FPGA Accelerated Stewart Platform Stabilization Using Xilinx Zynq - OpenHW'24 Submission AOHW-239

Summary

In this showcase for the AMD Open Hardware competition, the presenter introduces an innovative image processing system integrated into the video pipeline of a Pink Z2. Utilizing a Raspberry Pi for HD video capture, the system processes images in multiple stages: first, a smoothing filter reduces noise, followed by edge detection to highlight features. The module then crops unnecessary information, focusing solely on a ball. By analyzing histograms, the system accurately detects the ball's position in real time, allowing for rapid control of the platform using inverse kinematics. This solution offers performance advantages over traditional software methods.

Takeaways

• 🎥 The project showcases an image processing system integrated into the video pipeline of a Pink Z2.
• 📷 A Raspberry Pi captures a full HD video stream at the top of the structure.
• 🖥️ The processed image is displayed on a screen in the middle of the setup.
• 🔧 The system operates through multiple stages of image processing.
• 🧹 The first stage applies a smoothing filter to reduce undesired noise in the input image.
• ✂️ The second stage involves edge detection to identify specific features within the image.
• 🚫 Disabling the smoothing filter reveals increased noise around the image edges.
• 🔍 A cropping module is implemented to remove extraneous information, focusing on the main subject.
• ⚽ The ball's position is detected by analyzing histograms for both axes.
• ⚡ The system achieves a rapid update rate, allowing for real-time control using inverse kinematics to balance the ball.

Q & A

What competition is the submission for?

The submission is for the AMD Open Hardware Competition.

What hardware is used to capture the video stream?

The video stream is captured via a Raspberry Pi.

What is the resolution of the output image?

The system outputs a full HD image.

What is the first stage of the video processing system?

The first stage is the smoothing filter, which removes undesired noise from the input image.

What technique is used for feature detection in the images?

Edge detection is used to detect certain features in the image.

What issue arises when the smoothing filter is disabled?

When the smoothing filter is disabled, there is a significant increase in noise around the edges of the image.

How does the system handle additional unwanted information in the image?

The system includes a cropping module to remove unwanted information and focus on the object of interest, such as the ball.

How does the system detect the position of the ball?

The position of the ball is detected by determining the histograms for both axes and finding the center of the lines that represent the ball.

What happens when the ball is moved?

As the ball is moved, its position updates quickly on the screen, demonstrating the system's fast response rate.

How does this video processing system compare to software solutions?

The video processing system offers an incredibly fast update rate, which is unmatched by many software solutions.

What application does this video processing system support?

It allows control of the ST platform using inverse kinematics to balance the ball.