Efficient 3-D scene analysis from streaming data

Rich scene understanding from 3-D point clouds is a challenging task that requires contextual reasoning, which is typically computationally expensive. The task is further complicated when we expect the scene analysis algorithm to also efficiently handle data that is continuously streamed from a sensor on a mobile robot. Hence, we are typically forced to make a choice between 1) using a precise representation of the scene at the cost of speed, or 2) making fast, though inaccurate, approximations at the cost of increased misclassifications. In this work, we demonstrate that we can achieve the best of both worlds by using an efficient and simple representation of the scene in conjunction with recent developments in structured prediction in order to obtain both efficient and state-of-the-art classifications. Furthermore, this efficient scene representation naturally handles streaming data and provides a 300% to 500% speedup over more precise representations.

Authors

• Hanzhang Hu
• Daniel Munoz
• J. Andrew Bagnell
• Martial Hebert

Keywords

• Image analysis
• Algorithm design and analysis
• Data structures
• Prediction algorithms
• Robot sensing systems
• Inference algorithms
• Data Streams
• Scene Analysis
• Structure Prediction
• Point Cloud
• Simplified Representation
• Mobile Robot
• Efficient Representation
• Scene Understanding
• Order Prediction
• Scene Representation
• Precise Representation
• Data Structure
• Computation Time
• 3D Space
• Bounding Box
• Segmentation Algorithm
• Levels Of Hierarchy
• Local Map
• Hash Function
• Ground Truth Labels
• Number Of Grids
• Neighborhood Context
• Global Grid
• Overlap Region
• Inference Procedure
• Finest Level
• Validation Folds
• Single Grid
• Inference Algorithm
• Feature Calculation