Geekly Articles each Day
📜 ⬆️ ⬇️

Hazardous production: we are watching you,%% username (video analytics)


One friend - without a helmet, the second - without gloves.

There are many not very good cameras in the productions, in the squares of which not the most attentive grandmothers look. More precisely, they are just crazy about monotony there and do not always see incidents. Then they slowly call, and if it was a call into the danger zone, then sometimes it would make no sense to call the shop, you can immediately relatives of the worker.

Progress has reached the point that the robot can see everything and give it to everyone who violates. For example, recalling by SMS, light discharge of current to the siren, vibration, nasty squeaking, flash of bright light or just tell the supervisor.
')
Specifically:


And so on. The simplest example is the color differentiation of bricklayers and concrete castors by the color of the helmet. To help the robot. In the end, to live in a society with a lack of color differentiation, it means not to have a goal.

How to steal at a construction site


One type of widespread theft is when the contractor promised to bring 100 workers to the site, and in fact led 40–45. And the house is built and built. Anyway, no one can exactly count them. As in the well-known joke: if a bear settles at a construction site and eats people, no one will notice. So the general contractor has no opportunity to control the brigade. More precisely, even if you use the access control system, it will still be deceived, as in this post about the cat-terminator .

Usually there are no ACSs at construction sites or they are only at the entrance.

We went to change experience to highly developed civilizations and saw that each profession (or rather, role) has its own color of helmet. Here the bricks are laid by the handlers - their helmets are blue, the concrete is poured by the castors - they are green, all sorts of clever people are walking around - they are yellow, therefore before them they have to do two “ku” times. And so on.

And you need all this to very easily detect each role. The facility is worth several dozen fairly cheap cameras that give something like 320x200 in color. Workers are considered to be helmets in real time, and a specific construction site is tied to each chamber. As a result, at the end of the day, all this is merged into analytics of the charts by zones: who, in what quantity and in which sector he worked.

In general, we have adopted the experience. Only while we were watching it, neural networks stepped far forward, and many new detectors appeared. A few years ago they were quite capricious and unstable, and now they allow us to very accurately catch the most interesting situations. Not least because of the processing speed, the detectors are often mistaken on individual frames, and on the video stream with minor changes in the angle we get an excellent practical result.

And if I put on a second helmet on my belt?


First, we learned that a worker can get two helmets and fasten one of them on the ass. We had two detectors at once: the search for the skeleton and the determination of the color spot to match the apex of this skeleton and the search for synchronously moving objects. In the second case, it turned out to be easier to detect: for example, a person with a helmet on his ass almost never looks around with that helmet. Because for this you need to rotate the head. And this movement is very easily detected. More precisely, we do not know what is actually detected there (it’s a neural network), but it learned very quickly and catches violators, one might say, by walking.


We are building a human model.

Then we just build a heat map in real time and reports at the end of the day.

Accordingly, according to the same principle - by training a neural network - it is easily detected:


In total, 29 detectors have now been run in. The only point is that since we work in hazardous industries like chemicals or mining, there are requirements for the types of gloves. For example, long and short. In this case, it is necessary that they be of a different color: the length under the sleeve on the video camera is very difficult to determine.

And here there were often draws in rats. We do not have a separate rat detector, but there is a detector of objects that interfere with the operation of the machine:



What else is detected?


We ran detectors at chemical plants, in the mining industry, in the nuclear industry and at construction sites. It turned out that with minor efforts it is possible to close a few more requirements that were previously solved by the same grandmothers, who were frantically trying to see something in the picture through poor resolution and with a poor frame rate. Specifically:





Behavior control was implemented very easily - whether the employee sleeps specifically or not. While we were testing all this, the rules evolved from “In this zone there should be a man in a green helmet” to “In this area a man in a green helmet should move.” So far, there was only one wise guy who skipped the chip and turned on the fan, but that too turned out to be easy to fix.

Chemists, it was very important to fix all sorts of jets of steam, smoke. In the oil industry - the integrity of the pipes. Fire is generally a standard detector. And there is a check of closed hatches.



Lost things are detected in the same way. We ran it on one of the stations a couple of years ago, there it almost makes no sense because of the large number of events. But in the production, especially in the chemical, it is very convenient to keep track of things in the clean zone.

It is interesting that right from the video analytics we can read the readings of the instruments in the area of ​​the camera. This is true for the same chemists, whose production facilities have a high hazard class. Any change, like replacing a sensor, is a renegotiation of the project. It is long, expensive and painful. More precisely, LONG, EXPENSIVE and SICK. Therefore, their Internet of Things will come late. Now they want to monitor the counters and read the data, respond promptly to them and reduce losses due to unexpectedly and unnoticedly failed equipment. Based on the actual data of the counters, you can build a digital twin of an enterprise, implement a predictive MRO, but this is a completely different story ... There is already control: we are writing now proactive analytics using a combination of data. And separately - battery replacement prediction module.

Another incredible thing - it turned out that in granaries and in storage of materials such as rubble, you can remove a pile from 3-4 angles and determine its edges. And defining the edges - to give a volume of grain or material with an error of up to 1%.

The last detector, which we wrote, is the control of driver fatigue, such as pecking nose, yawning and blinking frequency. This is for HD cameras where you can see the eyes. Most likely, will be placed in the control room. But the main need is for BelAZ, KamAZ for careers. There, it happens, the cars are falling, so now they are making up something on mining to make the driver control. The robot is better than grandmother.

About cars. For example, the topic of fatigue control is actively used by automakers not only BelAZ, KamAZ and other MAZ. Already in the usual ordinary cars, manufacturers are building a warning system for driver fatigue, but so far they have fairly simple solutions that analyze only the position of the car relative to the markings and the nature of the movement of the steering wheel. We went further and detect human behavior, which is much more complicated.

Another case of surveillance of the driver is the detection of improper behavior when using car sharing machines. They can not talk on the phone without hands free, eat, drink, smoke and a lot of other things.



Oh, and last. For several years now, we have been able to do object tracking between cameras - when, for example, we stole something, we need to check which way and how. If there are 100 cameras on the object, then you’ll be raising the material. And here the system will form an action thriller about Ocean and his friends automatically.

What is the difference from the two-year-old system? Now this is not just a recognition like “bald in an orange jacket from one camera came out and almost immediately went into another”, but a mathematical model of the room is built, and according to it are hypotheses about the movement of an object. That is, it all began to work on areas with ceilings and places with blind spots, and sometimes extensive ones. Yes, and detectors are now much better, because there are libraries that determine age by face. On HD-cameras, you can set the orientation like "a man of 30 years with a woman of 35 years."

So, maybe in 5–7 years we will end up with production and go to your home. For security. It is in your best interest, citizen!

Links


Source: https://habr.com/ru/post/449560/

More articles:


All Articles