研究目的
To achieve autonomous aerial refueling by solving the position and orientation of the drogue based on stereo vision.
研究成果
The proposed method effectively tracks and positions the drogue for autonomous aerial refueling using stereo vision, demonstrating feasibility and reliability. Future work includes improving matching algorithm accuracy and parallelizing the algorithm for efficiency.
研究不足
The accuracy in the depth direction needs improvement. The method requires further parallelization for efficiency and hardware migration.
1：Experimental Design and Method Selection:
The method is decomposed into three stages: initialization, tracking, and calculation. A pre-training model and binocular constraint are used for initialization. Information fusion locates the drogue in each pair of images. Key points matching in binocular vision calculates the center point of the drogue's ring region.
2：Sample Selection and Data Sources:
Two video streams acquired by two fixed cameras are used as inputs.
3：List of Experimental Equipment and Materials:
Binocular vision system consisting of two GT1920 cameras with resolution 1936 × 1456, implemented in C++ with OpenCV 2.4.10, on a PC with Intel i7-6820HQ CPU, 16GB RAM, and NVIDIA Quadro M2000M GPU.
4：10, on a PC with Intel i7-6820HQ CPU, 16GB RAM, and NVIDIA Quadro M2000M GPU.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Initialization locates the target's initial state. Tracking continuously positions the drogue using an online tracker, offline model, and motion estimation. Calculation solves the center point coordinates from key points extracted in the drogue's ring region.
5：Data Analysis Methods:
The method's effectiveness is demonstrated through experiments on ABB robots, evaluating the RMS of errors in horizontal, vertical, and depth directions.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
