Search

Browse

• Communities
• Authors
• Year
• Titles
• Subjects

My Space

• Login

Reconfigurable Logic-Based system for image processing of fishery nets

Zacheilas Theofilos

Full record

---

Summary

The aim of this thesis is to develop an embedded system which is able to detect the deformed holes of fishery nets in an input video. In recent times, there is a need for production of automation systems which save time and manual efforts of human resources. The simulation of the system is done in MATLAB, whereas the final design is implemented and accelerated in Vivado HLS tool. The produced hardware component can fit into an actual FPGA platform (ZCU102), in order to accelerate the corresponding software part with less energy consumption. There are two main topics that affect the appearance of aquatic environment, the lighting variation with haze and the variation in size of holes along each frame. There is a progressive evolution on the production of final results. Firstly, the use of an edge-aware filtering, called 'guided image', takes place to face the haze removal and the lighting smoothing of each frame. Furthermore, morphological opening is used for detail enhancement. The image, also, breaks in windows for local identification of deformed holes and overcoming the problem in variation of holes' size along different parts of the frame. Finally, combination of consecutive frames happens for further improvements. Speaking with numbers, the dimensions of each frame are 270(height) x 480(width), and hardware achieves an average of 181 frames per second, and a speed-up of about 12.7 times compared to MATLAB. The significance of this study is that it can actually cope with each of the obstacles through a series of observations, help the user understand the circumstances and become the base for the creation of a real time system that operates well in every occasion.

Available Files

Services