Traffic Event Classifucation

Traffic Event Classifucation

Jay Shah

Jay Shah

Ahmedabad, Gujarat

The main idea in this work would be to extract high level of abnormal traffic event characteristics, such as severities and our initial focu

Virtual Reality, Robotics, Artificial Intelligence

Description

There are many city-owned cameras in public spaces, and the office collaborates and they share videos with other regional public safety agencies. Video analytics from these cameras can be used for traffic control, retail store monitoring, surveillance and security, as well as consumer applications including digital assistants for real-time decisions. The basic analysis will involve things like recognizing when a vehicle is traveling the wrong way on a one-way street. Also, it will have differences where walking on the grass might be aberrant in some situations but normal in a backyard of the house. Although, it's not practical to watch the feeds from all those cameras in real time. Instead, we can build models that reflect how people/objects under surveillance behave, based on that, predict deviations from the norm.

The system can be used to produce directional traffic volumes and on track to identify dangerous conflict patterns towards minimizing traffic deaths. We focus on traffic events at road intersections. Vehicle trajectories are clustered and common routes are learned in order to extract the traffic model at intersections. When an abnormal vehicle event is observed, the severity of this event is indicated by severity classification

Large cities have hundreds or thousands of traffic cameras installed and our extensive conversations reveal that they are very interested in analyzing video from these cameras for traffic safety and planning. Detecting “close-calls” between cars, bikers, and pedestrians helps preemptively deploy safety measures. To help install crosswalks, cities would like to detect areas where jaywalkers cross streets. To improve congestion, they would like to detect double-parked cars but exclude first-responders. Counting volumes of cars, pedestrians or bikes feeds into the traffic light controller to appropriately control the durations.

A significant part of the video surveillance program going forward will be video analytics, computer algorithms written to automatically alert officers to possible terror attacks or criminal activities. Both real-time monitoring and postevent forensics can greatly benefit from the development of autonomous video analytics tools. Today, this development is proceeding in three directions that are simultaneously pursued in both signal processing and computer vision communities

Video analytics is not limited to only traffic surveillance but potentially central to a wide range of future and existing applications ranging from surveillance and self-driving cars, to personal digital assistants and even drone cameras.

Medium 19510414 1414059632014333 5949366042080012019 n

Jay S. created project Traffic Event Classifucation

Medium 504b6ba4 c6e4 4b6c a9c0 e4c674c3a229

Traffic Event Classifucation

There are many city-owned cameras in public spaces, and the office collaborates and they share videos with other regional public safety agencies. Video analytics from these cameras can be used for traffic control, retail store monitoring, surveillance and security, as well as consumer applications including digital assistants for real-time decisions. The basic analysis will involve things like recognizing when a vehicle is traveling the wrong way on a one-way street. Also, it will have differences where walking on the grass might be aberrant in some situations but normal in a backyard of the house. Although, it's not practical to watch the feeds from all those cameras in real time. Instead, we can build models that reflect how people/objects under surveillance behave, based on that, predict deviations from the norm.

The system can be used to produce directional traffic volumes and on track to identify dangerous conflict patterns towards minimizing traffic deaths. We focus on traffic events at road intersections. Vehicle trajectories are clustered and common routes are learned in order to extract the traffic model at intersections. When an abnormal vehicle event is observed, the severity of this event is indicated by severity classification

Large cities have hundreds or thousands of traffic cameras installed and our extensive conversations reveal that they are very interested in analyzing video from these cameras for traffic safety and planning. Detecting “close-calls” between cars, bikers, and pedestrians helps preemptively deploy safety measures. To help install crosswalks, cities would like to detect areas where jaywalkers cross streets. To improve congestion, they would like to detect double-parked cars but exclude first-responders. Counting volumes of cars, pedestrians or bikes feeds into the traffic light controller to appropriately control the durations.

A significant part of the video surveillance program going forward will be video analytics, computer algorithms written to automatically alert officers to possible terror attacks or criminal activities. Both real-time monitoring and postevent forensics can greatly benefit from the development of autonomous video analytics tools. Today, this development is proceeding in three directions that are simultaneously pursued in both signal processing and computer vision communities

Video analytics is not limited to only traffic surveillance but potentially central to a wide range of future and existing applications ranging from surveillance and self-driving cars, to personal digital assistants and even drone cameras.

No users to show at the moment.

Bigger i card photo11
  • Projects 0
  • Followers 5

Nitin Mane

Aurangabad, Maharashtra, India