Realization of SURF algorithm based on Vivado HLS for FPGA platform

The SURF algorithm is a method which is used in image processing that is not affected by feature changes such as size, color and contrast. In this study, the SURF algorithm which has a lot of processes, was created with Vivado HLS tool. HLS can be implemented using languages as C, C++, etc. without using hardware description languages such as verilog, VHDL, etc. The SURF algorithm was implemented with a C-language. Hardware optimization of system has been implemented with directives in HLS. In this context, the data type has been chosen as "fixed-point" instead of "floating-point" for the faster system and less storage space. In addition, the runtime was reduced by executing the appropriate code blocks by the parallel operation. "Genetic Algorithm" has been used to determine the parameters which are affected the success of the SURF algorithm. It has been observed that the success rate significantly increased when the determined parameters were used. The proposed method has been tested for the character recognition application with the fixed font which is independent from the "rotation and character size". In this study, the tests were carried out with the images formed by the numbers. Reference picture and test pictures for all figures were created. Test pictures with different sizes and rotations (± 10) were selected. Thus, it was aimed that the reference and test pictures was not the same. The appropriate parameter values were used in HLS after being determined with matlab. The rate of high success were obtained using the proposed optimized SURF structure.