Trinetra

Urbanization and growing population have led to an exponential rise in the number of vehicles on roads, leading to higher traffic densities. This poses significant challenges for monitoring the road infrastructure in real-time for detection of traffic violations. This problem is compounded by poor driving behavior such as ad-hoc stopping, poor parking, over speeding and other traffic violations. Detection and monitoring of these conditions would significantly reduce traffic congestion and accidents.

We propose crowdsourced detection of violations using their hand-held mobile phones. Validated violations are stored on the Blockchain which is maintained by relevant organizations (government and corporate), allowing for auditability of data while ensuring security and privacy. 

Through concerted and collaborative efforts across all road users we believe we can improve the commute experience and safety by reducing violations. We envisage an ecosystem through which the citizens are rewarded for their active engagement in bringing a societal change.

Almost every road user in India is impacted by traffic violations. Illegal parking/stopping causes obstruction to the normal flow of traffic resulting in severe jams. Driving on footpaths causes tremendous inconvenience and fatal injuries to pedestrians. At least 98 two-wheeler riders without helmets died daily in 2017.  The report on Road Accidents published by the Government of India states that over 12,257 of 1,50,785 road accidental deaths are due to lane indiscipline and reckless driving. 

Traffic is usually monitored and controlled by policemen either in-person or from a centralized location. In-person monitoring are limited to the line of sight of the policemen and their ability to capture all the details of violating vehicle. 

Manual inspection of CCTV video are done by authorized personnel who create traffic tickets if any violation is observed. This mode of operation is time consuming. Additionally, limited CCTV coverage in cities poses challenges for micro analysis. Capital investment and operational costs are high to install and maintain cameras along every road segment. Accurate license plate recognition requires high quality images from specialized cameras and often the available network bandwidth is limited. 

We propose crowdsourced detection of traffic violations using hand-held mobile phones.

Being part of the choked streets of metropolitan Bangalore, we contribute to traffic jams everyday. Currently there are 8 Tier-1 cities in India, which include Bangalore, Chennai, Delhi, Mumbai, Hyderabad, Kolkata and Pune. In the city of Bengaluru, the population is 12.3 million with ~7 million registered vehicles. With an annual growth rate of 10%, the number of vehicles is estimated to be at 11 million by 2022! 

A study (https://www.who.int/news-room/fact-sheets/detail/road-traffic-injuries) by WHO states that if traffic laws are not enforced or are perceived as not being enforced, it is likely that the road users will not comply with and therefore will have very little chance of influencing behaviour. Effective enforcement includes establishing, regularly updating, and enforcing laws at the national, municipal, and local levels.

The department of transport has a shortage of manpower to enforce violations at the micro-level. We propose to solve this problem with technology at scale by involving citizens in the enforcement of road safety regulations. 

We envisage to pilot the solution within a small community in India and then scale to larger cities jointly with the transport department and other city administration authorities.

Trinetra empowers the community to become “citizen police” to not only ensure that traffic rules are followed, but also to beat the manpower shortage in traffic department. 

Mobile phones are an integral part of human life with 62.9% of the population worldwide already owning one. We leverage the penetration of mobile phones in the market by developing a mobile app to detect violations. The app, powered by Artificial Intelligence (AI) with blockchain authentication, can be used by anyone to report violations and get rewarded for their active engagement.  

Trinetra has three major components: 

1. Violation Detection using AI 

Artificial Intelligence (AI) based object detection, tracking and scene understanding are used in the application. The reported violations are processed in real-time on the mobile-phone itself. The app will prompt the user to record the scene until a valid violation record is captured. Once a valid violation is obtained, it is logged on to the blockchain. 

From detection to reporting, the entire pipeline is executed automatically with high confidence limiting manual intervention. This helps in saving time and manpower. 

2. Blockchain framework for violation proposals and enforcement

To provide authenticity to the violation report and to ensure that neither the public nor the law enforcement personnel can abuse the system, we use a blockchain backend to authenticate users, log violation complaints, log actions taken by the enforcement authorities in response to complaints, and penalize the violators in a secure, immutable and transparent way. The law-enforcement is notified of a new violation, who can access the blockchain for details and enforce fines on the violator. The fines collected are the main source of revenue, which is distributed among the law-enforcement agencies, violation reporter and blockchain miners.

3. Gamification

Game Play is and always has been an essential trait of evolutionary success and one of the strongest drivers of human behavior(1). Uber uses ludic loop and badges to encourage drivers to stay on the road and earn more money while ensuring quality service to its customers. Cure.fit app uses points and badges to motivate physical activity and healthy eating habits. We use the concept of gamification in Trinetra to incentivise users to better engage them. 

Alongside the incentive, each violation reporter is also assigned with a reputation score which is increased based on their transactions and motivation to continue using the app.  

References

1. https://yukaichou.com/gamification-examples/octalysis-complete-gamification-framework

• A highly scalable platform enabling citizens from any geographical region to engage, by downloading and using the app on their mobile phone. 

• The reported violation is validated by artificial Intelligence (AI) algorithms (object detection, tracking and scene understanding) in real-time on the mobile-phone itself before being logged on the Blockchain platform. The intelligence within the app will prompt the user to record the scene until a valid violation is captured.

• The Blockchain backend authenticates users, logs violation complaints along with associated actions taken by the enforcement authorities, and penalizes the violators in a secure, immutable and transparent way. This ensures auditability of transactions preventing any misuse of the ecosystem.

• Gamification improves user engagement - through concerted and collaborative efforts we envisage to develop an ecosystem through which the citizens are rewarded for their active engagement in bringing a societal change - improved safety and commuter experience. 

• Multiple users capturing evidence of the same violation from different viewpoints can result in several violation proposals. In this scenario, each reporter is rewarded for the effort based on the time of transaction and reputation score preventing user churn.

• Enforcement action is entirely automated, thereby reducing the number of defaulters of fines and increasing the revenue. This solution also addresses the significant challenges associated with limited manpower within the police & transport departments.

To the best of our knowledge there is no platform that provides the above mentioned innovative features for engaging citizens in a collaborative sphere with a common goal to  improve the safety on roads.

Fig 1: Architecture of our mobile application for violation detection

Fig 2: Blockchain framework 

Object detection: AI based object detection algorithms are used to detect vehicles, license plate, pedestrians, riders and other entities such as traffic signs in the captured scene. We use recent advances in deep learning for data-driven object detection with TensorFlowLite. 

Violation detection: Once all necessary objects pertaining to a violation are detected along with timestamp and GPS data, the confidence of violation detection evidence is estimated.

Blockchain framework for violation proposals and enforcement: We use the Ethereum blockchain framework, to serve as a backbone of the system. Consensus is achieved through traditional proof of work mining, and miners are rewarded either directly in Ether, or with a native system token (TrafficTokens) which can then be exchanged for Ether or any other flat currency. The mobile app, when downloaded, creates a blockchain wallet (private-public key pair) for the user by default. This enables users to post transactions signed with their private key for the blockchain network to verify. 

Gamification: The incentivization follows a gamification theory. The main revenue to the system is from the traffic fines paid. A major portion of the collected revenue is given to the law enforcement and the remaining is fed back into the system as incentives, divided among the violation reporters and miners. In addition to the incentive money, the participants are competing for a reputation score which will increase based on the quality of the violation report generated by each of them.

Currently, the solution is in the prototype stage and we are not actively serving any communities yet. We are working towards piloting the solution within a small community in the next 3 months and then scale to larger cities in India jointly with the transport department and other city administration authorities. 

To emphasize the magnitude of scale associated with this solution, consider the following statistic - currently there are 8 Tier-1 cities in India, which include Bangalore, Chennai, Delhi, Mumbai, Hyderabad, Kolkata and Pune. In the city of Bengaluru, the population is 12.3 million with ~7 million vehicles and 8.3 million traffic violation cases reported in 2018! With an annual growth rate of 10%, the number of vehicles is estimated to be at 11 million by 2022! 

In the next 1 year, we envisage to scale our solution to many of the above mentioned Tier-1 cities where traffic violation is a serious concern.