PURE VISION SENSOR

Pure vision sensors use cameras for perception only and they rely on visual data from them.

There are several types of pure vision sensors and they are; monocular, stereo, RGB-D, wide angle/fish eye, event, thermal and light field (Plenoptic) cameras.

Monocular pure vision sensor camera is a single camera that captures 2D images. They are simple low cost cameras with limited depth information. They are quite suitable for mobile and surveillance applications. Example includes; standard smartphone cameras etc.

Stereo pure vision sensor cameras are two cameras with depth perception. They are able to collect depth information and 3D reconstruction of the scene. They are complex and require calibration to operate efficiently. They find widespread application in robotics, AR/VR etc.

RGB-D pure vision sensor cameras capture color and depth data of the scene. They possess 3D data capability and gesture recognition. They have a limited range and outdoor issues. They are typically used in gaming, robotics, AR etc. examples include; Microsoft kinect, Intel real sense etc.

Wide-angle pure vision sensor cameras have a broad field of view (e.g. Fish eye). They are capable of capturing more scene but at the price of distortion and lower resolution. They find application in surveillance and VR. Examples include; security cameras and some action cams.

Event pure vision sensor cameras capture changes of events at low latency. They are fast and efficient for motion scene capture, but they have limited static information. They find application in robotics and high speed tracking. Examples include; Davis camera etc.

Thermal pure vision sensor cameras capture heat from the scene and works excellently in darkness. They can capture superb images in low light and heat. However their major disadvantage is their low resolution and high cost. They also find application in surveillance and search and rescue. Example incudes; FLIR cameras etc.

Light field/Plenoptic pure vision sensor cameras capture the light direction and are refocusable. Their main advantages are that they can post capture focus and depth. They are complex and of lower resolution. They find application in photography and computational imaging.

The advantages of pure vision sensor cameras are; they provide non-contact sensing, that is there is no physical interaction with the environment. They obtain rich data from scenes, that is they capture detail visual information (eg. Color, texture, etc.). They are versatile, that is they provide various types of cameras such as event, RGB-D, IR cameras etc. they are also low cost and can operate at low power and are very energy efficient.

The disadvantages of pure vision sensors cameras are; they are dependent on light as the medium of capture hence their performance varies with light conditions. They have complex processing needs, requiring time, heavy computation and human expertise. Their need for data interpretation requires algorithms for meaningful insights.

Pure vision sensor cameras find applications in the following; surveillance for security and monitoring. Robotics for navigation and object recognition. AR/VR for immersive experiences. Automotive; for driver assistance and autonomous vehicles. Healthcare; for diagnostics and patient monitoring.

The future of pure vision sensor cameras depends on the advances and development of the following technologies; AI integration such as incorporating enhanced processing and object recognition in camera capabilities. Edge computing to enable more onsite device processing. Advancing pure vision sensor research and technologies in areas as diverse as better sensors materials with more efficient and broader applications.

 

SOURCES:

 

• Computer vision: Algorithms and applications by Richard Szeliski.
• Multiple view geometry in computer vision by Richard Hartley and Andrew Zisserman.
• Computer vision: A modern approach by David A. Forsyth and Jean Ponce.
• Digital image processing by Rafael C. Gonzalez and Richard E. Woods.
• Event-based neuro-morphic vision edited by Teresa Serrano-Gotarredona and Bernabe Linares-Barranco.