VergenceControl

- Overview and credits

- How to install

- Demo video

- Code explanation

- Binocular energy model

- Vergence control

- Control Evaluation

Vergence Control

The distributed representation of binocular disparity information can thus be exploited to derive an effective control of horizontal and vertical vergence movement using a local weighting of the complex cell responses. Since the meaningful information for vergence comes from the perifoveal part of the image only, we used the response from a spatial neighborhood around the fixation point. A good approzimation of the effectiva rea for vergence control in humans is a spatial neighborhood Ω defined by a Gaussian profile with a standard deviation of 1.5 degrees of visual field.

The Vergence Control can thus be combuted by:

where r_c^ij is the complex cell response at the orientation i and phase difference j.

Thew^ij are the weights derived to obtain a vergence response sensitive to the horizontal component of binocular disparity, and insensitive to its vertical component, and are obtained minimizing the following functional:

Energy Normalization

A desired feature for a real-word system for vergence control concerns the capability of coping with changeable and unpredictable illumination conditions. The illuminationmay not be diffuse but coming from a single and bright source, thus providing dark shades and bright areas in the environment, the light source may move or change of intensity, the object itself may move and tilt with respect to the light source, thus drastically modifying the illumination. Moreover, significant differences might be present between the left and right images, due to imprecision of the two optics or different sensitivity of the sensors, which eventually affect binocular energy approaches. The robustness of the control against these issues has been obtained by implementing a normalization stage.

The response of each complex cell is then normalized by the response of the whole population, by:

where the normalization term is computed as:

For further details, read there papers:

Gibaldi, A., Vanegas, M., Canessa, A., & Sabatini, S. P. (2017). A portable bio-inspired architecture for efficient robotic vergence control. International Journal of Computer Vision, 121(2), 281-302.

Gibaldi, A., Chessa, M., Canessa, A., Sabatini, S. P., & Solari, F. (2010). A cortical model for binocular vergence control without explicit calculation of disparity. Neurocomputing, 73(7-9), 1065-1073.