SFM sparse reconstruction method, system and application

• Main Authors: LI JIAOJIAO, CAO KAILANG, WANG JIAHUI, WANG YANGLI, SONG RUI, LI YUNSONG
• Format: Patent
• Language: Chinese; English
• Published: 29.06.2021
• Subjects: CALCULATING; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention belongs to the technical field of computer vision and multi-view geometry, and discloses an SFM sparse reconstruction method and system and application, and the method comprises the steps: obtaining internal parameters of a camera; extracting point-line features and obtaining initial matching; selecting a feature neighborhood, and constructing a feature triple; performing mismatching elimination in a neighborhood by applying geometric position invariants to obtain an accurate feature matching corresponding relation; obtaining corresponding points on the line features in a neighborhood plane of the line features by using a cross ratio principle, and representing the lines by using the points; scoring the camera pairs, and selecting an optimal initial camera pair as an incremental seed point; arranging point feature representation points and line feature representation points into tracks respectively, and inputting the tracks into a point SFM process to obtain camera poses and sparse point clouds;