Ship Ranging Method in Lake Areas Based on Binocular Vision.

Area of Science:

• Computer Vision
• Robotics
• Navigation Systems

Background:

• Traditional ranging algorithms struggle with complex lake environments and catamaran hulls.
• Inaccurate depth information near ships impairs electric boat navigation and increases computational load.

Purpose of the Study:

• To develop an improved ranging method for accurate depth perception of ships in lake environments.
• To enhance the navigation of electric tourist boats by addressing limitations of existing stereo vision techniques.

Main Methods:

• Integration of improved Oriented FAST and Rotated BRIEF (ORB) feature detection with stereo vision.
• Utilizing a local feature weighting approach and quadtree structure for feature point refinement.
• Enhancing feature matching accuracy with Fast Library for Approximate Nearest Neighbors (FLANN) and Progressive Sample Consensus (PROSAC) algorithms.
• Applying triangulation principles for target distance and angle calculation.

Main Results:

• The proposed algorithm achieved a 6.5% improvement in processing speed compared to ORB-PROSAC.
• Ranging errors were 2.25% at 10m and 5.56% at 20m under ideal conditions.
• The method demonstrated effectiveness in compensating for stereo matching shortcomings in specific lake scenarios.

Conclusions:

• The novel ranging method offers a viable solution for depth perception challenges in lake navigation.
• Improved ORB and stereo vision integration enhances navigation safety and efficiency for electric boats.
• This approach provides a valuable alternative for addressing the "gap period" in stereo matching-based ranging.