--- title: 'openCV tracker 研究低fps bounding box 偏移解決方法' disqus: hackmd --- penCV tracker 研究低 fps bounding box 偏移解決方法(ubuntu18.04) === 文件版本.:v0.0.1 [TOC] ## 暫時筆記 Weekly report last week: (1) win10: 測試 multi-core tracking failed main process 與 child process memory 不共用 導致全域的 tracker object 無法再 process pool 內使用(先不要解) 30fps 跟5 fps做結合比較 (2) bounding box shifted issue ## 1. 筆記解說 目前測試低fps(5fps),會有 bounding box 偏移的現象,如下影片 https://youtu.be/5ZPR2aXiDEM 若是 使用 29.97 fps 無此現象 https://youtu.be/vu9Y9_2dlnM 如下圖可知其實在整體畫面 bounding box 的偏移不大,而是畫面移動太快  畫面移動,追蹤目標跑到右手邊了   ## 2. 測試 source code 至此 [github](https://github.com/masteree108/bbox_shifted_research) 可下載 測試程式 s選完一個人物後按下 Enter, 然後選下一個按一次 Enter 最後按下 ESC 讓程式繼續運作 ### 5 fps 測試方法 如下圖選擇 07_clip_5fps.mp4  ``` $ conda activate your_python_container $ ./run.py ``` ### 29.97 fps 測試方法 如下圖選擇 07_clip_30fps.mp4  ``` $ conda activate your_python_container $ ./run.py ``` ## 3.找出畫面偏移量 [參考論文](https://www.researchgate.net/publication/337261122_UAV_Real-Time_Video_Stabilization_Using_Opencv_Technical_analysis) A. Quick and Rich Feature Tracking (1) Grid-Based Feature Detection: (2) Integrating Feature Tracking With Feature Matching B. 2D Stabilization ### Lucas-Kanade 算法(先不要弄) detected points  tracking points1  tracking points2  ###### tags: `study`, `VoTT`
×
Sign in
Email
Password
Forgot password
or
By clicking below, you agree to our
terms of service
.
Sign in via Facebook
Sign in via Twitter
Sign in via GitHub
Sign in via Dropbox
Sign in with Wallet
Wallet (
)
Connect another wallet
New to HackMD?
Sign up