REAL TIME WEATHER RADAR SIGNAL AND DATA PROCESSING SYSTEM USING PARALLEL AND PIPELINING COMPUTATION TECHNIQUES

• Main Author: Satria Perdana, Riza
• Format: Dissertations
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75278
• Institution: Institut Teknologi Bandung

Weather radar is used to detect precipitation, calculate its speed, and predict its type (rain, snow, ice, and so on). Graphics Processing Units (GPU) can be used for radar data processing at the signal processing and data processing stages. This is necessary because the processed radar data is large in size and the computational process allows it to be parallelized. In this research, a signal processing system and weather radar data are developed in parallel using the GPU and using a pipeline technique so that the output products (both basic products such as reflectivity and Doppler speed, as well as derivative products such as CAPPI) can be produced in real time. If the weather radar output product can be generated in real time then the next analysis process (performed by weather experts) is expected to be faster and can be used for a wider purpose. In this research, a parallel weather radar signal processing algorithm has been created and has been implemented into the CUDA platform and its performance is compared to the serial version on the CPU. Furthermore, further optimization of the weather radar signal processing algorithm has been carried out in parallel with a number of techniques obtained from the literature. The algorithm was implemented into the CUDA platform and its performance was compared to an earlier version of the weather radar signal processing algorithm. In the final stage of this research, weather radar data processing has been carried out in the form of a scan volume matrix using the pipeline technique. By using this technique, the derivative output product of this system can be generated in real time.