FPGA-based Dynamic Image Processing using Zynq 7010
Published:
So after my internsip at IIT Madras, I became interested in Robotics Vision. I wanted to develop a fast image processing system that can essentially replace off-the-shelf cameras like RealSense and Oak-D. However, I was devoid of resources and funds to buy a cheap stereo camera, so I decided I will do it with what I have. I had only one monocular camera(OV7670) and I interfaced it with an FPGA to create a fast image processing logic.
This work was also accepted and published at IEEE International Conference on Advances in Computational Intelligence and Communications 2023 with me as the solo author.
Features
- Uses two Video-Direct-Memory-Access channels instead of one two parallelize the read/write process.
- I wrote a custom Dynamic Clock Generator script in VHDL to replace the Digilent Phase-Locked-Loop IP to adjust frequency on the fly. This made the system latency free.
- This system can now also be used with two OV7670 monocular cameras to create stereo vision that is capable of replacing high-end cameras. This is project is also a testament to my desire to make robots economically feasible to everyone.
Technologies Used
- Reconfigurable Computing
- VHDL/Verilog
- Robotics Vision