In computer vision, which technique explicitly relies on image templates (stored exemplars) to recognize or locate objects in scenes?

Introduction / Context:
Computer vision encompasses many approaches to understanding images, from low-level feature extraction to high-level object recognition. Some approaches match observed image data to previously stored exemplars or “templates.” This question asks you to identify which option describes a technique that relies on such templates to interpret a scene.

Given Data / Assumptions:

• The technique should use templates or exemplars of object appearance/shape.
• We contrast it with methods that focus on stereo geometry, basic edges, or the broader application of vision to robotics.

Concept / Approach:
Model-based vision typically uses known object models (which can be geometric or appearance templates) to explain observed image evidence. The system compares features or image patches to stored templates of the target object to hypothesize presence, pose, or segmentation. Template matching is a classic component here. By contrast, edge detection is a low-level operation producing gradient-based edge maps; binocular vision (stereo) estimates depth via disparity; and robot vision is an application context, not a specific template-driven technique.

Step-by-Step Solution:

Verification / Alternative check:
Standard texts describe template matching and model fitting (2D/3D) as canonical model-based methods for recognition and pose estimation, often using correlation, chamfer matching, or feature-based alignment to templates.

Why Other Options Are Wrong:

• Edge detection: extracts gradients; does not use stored templates by itself.
• Binocular vision: computes disparities; template use is not central.
• Robot vision: context of use, not a particular technique.
• None of the above: incorrect because model-based vision fits.

Common Pitfalls:
Equating any recognition with “robot vision”; assuming edge maps themselves are templates rather than inputs for later matching stages.

Final Answer:
model-based vision