Guiseppe, first question: Can the analytic be wrong every so often? I am suspecting the analytic needs to be fairly perfect, given the consequences of making a mistake (missing a person or an animal, etc.). Yes/no?
Secondly, can the analytic be disabled by an external input (i.e. signal is given when a plane is landing) or does the analytic need to automatically distinguish planes from other objects?
Norris, Inc., S. Portland, ME
Multiple analytics cameras running "left object". I was thinking laser optics or microwave skimming the tarmac, but not sure how you would make this work. Sensors would detect the plane and would shut off the alerts for a set amount of time of until it reaches a "landed" sensor. Probably call Sightlogix.
^^I had the same sort of ideas as Jeff, but as John notes, reliability is a concern. I don't know that I'd leave the actual raising of an alarm to the system; rather, I'd probably have it set to relatively high sensitivity (higher chance of false positive), but have that raise an alert that a human operator can check on to confirm. This is an instance where PTZs could be effective, tied into the analytics system so that when something is detected, a PTZ will zoom on the area, ready for a human to look and confirm the existence (or not) of a problem object.
At least a runway lends itself well to some sort of low-level surface-scanning idea, being super flat by necessity.
If your analytics are looking for "left objects" a plane shouldn't be a problem for it, as that will just pass through the scene.
There is a product called a ground surveillance radar which is typically used along with other radars in airports. That in conjunction with a good thermal PTZ camera (for all weatheruse - including fog) should be able to do what you are interested in. A google search will yield the names of several manufacturers/vendors.
The last thing I'd suggest is a "left object" video analytics approach. In an outdoor area like an airport that would be a nightmare to maintain. Your best approach is most likely a rader-style approach, whether it be actual radar or a similar technology.
There is a group in TN, The National Safe Skies Alliance, that tests various technologies specifically for interior and exterior airport security purposes. I can give you some contact names there if you like, they might be able to help tell you if anything has been tested, or submitted for testing, for this purpose.
The level of precision of the analytics should be quite high, considering the possible consequences and they should be able to distinguish a plane with something else.
I've spoken with some manufacture of laser sensors but they did not feel very comfortable in fog condition so it may not be the right choice.
I was also thinking of having video supporting the system, as you already mentioned.
After the detection, the system should generate an alarm and the object should be shown in video so that a human can confirm the situation or recognize a false alarm. This will also help in tuning the system.
My first idea was to use fixed camera with a wide angle lens.
Then you have the VCA algorithm to detect abandoned (still) objects and it is already filtering objects in motion.
But what about rabbits and birds? they are also in movement...
I have still some research to do on this.
I agree that a standard left object apprach would fail.
We can offer a learning algorithm that can learn the environment, people, animals Planes etc. This works well with moving traffic, stationary traffic, pedestrians etc. and could be adapted.
We can also decode high resolution images like the Avigilon 16mp to get more resolution for small objects like birds. Although I would feel we would need around 5-8 frames.
The whole system does need to be thought about.
Vision – is fog/smog going to affect the view?
Night time – what are the light leves? b/W with IR is fine for analysing.
What other traffic/animals/people are likely to be there.
The simple thing would be – record some footage with Planes and other traffic send it to us – we will run it through our engine. We will overlay the metadata information so you can see what the system is detecting.
If you like the results we can then develop the software to your needs.
We are happy to discuss the concept.
We need more information from you to help find an appropriate solution. Could you give us a better description of your problem and the environment in which you are working? If fog, dust haze or flying leaves are a part of your environment, Video Analytics may prove unequal to the task.
Your use of the term airport indicates that you are probably referring to a place which does have a control tower with a clear view of at least a part of the runway, otherwise it would have been called an airstrip.
Generally, left objects should not be a problem because if it has only a few flights per day, a jeep or pickup should have driven down the runway sometime before the ETA of the aircraft to check for such objects. A larger airport would probably already be equipped with a ground surveillance radar.
The real threats are from one or more of the folloowing, moments before a landing, (i) a stray jeep, pickup or worker coming on to the runway after ignoring the SOPs of the airport (ii) if the perimeter is not adequately protected, larger animals like deer, sheep, goats and others straying onto the runway, (iii) larger birds like eagles and vultures, if by chance there are any dead animals or other food in the neighbourhood; and all these types of obstructions could severely damage or wreck the aircraft as well as the intruder.
In all such cases, till the runway is clear, the aircraft would have to flyover on the instructions of the ATC.
Since the solution does not lie merely in the detection of an intrusion on the runway, but also in responding to such intrusion, a few small birds or rabbits are somethings which would have to be ignored (as far as the solution to be provided by you is concerned, since often the response time available after the intrusion is extremely short) and may get runover (most unfortunately). Every year, worldwide, several jet engines get ruined by sucking in birds on runways. Most controllers at airport ATC towers live with these problems.
It certainly depends on the specifics of the detection and the CONOPS of the scenario. VCA's can perform analytics that "ignore" aircraft of various sizes but detect when non-aircraft are on the taxiway. See this video. This is done through real size calculations and classification algorithms. However, this particular application is looking for non-aircraft objects that are in the size of humans and larger. If you want to detect smaller objects, whether stationary or moving, it comes down to coverage of the runway and available resolution to detect those smaller objects. In my experience in the aerospace world, it also comes down to the type of object. Airports and pilots are less concerned with things like coyotes or dogs. They stand little chance with an aircraft moving at high rates of speed. However, they do appreciate "bird detection", especially the detection of "flocks of birds," as these can cause some serious engine dmanage.