Computer Vision

Computer Vision

Science and technology that enables machines to “see.” This can range from simple light detection, like my garbage door opener’s safety system, to advanced systems that recognize objects and interpret events.

Computer Vision Two Branches

1. 1st focuses on developing systems that extract information for images, advancing theory and algorithms

2. 2nd is more applied, using cameras & sensors to control machines, such as robots or autonomous vehicles

Autonomous Vehicle: Case Study

These be chole’s rely on sensors their surroundings in real time. U.S. Dept of Transportation National Highway Traffic Safety Administration defines 6 levels of vehicle autonomy from level 0 (fully human driven to level 5 (Completely auto under all conditions. Despite the progress, commercial vehicles remain Level 3 or 4.

Level 1 Vehicle Autonomy

Basic driver assistance- such as cruise control- automates 1 task, but the drive handles everything else

Level 2 Vehicle Autonomy

The vehicle can street & control acceleration/deceleration, but the driver must remain alert & ready to take control

Level 3 Vehicle Autonomy

The vehicle can drive itself in controlled environments, such as highways, but the driver must take over in complex situations

Level 4 Vehicle Autonomy

The vehicle can operate autonomously in well-mapped areas like urban centers but still requires human intervention in extreme conditions, such as severe weather or unpaved roads

Controlled Environment & Human Oversight

Controlled conditions are necessary because today’s computer vision systems struggle with the unpredictability and complexity of real world scenarios. Many autonomous vehicles rely on remote operators to take control.

Capabilities Autonomous Vehicles Need

Vehicles need to predict and react to their surroundings like human drivers intuitively do. They must anticipate pedestrian movements, adjust to weather changes and interpret road signals while coordinating with other vehicles & infrastructure.

Hawk Eye System

Uses multiple high speed cameras to track position of tennis ball and determine whether serve is in and out. Has predicative ability, uses trajectory data from immediate preceding moments to esti where the ball is going to next

Hawk Eye System & Human Vision

Human perception often relies on prior experience to predict where an object will go. Hawk Eye is an esti, just like human perception based on inferences from incomplete visual data.

PITCHf/x

Tracks pitched ball speed, trajectory and spin. Using multiple high speed cameras around stadium

Connecting Human & Machine Vision

Human brain relies on regulators learned from experience such as assumptions that light comes from above or that objects in motion tend to stay in motion

Robotic & Computer Vision

Robots use vision to navigate & interact with their environment. Ex. Flippy use computer vision to monitor cook times can also track nearby humans