代表系統Featured Systems
次世代檢測：6D 姿態估測結合雙臂瑕疵檢測系統Next-generation inspection: 6D pose estimation for dual-arm defect detection
回到專案總覽Back to project archive6D pose 與 PSO refinement workflowProject image 1 
姿態校正與瑕疵檢測結果對照Project image 2
面向反光金屬工件的雙臂 AOI 系統，將 6D 姿態估測、瑕疵檢測、座標校正與雙臂檢測流程整合到同一條 pipeline。A dual-arm AOI system for reflective metal parts, integrating 6D pose estimation, defect detection, coordinate calibration and robot inspection into one pipeline.
專案內容Project scope
- 團隊建立面向反光金屬工件的雙臂 AOI 系統，結合 6D 姿態估測、YOLOv7 瑕疵檢測與雙臂檢測流程。
- 團隊系統測試最高約 95% 檢測準確率並可辨識 1 mm 瑕疵；此量化結果以團隊成果呈現。
- 使用 3D CAD 模型產生合成訓練資料，並以 PSO、手眼座標轉換與姿態矩陣計算對接機器人檢測動作。
- The team built a dual-arm AOI system combining 6D pose estimation, YOLOv7 defect detection and robotic inspection.
- Team testing reached approximately 95% detection accuracy and identified defects down to 1 mm; these figures are presented as team results.
- 3D CAD models generated synthetic training data, while PSO, hand-eye calibration and pose matrices connected perception to robot motion.
我的工作範圍My contribution
我主要負責Owned
- 6D 姿態估測模型訓練。
- 3D CAD 合成資料生成與自動標註流程。
- PSO 姿態校正、手眼座標轉換與相機/物件姿態矩陣計算。
- 雙臂檢測流程整合。
- ChatGPT / LINEBot 文字、影像、語音互動介面。
我協作的部分Collaborated
- 專案整體為團隊合作，YOLOv7 瑕疵檢測與量化檢測成果以團隊系統成果呈現。
非我主要負責Not my primary scope
- 動態相機或光源視角規劃尚未確認為本人主責。
Owned
- 6D pose-estimation model training.
- Synthetic-data generation and automatic annotation from 3D CAD.
- PSO pose refinement, hand-eye calibration and pose-matrix computation.
- Dual-arm inspection-flow integration.
- ChatGPT and LINEBot text, image and voice interfaces.
Collaborated
- The overall project was a team effort; YOLOv7 defect detection and quantitative testing are presented as system-level team results.
Not my primary scope
- Dynamic camera or lighting viewpoint planning has not been confirmed as my primary responsibility.
專案影像Project evidence
技術與回顧Technology and review
- 技術深度Technical depth
- 待評分Not rated
- 個人主責Personal ownership
- 待評分Not rated
- 展示成熟度Presentation maturity
- 待評分Not rated
- 回顧價值Review value
- 待評分Not rated