Intozi's Video Analytics Solution helped LnT overcome safety and security challenges during tunnel construction

In India, construction under tunnels for making passages is one of the most challenging tasks if it comes to safety and security measures. Any unsafe act can indulge the whole program into a big tragedy at such delicate places especially when it comes to excavation of rocks. Mass excavation takes place by several methodologies where a large number of manpower is involved to make this happen successfully. Now, one of the key challenges is to track the movement of vehicles, workers safety handles-Jackets & Helmets, and their attendance.

Solution: Intozi’s AI-Based video analytics solution is implemented at L&T tunnels-(ADIT-2) where our solution is meant for Vehicle detection, Number Plate detection, OCR reading, Face detection, Attendance via Face matching, Without Helmet detection? and without safety jacket etc.

How the System Worked: An upgraded hardware is used keeping delegacy of sites/scenarios in notice. Clear images of workers are used to register for FRS purposes. Custom tags used for heavy machines/ loaders/ cranes etc. This complete solution helped L&T management in a highly appropriate manner as they were able to detect the vehicles passing through, attendance of the manpower was managed in a proper manner as well as No helmet detection keeping the safety of workers as priority.

Challenges under Tunnels

Severe weather conditions, Dust due to continuous excavation which leads to poor visibility. Lack of power supply or unavailability of power supply as well as poor network signals.

1. ? ? ? Positioning of Artificial Lights

2. ? ? ? Dust

3. ? ? ? Poor visibility of vehicle number plates.

4. ? ? ? Unfavorable weather conditions.

5. ? ? ? Width of tunnels etc.

6.? ? Face detection of workers for attendance

7.? ? To manage the clear view of workers in camera for safety measures detection like Helmet and safety jackets.

Positioning of Artificial lights: Challenge- lights mounted on the both sides of tunnels were not fixed properly and their position gets changed from time to time. These lights were creating problems for cameras to view the vehicles & workmen properly as light source facing cameras directly diminish the camera visibility capacity.

Dust: Challenge- video analytic solution finds it tough to read the vehicle number clearly when the image is taken in a dusty environment as extreme dust results in poor visibility. Vehicles entering or leaving the camera view site at high speed are likely to erupt more dust as compared to vehicles moving at slower speed. High speed will result in more and more dust coming up from the surface.

Poor Number Plate Visibility: Challenge: a majority of vehicles were either without number plates or number plates were covered by dust and mud. It was impossible to read the number plate characters properly as characters were not visible clearly. Machines like Cranes, man lifter machines etc. were without number plates. System was able to count the vehicle instead of reading the vehicle registration number.

Width of Tunnel: Challenge: there are sites (ADIT-2 particularly) where tunnels are very wider as compared to other sites. Higher degree of angle between Camera position and vehicle’s number plate resulted

?How Intozi tackle these challenges with AI-Based Video Analytics solution

Intozi analyzed the issues at the tunnels where the solution was to be implemented.

?? Firstly, our team overcame the camera vision impact issues by fixing the lights' positions.

?? Secondly, to improve the visibility on the site and reduce the dust all over; regular water spray so that dust can be settled. The daily practice of water spray adapted by man force at site.

?? In addition to this, to control the speed of the vehicles, implementation of speed breaker as well as written guidelines to slow down the vehicle to maintain the hardware integrity and enhance the accuracy of the solution.

?? Custom tags (asset codes) are pasted on all the registered vehicles and the system was trained to read the asset codes along with vehicle registration numbers. System was able to track the vehicles without number plates by reading custom tags. Chances of custom tags getting distorted are less as compared with number plates as these tags are pasted on the inner side of the windshield so that external factors like mud and dust should not distort the font written on the tag.

?? To read the vehicle number under a wiser tunnel, a path is defined for entry-exit so that vehicle movement is captured by camera clearly and the number is detected in the required manner.

?? An additional measure is to use PTZ cameras for better accuracy.