This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

AM5728: Optimization of 3D vision system

Part Number: AM5728
Other Parts Discussed in Thread: TIDEP0076

It takes twice as long as windows to transplant 3D vision system to am5728, scan and detect objects, and generate 3D point cloud data.

Now I want to optimize the system on am5728. What's the good advice of Ti experts?

1) Because I'm not familiar with DSP, the current 3D vision system is running on the cotex A15. I don't know which parts can be run on DSP?

2) At present, before generating point cloud data, it is implemented with functions in opencv library. In the system, I did not configure the environment variable OPENCV_OPENCL_DEVICE. Does it mean that the optimized Ti opencv is not running? How do I use the optimized Ti opencv library? Is it to configure export OPENCV_OPENCL_DEVICE=TI AM57:ACCELERATOR:TI Multicore C66 DSP' That's it? What do I need to do on DSP?

3) There is GPU function on am5728. How can I use GPU? Do I call the interface in opengles2.0, which means I have used the function of GPU?

4) Point cloud display is currently implemented by QT virtualization module? The effect is not very good. It is implemented with VTK library on windows. The library of VTK is only based on X11 on Linux, and the file system of am5728 is based on Weston, so VTK cannot be used. What other libraries can point cloud display run on Weston?

5) After the point cloud data is generated, the point cloud data needs to be processed. At present, the PCL library is used to realize the point cloud data processing. It is slow for am5728 lines to use PCL library to process the point cloud data. How to optimize the PCL library on am5728? Need to transport PCL library to DSP?