Vladan Papic (University of Split, Croatia)
NVIDIA graphic processor unit (GPU) using Compute Unified Device Architecture (CUDA) and Intel Xeon Phi are two different parallel computing platforms. The challenge using the platforms lies in achieving the best possible performance. The real value in GPU and Xeon Phi coprocessor programming lies in the ability to run large number of parallel threads. To provide fair comparison of the processors, OpenMP technique, hand-tuned CUDA kernels and shared memory have been used. For the Xeon Phi, the application has been executed using offload mode. Memory capacity represents the main limitation for current generation of Intel Xeon Phi coprocessors, which is also a limitation for GPU. For that reason, offload programming is the most popular mode as the memory capacity of the host can be used to prevail that limitation. In this paper, the performance of NVIDIA CUDA and Intel Xeon Phi using matrix multiplication code and K-means algorithm has been compared. Performance tests measure and compares kernel execution times and end-to-end application execution times for both CUDA and Xeon Phi.
- FPGA based digital image processing system for aerial search and rescue assistance
Ante Gotovac; Sven Gotovac; Vladan Papic (University of Split, Croatia)
In recent years the technical advancements of Unmanned Aerial Vehicles (UAVs, also called drones) give new opportunities in preventing accidents and saving lives when disasters occur. During Wilderness Search and Rescue (WiSAR) operations it is crucial to make quick real time decisions, as the response time can make a huge difference in some critical situations. The drones can be used to locate missing persons, discover debris on sea and wildlife areas, to scan wildfires and in military surveillance and reconnaissance operations. These tasks can be accomplished by using intelligent data processing equipment where the supervising personnel would only need to define search targets and to evaluate the recorded images of lost person or other object of interest. The images taken by the drones need to be processed in order to give information on the nature and the exact geographical location of the detected objects. In this paper a novel FPGA based digital image processing method will be discussed and evaluated. The results have been presented and analyzed.