Status Report 4(10.13-10.19)

Team

Most Significant risks and contingency plans

  1. No risks

Changes made

  1. No changes made

Schedule Update

  1. Schedule on time, we made a lot of progress this week to bring schedule back on track

Ruohai Ge

  1. Abandoned the OpenCV Kalman filter and try to write my own with the self-defined dynamic system, the algorithm and dynamic system is shown in the figure below
  3. Find a more suitable video to test
    1. Get its fps by calculating how many times OpenCV takes to show 120 frames. fps = frames/time
    2. Get the pixel per inch in order to get the acceleration(9.8m) in pixel
  4. Rewrote the pipeline of Kalman filter trajectory prediction
    1. Implement my own Kalman filter
    2. Define the system model, fps and acceleration in pixel
    3. Write the new pipeline to update the trajectory based on new observed points
    4. Able to get adjusted and predicted trajectories instead of single points
    6. Schedule situation: on time
    7. Deliverable for next week
      1. Try to integrate the whole vision pipeline with Zheng
      2. Try to access 3d coordinates

Xingsheng Wang

  1. Solved the motor controller problem. The mobile platform ordered does not come with controllers and the controllers provided by the mobile platform company had to be purchased separately. The controller board was manufactured in China so it will take weeks to arrive. Luckily, a friend of mine is coming to Pittsburgh this week and he will bring the board for me. The board will arrive on Saturday afternoon so I am still on schedule.
  2. I am continuing my work on the coding part of the mobile platform. The mobile platform company has provided sample code but there need to be modifications to the code since not every component on the board is used and I need to add our code for the trash can to the skeleton code they provide.
  3. Last week I was a bit behind schedule but now I am back on schedule. Code modification is in progress and testing will start soon, as is indicated on our Gantt chart.

Zheng Xu

  1. Accomplishments
    1. Tried various tracking/recognition options to decide which is the best option.
      1. OpenCV trackers. The OpenCV trackers cannot track fast moving objects and thus is not suitable for the tracking part of this project.
      3. Optical Flow. The optical flow does not work as expected in our implementation. However, I would like to spend an extra hour or two to further explore this idea.
      5. Camshift. Camshift still need some work to improve, as it is not robust in our test. I will spend the next two days working on it, including combining it with background removal.
    2. Schedule situation: on time
    3. Promises for next week:
      1. Further improve object tracking to get improved results.
      2. Use tracking information to get 3D coordinate information.
      3. Work on transformation from camera coordinate to trashcan coordinate.