Target-positioning method for a four-aperture infrared bionic compound eye based on the reprojection Kalman filter...

Area of Science:

• Optics and Photonics
• Robotics and Automation
• Computer Vision

Background:

• Single-aperture infrared cameras have limited fields-of-view.
• Existing multi-aperture systems often neglect sub-eye pose differences, leading to slow convergence and reduced accuracy.
• Bionic compound eye systems offer potential for enhanced imaging and positioning.

Purpose of the Study:

• To develop and validate a target-positioning method for four-aperture infrared bionic compound eye systems.
• To improve target-positioning accuracy and convergence speed by accounting for spatial and pose variations among sub-aperture cameras.
• To address limitations of traditional infrared imaging systems.

Main Methods:

• Captured static and dynamic target images using a four-aperture infrared bionic compound eye system.
• Developed a target-positioning method incorporating spatial and pose weights.
• Employed a reprojection method for camera relationship establishment and a Kalman filter with spatially weighted projection for all camera combinations.
• Performed single-aperture and dual-target calibrations.

Main Results:

• Static target-positioning achieved horizontal error of 1.76%, vertical error of 2.44%, and depth-estimation error of 0.40% at 2660 mm.
• Ablation studies confirmed the importance of spatial weighting.
• Dynamic target-positioning error was controlled within 2.5% at 2840 mm in the overlapping field-of-view.

Conclusions:

• The proposed target-positioning method significantly enhances accuracy in four-aperture infrared bionic compound eye systems.
• Accounting for spatial and pose differences is crucial for optimal performance.
• The system demonstrates robust capabilities for both static and dynamic target tracking.