Get best of the week

Use of a smoke detector in vehicles



Earlier we talked about smoking detection through object video analytics. Now let’s try to consider the practical aspects of the application of these solutions and specific implementation sectors, as well as their advantages for the business.

In our opinion, the most interesting area of ​​application is transport, in particular - car sharing, where penalties in the form of fines for smoking in the salons of rented cars are already provided. The amount of the fine varies from 5 to 15 thousand rubles, depending on the company. Returning to the comparison of object video analytics and sensors, the sensors do not pick up vapes and other devices for smoking mixtures, and are also practically not sensitive when the car's windows are open. But this does not cancel the fact of violation and, accordingly, the legal punishment in the form of a fine in accordance with the contract.

In addition, in transport, several neural networks can be cascaded (sequentially), such as smoking detection and detection of the fact / time of use of a mobile phone. It is clear that further such systems should be scaled, for example, with the integration of telematics and connecting to the vehicle’s CAN bus to track the use of telephones only when the vehicle is moving, but these are already integration details.

A good example of what we specifically detect and what we get as a result:





Demonstration on bots in Telegram (input - picture from the smartphone’s camera or gallery, exit - probability):


Hardware component


If in the first article we talked about Intel NUC and servers based on them, as calculators for inference, now we are talking about operating the solution in vehicles, that is, the influence of weather conditions (heat, cold, dew point, etc.) appears . AAEON, VPC- 3350S, turned out to be a good solution :

AAEON, VPC-3350S

Specifically, our version is with Intel Atom x5 E3940 processor. Interface - on MyriadX on the expansion board. FPS in inference:



Decoder tests:



What is a good piece of iron and why did our choice fall on it?


We liked:

  • The presence of a built-in LTE-module.
  • The availability of expandable VPU accelerator Intel MyriadX.
  • Integrated Intel HD Graphics 500, on which you can use hardware decoders and encoders to process video streams.
  • LAN- .
  • (-20+70).

?


  1. Ethernet, POE ( : , ).
  2. , AAEON NVR 3350.
  3. .
  4. .
  5. ( ). . , , 50%, ( ).
  6. Based on the number of recurring events, the action / violation time is recorded.
  7. If the action time exceeds the specified constant (10 seconds), then the fact of the event is recorded in the database. The event includes the following information:
    • date Time
    • violation photo
    • event duration in sec.
    • vehicle identifier (static GUID)
    • camera number (0, 1)
    • event type
  8. Data on events upon the availability of 3G / LTE is transmitted to a central data processing server with integration with the existing information sharing system for billing operations.

Instead of a resume


In the article, we tried to share our experience in implementing and integrating AI solutions using the example of transport infrastructure. Most importantly, most automation facilities are already equipped with cameras, and you can process existing flows without any significant modernization.

More articles:


All Articles