Modeling heading and path perception from optic flow in the case of independently moving objects

Area of Science:

• Visual perception
• Computational neuroscience
• Robotics

Background:

• Humans exhibit robust heading and path estimation from optic flow.
• Independently moving objects (IMOs) can bias heading perception, especially when large and obscuring the focus of expansion (FOE).
• Previous models struggled to fully explain human robustness against IMOs.

Purpose of the Study:

• Investigate the mechanisms underlying human robustness in optic flow-based navigation.
• Identify and model segmentation cues that enable robust perception in the presence of IMOs.
• Develop a unified computational model explaining human psychophysical data.

Main Methods:

• Derived analytical models of optic flow for linear and curvilinear self-motion.
• Developed a linear least squares method for heading estimation based on optic flow models.
• Modeled optic flow segmentation cues: accretion/deletion, expansion/contraction, acceleration/deceleration, and curvature.
• Integrated segmentation cues into the heading and path estimation method.

Main Results:

• Analytical models revealed significant heading biases for a basic least squares method, exceeding human performance.
• Identified specific optic flow cues (accretion/deletion, curvature, etc.) crucial for segmenting IMOs.
• The integrated model incorporating segmentation cues successfully explained human psychophysical data on heading and path perception.

Conclusions:

• Optic flow segmentation cues are critical for robust heading and path perception in complex visual scenes.
• These cues allow the visual system to effectively segment IMOs, mitigating their influence on self-motion estimation.
• The findings unify previous research and provide a computational basis for human navigation robustness.