Range clusters based time-of-flight 3D imaging obstacle detection in manifold space

Area of Science:

• Robotics and Computer Vision
• 3D Imaging and Sensing

Background:

• Obstacle detection is crucial for autonomous systems.
• Traditional methods struggle with noise and outliers in 3D range data.

Purpose of the Study:

• To propose a novel obstacle detection method using time-of-flight (ToF) 3D imaging.
• To improve the robustness and accuracy of obstacle detection in noisy environments.

Main Methods:

• Utilizing intensity images to estimate noise deviation in range images.
• Applying weighted local linear smoothing to project data onto a new manifold surface.
• Segmenting 3D imaging data into adaptive range clusters based on pixel distance and ambient relations, with criteria for optimizing cluster shape and size.

Main Results:

• The proposed method effectively reduces the influence of outliers and noise.
• Range clusters are dynamically adjusted for optimal shape and size.
• Experimental results on SwissRanger sensor data demonstrate superior precision compared to traditional methods using regular data patches.

Conclusions:

• The novel range cluster-based obstacle detection method offers enhanced precision.
• This approach provides a more robust solution for 3D imaging-based obstacle detection.