Fast Threshold Image Segmentation Based on 2D Fuzzy Fisher and Random Local Optimized QPSO

Area of Science:

• Computer Vision
• Robotics
• Image Processing

Background:

• Spacecraft docking requires precise navigation image segmentation.
• Traditional methods like entropy and Otsu criteria are inadequate for complex navigation images.
• Existing 2D Fisher criteria can lead to over-segmentation in these scenarios.

Purpose of the Study:

• To develop a real-time segmentation method for separating target signals from non-target signals in spacecraft navigation images.
• To address the limitations of traditional segmentation methods in complex docking scenarios.
• To achieve a balance between accurate target positioning and preserving fuzzy boundaries.

Main Methods:

• A revised 2D fuzzy Fisher criterion is proposed for image segmentation.
• An integral image based on the 2D fuzzy Fisher criterion is utilized to simplify fuzzy domains and reduce computation.
• A random orthogonal component is incorporated into the quasi-optimum particle to enhance local search capacity and accelerate convergence.

Main Results:

• The proposed method demonstrates effective real-time segmentation of navigation images during the approaching docking stage.
• Experimental results validate the method's ability to precisely segment target signals (bright spots) from the background (space vehicle).
• The revised 2D fuzzy Fisher approach successfully overcomes the segmentation shortcomings of traditional and existing 2D Fisher methods.

Conclusions:

• The novel real-time segmentation method offers a robust solution for spacecraft docking navigation.
• The technique provides accurate segmentation by balancing target localization and boundary preservation.
• The enhanced convergence speed and segmentation accuracy make it suitable for critical space missions.