Automatic illegal recognition

A project under the german "safety research" tried to join "intention diagnostics" to prevent strikes on U-bahnthofen. This is just one example of the desired digital pattern recognition as security technology

It’s spat at night. The subway station in a german coen town is almost empty, only a passenger is still waiting for the last train. There is a crackling group of drunken teenagers. It comes to a change of words, the situation threatens to escalate. As one of the teenagers stabbed the man in front of the chest, suddenly turns a voice from the speaker: "attention, they are observed! Horen you immediately!"

So about the application scenario of adis looks short, short for "automated detection of intervention tiger situations by classification of visual pattern". This research association develops a kind of alarm system for public transport: the system of computer software and digital cameras should automatically recognize whether a constraint is notified in a certain situation and then the wax personnel point out.

Automatic illegal recognition

From the fact leaf adis of the BMBF

With the help of pattern recognition of dangerous situations automatically filter out from the tide of video recordings, goods for the safety force on traffic junctions such as airport or trainhofen indeed immensely practical. In the past decade, staff was dismantled on-site, for example on the platforms. Today, employees are remotely with the help of video cameras. These cameras provide you with a plethora of image recordings. The fact that the overwaxes are rigid to the monitors in the control centers, but is not only boring for them, but also a little successful. Already after a few minutes the attention decreases clearly. Highest for half an hour on the stucco, the employees can watch until they have to be delivered. If it really comes to a criminal offense, the recordings are usually only evaluated fulfilled. To prevent violence, the existing video monitoring hardly is good. More: it is becoming increasingly clear that the praventive and repressive effect of the open comecright monitoring, because your deterrent effect was based on supercurrent expectations.

To automate the recognition of dangerous situations, there is an attractive way out – but technically and conceptually ambitiously ambitious. Among other things, the systems have to compensate for perspective distortions and changes from the zoom from the digital image data to detect the specified features. So the software "intervention requirements" can recognize typical body movements of fashion and sacrifices for strikes and modeled mathematically.

The fraunhofer institute for production facilities and design technology develops a dedicated to this purpose "motion description language". ADIS captures all forms of corporary injury, except crimes that are accompanied by body injury, such as robbery. In the relevant publications, however, medical emergency cases are also mentioned, for example if a passenger collapses because of a stroke of stroke or, little concrete, "other intervention-approved scenarios".

With the help of pattern recognition in the future?

But the project is even more ambitious: the system should recognize aggressive intentions, before show this overhead in violence. The research idea: as soon as the "intelligent cameras" discover patterns that point to violent motifs will be notified. For this purpose, the corresponding shots on the screens in the control centers could be highlighted acoustically and visually. By "intention diagnostics" based on facial expressions, gestures and corpers of the taders and the victims, forecasts should be generated automatically.

In order to meet statements about the future actions of individuals, the psychologists from adis to approaches access to influence human behavior on quantifiable factors. These are among others around the "theory of the planned behavior" or the "theory of the uplieved action", which were developed by the american psychologists icek ajzen and martin fishbein. Their motivation models are previously used mainly by the advertising industry in order to predict the behavior of consumers. It’s about buying decisions. Whether physical activities really accurately predict enough for body movements, is open, is open.

Between one and two minutes, the ADIS planners want through the "intention diagnostics" win – time in which the employees could react to the situation. If that succeeds? Capital conflicts in public space are finally a coarse part of spontaneous affect actions, and most clashes escalate quickly. Involved scientists are exposed to say that the success of the research project is by no means established.

Like any alarm system, the ADIS system must also be sufficiently specific and sensitive at the same time: it may "right situations" neither miss nor produce too many errors; otherwise, the system was more likely to hinder the surveillance than to use them. The "automated detection intervention tailor situations" goods not the first monitoring technology that fails in practice. The introduction of the so-called corperscanners did not ultimately fail because of privacy concerns, but because () the guard people could not start the airport with the device.

The previous application scenario of ADIS provides that the platform does not fully overwake, but only a marked area. The passenger hour then free to stay inside or outside this supercover zone. How practicable or meaningful that is, fruesters will turn out in marz next year when a first plant will be presented. The end user is called the munchner transport company. Field trials under realistic conditions, however, are not planned, only prerequisites where project employees mimen the passers-by in public space.

Auber the fraunhofer institute for production plants and construction technology psychologists by the ruprecht-karls-university in heidelberg and the munchner company indanet information and data networks developed, which develops control center technology, to the ADIS research association. Indanet is also with orgamir, another security researcher for the defense of terrorism on tunnels in coarse-cast rail transport.

Basic research for the export of monitoring technology

The federal government calls for ADIS with about 1.2 million euros. The research project is based on the program line "pattern recognition" the so-called security research. When it comes to this "research for civil security" go, then the federal research department likes to talk about new equipment for firefighters and rescue anitators, or how to prevent technical catastrophes in coarse events such as the duisburg love parade. The fact that the research program is also developed with the research program also monitoring technology (ruckwarts and waiting monitoring) that tends deep into the privacy will rarely mention.

The "research for civil security" serves primarily in the economic demand. Not randomly the program is part of the "high-tech strategy", with which the federal government wants to secure the german position on the world market. By a targeted "innovation transfer" target "the international pioneering of german providers of civil security products and technologies long-term expanded" become, it is called from the federal research department. However, the german industry is strong competitors, especially from the USA, great britain and israel. In the new framework program, the economic significance of the "safety research" the following remains described:

The market for security products and services with a global volume of approx. 100 billion euros in 2008 growth for the organization for economic cooperation and development (OECD) for five to seven percent. At the european economic area?. 30 percent of the total volume. Civil security research will help to further develop the market for security technologies and services, so that innovative german companies as well as researching facilities can benefit from this booming market.

In 2015, according to federal ministry of economics, 31 billion euros for security technology will be implemented in germany alone. Pattern recognition applies as key technology for the monitoring technology of the future. Auber adis try four more security research projects to use this technique for traffic hubs. During ADIS busy with U-bahnhofen, apple is dedicated to the airport terminal and ASEV to the airport run. All of these projects try to evaluate image data automatically by detriment, control assistance and danger potentials. Such systems are also "intelligent video monitoring" called (smart CCTV). This is done on the one hand that "tracking", so the automatic identification and tracking of a suspicious person or a vehicle over several cameras, on the other hand, the behavioral analysis of individuals or groups.

In german security research there are both: apple, for example – "analysis of passenger movements at the airport using temporally ruckwarts- and prerequisite video data streams" – develops a system with which the safety faith can find in an airport terminal, where a suspect has previously stopped and what it probably will take as a next. When a deviation from the gear of movement patterns, the system generates a hint to "at home the risk potential". As part of ASEV ("automatic situation assessment for event-driven video monitoring") is a facility that is to automatically discover objects or persons in the airport for airport, which have nothing to look for there.

And privacy?

At the projects of security research, not only technicians and computer scientists, but also many lawyers, psychologists, mind and social scientists are involved. However, these thematize "security researcher" not the question under which threats the population actually suffers and what to do about it. Your stake is little on the orientation of the research program on application and marketing based on fixed scenarios of civil protection, terrorist defense and criminal title.

On the other hand, those responsible have quite right if they emphasize that social scientists are more than just "accompanying research" have been operated. The non-technicians fulfill the necessary tasks in order to make the technique used for use: they explore what the users want to have in the security organizations which contain the contents in the population and develop the basics for the models of danger, which ultimately technically with the sensor data be matched.

So far, quite a few people are well known, what "intelligent video monitoring" is actually and what it serves. However, if you experience control and monitoring applications that work with digital pattern recognition, these often mistrust on mistrust. Especially the fact that instead of people are taking a part of the monitoring, causing more or less diffused discomfort.

The project staff of adis, ASEV etc. Stress therefore that they only "assistant systems" to develop. People still became the last decision to meet whether a suspiciter is further observed or even arrested. Nevertheless, the machines will play a crucial role by discovering certain detections and standard deviations – and others. Adis, for example, works with a catalog of facial expression and body movements with one "sufficient" probability of artistic violence. If the security personnel uses the technical systems actually and intensively, they draw their attention to people who otherwise did not target the surveillance.

Is something compatible with the existing (privacy)? The legal answer is, like ubrow: depends on it. In germany, the fundamental right to informational self-determination is considered a prerequisite for a democratic society. As a defense law, it turns against the stitching prere that arises when the burgers can not understand whether and how they are just monitored and who has what information about them has. The data protection regulation is therefore concerned with the question of which behavioral chants are acceptable and what information must receive the monitoring to the "informational self-determination" genuge to do.

The safety research association muvit ("pattern recognition and video tracking") analyze the psychological, sociological and legal consequences of part-automated video monitoring. This is also attended by the search for fundamental participation opportunities for the "intelligent cameras". Psychologists surprise the so-called animer diaphragm empirically by testing whether the behavior of subjects is concerned if they believe they are watching of pattern recognizing cameras. Allegedly, many types of experiments in these laboratory studies are indeed their behavior, have increased self-perception and avoid if possible, monitoring.

However, that could be because hardly anyone has been practical experience with the next generation of video monitoring and the technology is too much. Even the classic video monitoring initially looked much inhattering than today.

According to the previous concepts, no personal data is stored for the pattern recognizing safety technology. Technically, however, it is possible to connect a behavioral detecting with an identifying, for example a biometric system. Because the costs will fall through the progressive automation, a new control density will be possible. It is neither a lot of imagination nor boser’s will to count out what the "intelligent video monitoring" for example, in a chinese factory or public space in saudi arabia, it soon be used – and that people will soon be heavier who lie or sit in public space, rather than protecting themselves from shop windows to shop windows.

This repressive potential of pattern recognition leads to uncomfortable socio-political ies, but to answer the data protection, however, nothing can contribute. For example, these: how far should social control be technicalized, professionalized, and remotely removed? For which purposes such systems are reasonable? Whether the corresponding research and development should be supported with taxpayers is still comparatively easy to answer.