Automated Geometric Calibration of Spatial Augmented Reality for Architectural Design Visualization

Abstract

Spatial Augmented Reality (SAR), A Branch Of Augmented Reality (AR), Has Great Potential For Collaborative Architectural Design Visualization By Projecting Building Information Modelling (BIM) Images Onto Real Objects (E.G., Small-Scale Mock-Up Models). In SAR, The Building Surface Texture Is Pre-Produced And Assigned To Projectors Such That The Texture Matches With The Physical Building Model In The 3D Real World. To Enable SAR Registration, Geometric Calibration For The Alignment Between Projected Images And Physical Models Is Essential. Previous Research Efforts Have Proposed Various Approaches For The Geometric Calibration Of SAR, Ranging From Manual Calibration Of Reference Points To Automatic Projection Mapping. Although the Marker-Based Calibration Method Has Been Developed For AR, It Is More Challenging To Apply To SAR. Using Geometric Calibration Information, The Output Display Panel Of SAR Is In The Real World, While Its Outcome Display Remains On A 2D Screen Of AR. To Tackle This Challenge, We Propose A Marker-Based Approach For SAR-Based 3D BIM Visualization, Which Uses A Projector-Camera System And A Marker Detection Method To Achieve Automatic Geometric Calibration And Mapping On 3D Building Models. Markers Are Attached On The Corners Of The Model’s Detected Surfaces, And The Projector-Camera System Can Be Easily Placed On Any Position Within The Shooting And Projecting Range. We Used A Series Of Computer Vision Methods For Image Processing And Object Detection, Including Perspective Distortion Correction, Image Segmentation, Position Estimation, And Projective Transformation. We Also Conducted An Experimental Test To Perform And Evaluate The Whole Process And The Outcome Model Appearance Intuitively. Moreover, We Validated Registration Accuracy By Measuring The Projection Error. The Preliminary Result Showed Satisfying Accuracy Of The Projection On The Front Surface Of The Model And the Realness Of SAR-Based 3D BIM Visualization. However, There Was a Test Error On The Side Surfaces That Need To Be Further Eliminated, And Potential Solutions Were Proposed. The Proposed Method Involved An Alternative For 3D BIM Visualization In A Shared Environment To Facilitate More Efficient Architectural Design Collaboration Between Participants.