ZYNQ Lidar-Visual SLAM Algorithm Porting and Design

FPGA-accelerated stereo vision 3D reconstruction on ZYNQ, generating real-time depth maps and rainbow maps.

1. Completed stereo image acquisition and PS+PL collaborative processing on ZYNQ.
2. Completed camera calibration and lens correction on ZYNQ.
3. Completed stereo matching algorithms on ZYNQ, currently implementing local stereo matching BM and semi-global stereo matching SGBM.
4. Completed the conversion of post-stereo matching depth maps to rainbow maps (i.e., colorized depth maps) on ZYNQ.
5. Also completed the ranging algorithm on ZYNQ, but currently without a recognition algorithm, it can only calculate distance from a single point in the image. Accuracy is within 10mm. The effective range is currently tested within 5m.
6. Optimized all algorithms, currently achieving near 720p image quality with a frame rate of around 30fps.

A dedicated column will be published detailing the implementation process, updated weekly, covering the building of a real-time 3D reconstruction project on ZYNQ from scratch.

![](https://pub-048dcb96257f476697b113fcb5939cb9.r2.dev/blog/124908639/01_2be6c0c91b2055cbfd48d