CLCFM3: A 3D Reconstruction Algorithm Based on Photogrammetry for High-Precision Whole Plant Sensing Using All-Around...

Area of Science:

• Plant Science
• Computer Vision
• Photogrammetry

Background:

• Accurate 3D plant phenotyping is crucial for understanding plant growth.
• Plant structures with overlapping parts (leaves, stems) pose challenges for 3D reconstruction.
• Occlusion and image-matching errors lead to incomplete or inaccurate 3D point clouds.

Purpose of the Study:

• To develop a novel technique for acquiring high-density, high-precision 3D point cloud data for plant phenotyping.
• To address challenges of occlusion and erroneous points in 3D plant reconstruction.
• To enable large-scale comparative analysis of plant phenotypes using 3D information.

Main Methods:

• Proposed a closed-loop coarse-to-fine method with multi-masked matching (CLCFM3).
• CLCFM3 utilizes repeated local point cloud generation (multi-matching) to suppress occlusion.
• Employed masked matching to remove noise points and a closed-loop coarse-to-fine method (CLCFM) for accurate Structure from Motion.

Main Results:

• The CLCFM3 method effectively reconstructs high-density, high-precision 3D point clouds for complex plant structures.
• Suppression of occlusion and reduction of erroneous points were achieved.
• Improved accuracy in Structure from Motion through CLCFM enhances overall 3D data quality.

Conclusions:

• The developed photogrammetry technique enables efficient acquisition of detailed 3D plant data.
• This approach facilitates comparative analysis of plant phenotypes across various species during growth.
• The method is expected to advance plant science research through precise 3D phenotyping.