A group of students have created hardware to eliminate fog in smart cameras, contributing to ensuring traffic safety for people.
In bad weather conditions such as fog and dust, the captured images can be blurry and unclear, reducing visibility and making it difficult to identify objects, leading to many accidents every year. This places high demands on the development of effective solutions to improve image quality and increase visibility in extreme weather conditions. That is why the student group came up with the idea of a hardware design project to remove "fog" in smart camera images.
A group of students and their instructor (right cover) received the award in the competition.
THAI PHUC
The project owner is a group of 4 students from Ho Chi Minh City University of Technical Education, including: Nguyen Huy Hoang, Phan Minh Nhat, Tran Tuan Kiet, Pham Nguyen Hoang Hai. After having the idea, the group spent about 6 months to complete the product. The project won second prize at the first "Smart City Microcircuit Design" contest in 2023. Huy Hoang shared: "The product can be used as a processing premise for AI applications in object recognition. The fog removal application will play a role in cleaning data, enhancing image quality and removing noise. This improves the accuracy of AI models". In terms of superiority, compared to traditional input image enhancement methods, which are less effective because they have to calibrate basic image parameters such as brightness, contrast, color saturation, etc., the dark channel method allows deeper intervention by assuming and re-mapping the field of objects in the image. Therefore, the image quality is improved more clearly and has reasonable processing performance and power consumption compared to methods using machine learning and deep learning techniques.
Product design and simulation results
THAI PHUC
Sharing about the implementation process, Huy Hoang said: "First, the group looked for a suitable de-fog algorithm. Our group chose Dark Channel Prior, this algorithm ensures short execution time and relatively high efficiency. Then, the group completed the code using the Vitis HLS library to use a third party Xilinx - Vitis HLS, which can be converted to Verilog language. Xilinx's function can also support us to initialize Generate to create a Testbench test set to ensure that the group's code executes correctly when converted to the new language. Finally, the group will use Synopsys's tool to complete the chip. After being trained and guided by members of Sunedu and ADT company, we were able to execute it." Assessing the feasibility of the project, Master Truong Huu Ly, Deputy Director of the Research and Development Center, Ho Chi Minh City High-Tech Park, said that the model of the group is currently of great interest to businesses, especially Truong Hai Auto Corporation. Mr. Ly also said that this project has great potential if further developed. For example, when driving in the fog, without observing from the outside, the group's camera system will recognize and capture it. If you develop the software to display on the windshield, it will be easier to observe instead of looking directly. But developing on a small screen like the current one will cause distraction for the driver. "In general, this project will have a lot of potential for development, especially in the automotive engineering industry," Mr. Ly emphasized. Huy Hoang added that in the near future, the group will try to complete the integration into an ASIC block so that it can be inserted into the camera like a car camera. From there, the driver can monitor the image and ensure safety when participating in traffic. In addition, in the future, the group plans to develop hardware that can filter images from security cameras in extreme weather conditions. This will better ensure public order and safety for the people.
Comment (0)