Traffic AI

Traffic AI

Traffic congestion is one of the major problems in big metro cities due to the increase in the number of vehicles on the roads. In today’s day & age, it is widely acknowledged that Indian cities are increasingly becoming congested and difficult to live. With the rapidly increasing population, the traffic on the roads is also increasing proportionately. The challenge can be addressed through various methods using modern technology. 

Vehant’s traffic management and enforcement system is artificial intelligence and machine learning-powered traffic enforcement and management solution. It offers several violation detection capabilities such as red-light jump, over-speeding, no helmet riding, triple riding, wrong lane movement, stop line violations, speaking over the phone while driving, and various other traffic violations in real-time. The system is connected to the traffic control rooms and is monitored round-the-clock. The ANPR reads the number plate of the offending vehicle and stores it in the database which is then used to generate e-challan. The system also facilitates e-challan with the photo or video evidence which is then sent to the violator’s mobile phone via SMS. Furthermore, accident and other anomaly detection on the road along with vehicle attribute recognition and traffic flow analysis are few more applications which we intend to incorporate in the current Vehant’s traffic management system in upcoming years.

Traffic flow analysis

Traffic flow analysis is a tool to analyze various traffic-related issues that include traffic congestion, traffic accident, road construction, and noise pollution to name a few. The growing transportation demand with an inadequate supporting infrastructure makes the traffic condition chaotic in countries like India and therefore it becomes necessary to analyze the traffic flow using a traffic monitoring system. The research focuses on designing and developing such systems using the latest computer vision and AI technologies.

Vehicle Attribute Recognition

The traffic monitoring system is a necessary tool to capture real-time traffic statistics for better planning of transport infrastructure. Simple vehicle counting is sometimes insufficient, and it is essential to record additional vehicle features such as make, color, type, etc. For detecting these attributes, separate classification models exist in the literature to use. Still, the challenge is in training the generalized model for detecting all the relevant attributes, including the finer like state of the headlight/backlight, the vehicle is overloaded or not, the vehicle is damaged or not. Our research focuses on training a generalized attribute recognition model for surveillance scenarios using the latest AI technologies. 

Traffic Violation Detection and Analysis

 There is a need for automatic violation detection systems to avoid accidents with the increasing traffic on the road. This violation includes seat belt, triple rider, helmet, mobile calling, red light, over-speed, etc. Although Vehant TrafficMon solution is already available to use, but there is a scope of improvement in dealing with certain challenges like illumination, dark light, occlusion, etc. The research team at Vehant is working on additional features to improve traffic monitoring solution further.

Autocalibration of Traffic Cameras

Calibrated cameras allow measuring the real-world distances from the captured image and video data, thereby addressing a wide range of problems in traffic scenarios like speed measurement, vehicle collision, etc. Moreover, manual calibration of the individual cameras is a tedious task when deployed at a large scale.  Our research focuses on the automatic calibration of traffic cameras using the vehicles’ information moving on the roads. The challenges include identifying the true category of vehicles and detection of key points from the vehicles as traffic cameras typically have low resolution.