Image processing techniques on embedded system

Image processing and embedded systems are used in many applications such as object recognition, robotic applications, autonomous and remote control systems developed for the defense industry, medical applications, face recognition, and vehicle license plate recognition. The size and complexity of the design depend on the algorithms. Therefore, the design must be implemented in the most suitable hardware structure. Especially, Power consumption, desired memory space, and speed of processing are crucial in real-time image processing applications. In the scope of this thesis, image processing applications that can be executed with appropriate embedded system selections have been developed. The Applications were carried out using artificial intelligence models on computer-based software programs, Raspberry Pi, and FPGA embedded system platforms. In the first step, basic image processing algorithms were implemented in different software languages and related embedded platforms. Scientific contributions and innovations of the applications carried out within the scope of the thesis; real-time applicability of image processing algorithms on embedded system platforms was investigated and developed. Deep learning models have been developed with comparative results in order to transform these applications into smart systems. A new normalization technique is proposed for feature extraction on the image and diagnosis of the infected region on x-ray images. The rate of detecting the Covid-19 virus was increased from 83.01% to 96.16% with the proposed normalization technique by using deep learning models to detect the Covid-19 virus in X-ray images. In addition, using the image processing techniques and deep learning model developed within the scope of the thesis, car license plate recognition processes were performed on the images obtained under different conditions, and the performances of these applications were verified with experimental results.