WO2021064591A1

WO2021064591A1 - Improved method and system for camera security

Info

Publication number: WO2021064591A1
Application number: PCT/IB2020/059146
Authority: WO
Inventors: Frederick Jacobs
Original assignee: Secury360 Bv
Priority date: 2019-09-30
Filing date: 2020-09-30
Publication date: 2021-04-08
Also published as: BE1027605B1; BE1027605A1

Abstract

Method of securing an area, the method comprising digitally capturing (thermal) images; processing the captured (thermal) images; sending the (thermal) images upon detection of one or more unauthorised persons to a community comprising a plurality of users, wherein the users can individually evaluate received (thermal) images as being an alarm situation or not an alarm situation; wherein an alarm is initiated upon evaluation of the (thermal) images when the received (thermal) images are evaluated as an alarm situation by at least a predetermined number and/or percentage of users.

Description

I MPROVED METHOD AND SYSTEM FOR CAMERA SECURI TY

TECHNI CAL FI ELD

The invention relates to a method for securing an area. I n a second aspect, the invention also relates to a security system .

PRI OR ART

Cameras, and thermal cameras in particular, continue to make strides in the electronic security space, driven by declining prices and the ability of (thermal) cameras to power intelligent video surveillance solutions.

US 10 140 832 describes systems and methods for providing a surveillance and/or an alarm system . The system may comprise an imaging system , where the imaging system may comprise a camera. The camera can create thermal images based on infrared radiation. Furthermore, the document describes that the system can comprise a controller that can analyse the images and determine the presence of an object of interest. The controller can detect deviations in behaviour and generate an alarm based on this. Alternatively, the controller can generate a communication in which it must be determined whether it is an emergency situation and whether help must be sent. After the image has been processed, the image can be displayed, saved or sent back to a user device.

US 8 131 012 describes a first element that may be configured to analyse video frames to identify targets of interest. The analysis is then forwarded to a second element. The second element analyses the received data, builds semantic representations of the behaviours / events depicted in the video frames, determines patterns and learns from the observed behaviour to identify normal and/or abnormal events. Data describing a normal (or abnormal) behaviour / event, along with the semantic labels that can be applied to such an event, can be sent to an output device to send out a warning (for example, a warning message presented on a GUI interface screen).

US 2012/307066, US 2018/130335 describe alarm systems and methods where, however, standard control rooms are still used and the margin of error remains extremely high, while the speed of processing often also remains problematic. Other documents, such as ‘Abnormal behaviour detection on queue analysis from stereo cameras’ by Luis Patino et ai . , Mei Wang et al. their ‘Deep face recognition: a survey', Bauer et al. their ‘Biometric verification of persons based on thermovision’, US 2015/049191 and US 2007/182540 describe related themes, but fail to eliminate the well-known issues.

The present invention aims at least to improve known security systems and methods. The object of the invention is to provide an improved security method and security system .

SUMMARY OF THE I NVENTI ON

To this end, the invention provides a method according to claim 1 . Preferred embodiments of the method are set out in claims 2 to 13. In a second aspect, the present invention relates to a security system according to claim 14. Preferred embodiments of the method are set out in dependent claims 15 to 17.

I n addition, the present invention allows potential alarm situations to be identified before they occur. Further advantages and effects are described in the detailed description.

DESCRI PTI ON OF THE DRAWI NGS

Figure 1 shows a schematic representation of an embodiment of the method according to the present invention.

DETAI LED DESCRI PTI ON

Unless otherwise defined, all terms used in the description of the invention, including technical and scientific terms, have the meaning as commonly understood by a person skilled in the art to which the invention pertains. For a better understanding of the description of the invention, the following terms are explained explicitly. In this document, ‘a’ and ‘the’ refer to both the singular and the plural, unless the context presupposes otherwise. For example, ‘a segment' means one or more segments.

When the term ‘around’ or ‘about’ is used in this document with a measurable quantity, a parameter, a duration or moment, and the like, then variations are meant of approx. 20% or less, preferably approx. 10% or less, more preferably approx. 5% or less, even more preferably approx. 1 % or less, and even more preferably approx. 0.1 % or less than and of the quoted value, insofar as such variations are applicable in the described invention. However, it must be understood that the value of a quantity used where the term ‘about’ or ‘around’ is used, is itself specifically disclosed.

The terms ‘comprise’, ‘comprising’, ‘consist of’, ‘consisting of’, ‘provided with’, ‘include’, ‘including’, ‘contain’, ‘containing’, are synonyms and are inclusive or open terms that indicate the presence of what follows, and which do not exclude or prevent the presence of other components, characteristics, elements, members, steps, as known from or disclosed in the prior art.

The term ‘thermal image’ as used herein refers to an image obtained by recording heat or infrared radiation (or energy) emanating from objects and living organisms.

The term ‘thermal pattern’ as used herein refers to the temperature profile of an object or living organism in a thermal image.

Unless otherwise stated, the term ‘authorised person' as used herein refers to a person who is authorised to enter an area, for example, a postman or an owner of the area.

The term ‘alarm situation’ as used herein refers to a situation where an unauthorised person is within an area and engages in suspicious behaviour (e.g., a burglary or a robbery).

Quoting numerical intervals by endpoints comprises all integers, fractions and/or real numbers between the endpoints, these endpoints included. In a first aspect, the invention relates to a method for securing an area, the method comprising: digitally capturing images; processing the captured images for detection of persons, comprising detecting shapes, in the captured images, wherein the processing is performed based on shape and pattern, of the detected images, said processing comprising identifying the detected shapes as one or several people; further processing the images if in the first processing one or more persons were identified in the detected forms, the further processing comprising classifying the detected persons as authorised or as unauthorised based on behaviour; sending the images upon detection of one or more unauthorised persons to a community comprising a plurality of users, the users each individually evaluating received images as being an alarm situation or not an alarm situation;

In this regard, an alarm is initiated upon evaluation of the images when the received images are evaluated as an alarm situation by at least a predetermined number and/or percentage of users.

The applicant noted that in most security systems, a large number of false positives are allowed and communicated as real alarm situations. In practice, this concerns, for example, animals, wind or innocent human movements, wherein no or insufficient differentiation is made. To reduce these false positives, further processing should be performed on initial results. By allowing this further processing to take place at several levels, which differ greatly in terms of procedure, the applicant achieves a strong reduction in false positives. To this end, use is made of, among other things, an (Al) classification of the images observed as relevant, which produces a first filter. Furthermore, use is also made of a human review of the situation, which can remove further false positives, and in particular those where human shapes are detected, which, however, do not appear to warrant triggering an alarm (mistakes in interpretation of movements, errors in images, etc.). Moreover, by working with a predetermined ‘threshold value’ in the human review, a correct evaluation can be quickly reached, which is not the case with any existing technology. Finally, in the present invention there is preferably no limitation with regard to members of the community, and it does not only concern (locally) trained security personnel. This reduces the burden on such personnel, but also avoids bias or misconceptions that can prevail in certain populations, and ultimately increases the likelihood of recognising both risky behaviour and behaviour triggering false alarms (by allowing a broader spectrum of the population can be present in such a community). One of the most important aspects concerns the individual evaluation by the users. Whereas in existing a report can enter a control room, it can possibly be viewed there by several people. However, these will influence each other in an unstoppable way in their evaluation, ranging from small behaviours to comments and statements. The present invention explicitly counteracts this by decentralising the evaluation to ensure objective evaluation. I n addition, the current method also opens up the group of users further than the typical operators in a control room, further avoiding bias, as well as time pressure (in control rooms several calls can come in at the same time, and a shift typically lasts 8 hours, allowing for loss of concentration, but also a discolouration of the evaluation ability). It should be noted here that the captured images are not used for the processing of personal information, and in that way do not violate existing privacy legislation (GDPR). I n certain embodiments, the images will also not be saved and/or faces automatically made unidentifiable before members of the community can access them .

In a preferred embodiment, an alarm is initiated when the received images are evaluated as an alarm situation within a predetermined time span by a predetermined number of users, the predetermined number of users being at least 3 (preferably more, such as 4, 5, 8, 10; 15 or more users, partly depending on time span and size of available community, but also on the requirements of the owner / manager of the area / client), and where the predetermined time span is at most 1 minute, preferably at most 30 seconds, and even more preferably at most 10 seconds. Preferably, an additional requirement of a certain minimum percentage of the users within a time span applies (whether or not the same as the predetermined time span, preferably the same, more preferably even earlier, such as, for example, the moment when the required number is reached, and the further evaluation may or may not be discontinued). This applies to both the method and the system .

The time span in question should be regarded as the time elapsed since the images were sent to the users, assuming extremely limited delay between sending and receiving. Other numbers of predetermined users are of course also possible and can be adjusted according to the user’s requirements (more or less, depending on criticality), but also according to the available users in the community.

The predetermined time span can also be adapted to the situation. I n some areas, a very fast response is extremely important and must be kept very low, while in others a longer turnaround time is less important, but rather a great level of certainty (for example when it comes to remote areas, or zones that cannot / may not be entered just like that).

In a preferred embodiment, the invention relates to a method for protecting an area. The method specifically comprises digitally capturing thermal images. The method further comprises processing the captured thermal images for detection of persons, comprising detecting thermal shapes in the captured thermal images, wherein the processing is performed based on shape and thermal pattern of the detected thermal shapes. The said processing herein comprises identifying the detected thermal shapes as one or more persons, and the method further comprises the further processing of the thermal images if one or more persons were identified in the detected thermal shapes in the first processing. The further processing comprises classifying the detected persons as authorised or as unauthorised based on behaviour. Subsequently, the thermal images are sent upon detection of one or more unauthorised persons to a community comprising a plurality of users. The users can each individually evaluate received thermal image as being an alarm situation or not an alarm situation. An alarm is initiated upon evaluation of the thermal images when the received thermal images are evaluated as an alarm situation by at least a predetermined number and/or percentage of users.

The present invention improves the reliability of security systems. The use of a thermal camera makes it possible to leverage a community to reduce false detections. The use of images from conventional security systems is often strongly limited by privacy regulations (including GDRR) , so that only certified persons are allowed to process them , typically under further strict conditions. This leads to a situation that is not feasible in practice to be able to intervene quickly in alarm situations, partly due to the limited number of ‘analysts’ available. By using thermal cameras, and thus thermal images, an arbitrary user in the community cannot see (and therefore cannot recognise) the face or body of a detected person, thus ensuring the privacy of the person. In this way, the community can be opened up much more widely, and with a larger audience, a faster response and more reliable judgement can be assured. For example, by having the burglary confirmed by different (independent) persons, it can be confirmed with greater certainty that a potential emergency situation is actually occurring. I n addition, the method allows potential unauthorised persons to be quickly identified before an alarm situation actually occurs. I n this way, breaches or other alarm situations can be prevented.

‘Initialising an alarm’ can comprise one or more of the following, but is not limited thereto: notifying law enforcement agencies (official such as police, but also unofficial such as security firms and the like), notification of owner / responsible party / security officer for the area, activation of a visual / audible alarm to deter the offender.

In a preferred embodiment, the alarm is initiated when the received (thermal) images are evaluated as an alarm situation by at least the predetermined number and/or percentage of users within a predetermined time span.

Defining a time span ensures, on the one hand, that an alarm is initiated in time to at least deter the burglars. On the other hand, it is ensured that an authorised person (e.g. the police, the neighbourhood watch or the owner(s) of the protected area) can respond in a timely manner.

In a further preferred embodiment, the alarm is initiated upon evaluation of the (thermal) images when the received (thermal) images are evaluated as an alarm situation by at least 80% preferably by at least 81 % more preferably by at least 82% still more preferably by at least 83% still more preferably by at least 84% still more preferably by at least 85% still more preferably by at least 86% still more preferably by at least 87% still more preferably by at least 88% still more preferably by at least 89% still more preferably by at least 90% of the users.

Specifying the percentage of users who evaluate the (thermal) images ensures that an expected reliability is obtained. The inventor of the present invention has found that if 80% of the users evaluate the (thermal) images, a reliability of at least 92% is obtained. If 90% of the users evaluate the (thermal) images, the reliability increases to at least 98% . In a further preferred embodiment, the alarm is initiated when the received (thermal) images are evaluated as an alarm situation by at least the predetermined number and/or percentage of users within 1 minute, preferably within 55 seconds, more preferably within 50 seconds, still more preferably within 45 seconds, still more preferably within 40 seconds, still more preferably within 35 seconds, still more preferably within 30 seconds, still more preferably within 25 seconds, still more preferably within 20 seconds, still more preferably within 15 seconds, still more preferably within 10 seconds.

The applicant notes that in practice a few minutes is enough for an unauthorised person to enter and exit an area or a building. By initialising an alarm within 1 minute, the unauthorised person can be deterred before committing a serious crime. Furthermore, the applicant has noticed that by initialising the alarm within 10 seconds, the chance of deterring the unauthorised person is increased. I n addition, an authorised person (e.g. the police, the neighbourhood watch or the owner(s) of the protected area) can respond more quickly.

In a further alternative preferred embodiment, the alarm is initiated upon evaluation of the received (thermal) images as an alarm situation by at least three users within at most 2 minutes, preferably within at most 1 minute, more preferably within at most 50 seconds, still more preferably within at most 40 seconds, still more preferably within at most 30 seconds, still more preferably within at most 20 seconds, still more preferably within at most 10 seconds. Preferably, the (thermal) images are evaluated by at least four, more preferably by at least five, more preferably by at least six, more preferably by at least seven users.

In a preferred embodiment, every user has a user score. The user score is preferably adjusted on the basis of correct or incorrect evaluation. Preferably, the user score falls if the user has incorrectly evaluated the police alarm situation. Preferably, the user score increases if the user has correctly evaluated the potential alarm situation.

The system allows the method to also include the reliability of the users in evaluating whether an emergency situation occurs or not. I n this way, the reliability and accuracy of security is increased.

The method may comprise rejecting the evaluation of users with a user score below a predetermined value. This increases the accuracy of the assessment. In a preferred embodiment, the method further comprises confirming whether there is an alarm situation or not by a user from whom the alarm situation originated. Correct feedback is beneficial to ensure correct operation of the method.

In a preferred embodiment the community is at least partially in a different time zone from the area under protection. An advantage can be found in the alertness of the users who evaluate the (thermal) images. Users in a different time zone can easily evaluate night-time alarm situations in the early morning, during the day or in the late evening. Preferably the time difference between at least part of the community and the area under protection is at least 3 hours, more preferably at least 4 hours, still more preferably at least 5 hours, still further preferably at least 6 hours.

In a preferred embodiment, the (thermal) images are sent to the users via a mobile application, whereby the users evaluate whether the (thermal) images are an alarm situation or not an alarm situation via the application. An application allows the images to be instantly sent to the users and the evaluation of the users to be processed. Time loss is minimised between registering potentially unauthorised persons and confirming whether or not an alarm situation is present.

Preferably, the images are delivered to the users via a mobile application or app on a smartphone, tablet or similar portable electronic devices, for example via a push notification. This makes it possible to react very quickly to the images, and to respect the predetermined time span.

In particular embodiments, the users themselves can set the times or periods for which they are available as evaluators. Alternatively (or additionally) this can be done based on location (time zone).

In a preferred embodiment, the images are provided to the users as a clip of at most 10 seconds, preferably at most 5 seconds. This length is sufficient for a user to thoroughly evaluate the images and provides sufficient time to estimate whether it is a person or not, and whether it is an alarm situation or not.

In a preferred embodiment, processing the perceived (thermal) images comprises defining a zone in the (thermal) images and evaluating motion in the zone. Preferably, evaluating movement comprises evaluating whether (thermal) shapes are moving in a particular direction, (thermal) shapes enter the zone and/or (thermal) shapes are in the zone for a particular period of time. I n order to optimize the method, and thus increase the reliability and/or accuracy, it is advantageous to define a zone under surveillance in (thermal) images.

In a preferred embodiment, the further processing comprises evaluating the distance travelled by one or more (thermal) shapes in the (thermal) images within a certain time, the trajectory travelled by one or more (thermal) shapes in the thermal images, the posture of one or more (thermal) shapes in the (thermal) images, the number of observed (thermal) shapes in the (thermal) images and/or the time at which one or more (thermal) shapes are recorded in the (thermal) images. The accuracy of classifying an identified person as authorised or unauthorised is increased. In a preferred embodiment, the (thermal) images are processed and/or further processed on the basis of artificial intelligence. Preferably, the (thermal) images are processed and/or further processed on the basis of Deep Learning. Processing via artificial intelligence is beneficial for increasing the reliability and accuracy of the surveillance. Artificial intelligence, and more specifically Deep Learning, allows to learn from the already registered (thermal) images and evaluations, and the method for each evaluation becomes more accurate in identifying persons from (thermal) shapes and classifying the registered persons as authorised or unauthorised. In addition, this reduces the workload of the community. In a preferred embodiment, the method comprises evaluating whether one or more (thermal) shapes show behaviour analogous to one or more authorised persons. The behaviour (such as speed of walking, direction of walking, posture, etc.) of a person is unique to the person and can thus be used to distinguish an authorised person from an unauthorised person. The applicant has found that comparing the behaviour improves the reliability and accuracy of the method.

In a preferred embodiment, the method comprises learning a historical movement pattern. The historical movement pattern can comprise motion paths. Each of the motion paths may comprise a movement probability associated with the motion path. The movement probability can be associated with a percentage and/or frequency with which the object traverses the motion path. For example, the motion path comprises an associated percentage of 61 %, indicating that 61 % of the time the object travels within the area shown by the historical movement pattern. The processing and/or further processing of the (thermal) images can be carried out via a cloud, using one or more cloud devices, by sending the (thermal) images over the cloud. After the (thermal) images have been processed, the (thermal) images can be displayed, stored or sent back to a user device and displayed or stored on a user device.

In a preferred embodiment, the (thermal) images are processed and further processed in real time. The time scales for the processing and further processing of the (thermal) images may differ. I n an embodiment, the (thermal) images are processed image by image, while the thermal frames are further processed every N frames. Here, N frames refers to frames where a person has been identified by the processing.

In a preferred embodiment, an authorised user (for example an owner of the area under surveillance) can remotely view the (thermal) images in real time, preferably via an application.

In a second aspect, the invention relates to a security system comprising an alarm system. Preferably, the security system further comprises one or more, preferably thermal, cameras configured for digitally capturing (thermal) images. Preferably, the security system further comprises one or more analytical elements. Preferably, at least one element is configured to process the captured (thermal) images. Preferably, the processing comprises detecting (thermal) shapes in the captured (thermal) images, for the detection of persons. Preferably, the at least one analytical element performs said processing based on shape and thermal pattern of the detected (thermal) shapes in the (thermal) images. Preferably, said processing comprises identifying the detected (thermal) shapes as one or more persons. Preferably, at least one element is configured to further process the observed (thermal) images upon detection of one or more persons, thereby classifying the detected persons as authorised or as unauthorised on the basis of behaviour. Preferably, at least one analytical element processes the (thermal) images using artificial intelligence, preferably using Deep Learning.

Preferably, the security system further comprises a community comprising a plurality of users, each user receiving the (thermal) images upon detection of one or more unauthorised persons by the one or more analytical elements, the users each individually evaluating the received (thermal) images as being an alarm situation or not an alarm situation. Preferably, an alarm is initiated during evaluation of the (thermal) images when the received (thermal) images are evaluated as an alarm situation by at least a predetermined number and/or percentage of users.

Preferably, an alarm is initiated upon evaluation of the images when the received images are evaluated as an alarm situation by at least a predetermined number of users within a predetermined time span, the predetermined number of users being at least 3, and the predetermined time span being 1 minute at most, preferably at most 30 seconds, and still more preferably at most 10 seconds.

In a preferred embodiment, the one or more analytical elements comprise a Machine Learning platform . The platform can be configured to learn, evaluate and/or remember. The platform can identify certain behaviour as abnormal.

In a preferred embodiment, the alarm system comprises an audible alarm. The alarm system can comprise at least one light element.

The alarm system, the one or more thermal cameras and/or the one or more analytical elements can comprise components that are distributed over one or more devices. The devices can be connected to each other by means of Bluetooth, Wi-Fi, a blockchain network, near-field communication, local area network and/or another wireless network.

The alarm system , the one or more thermal cameras and the one or more analytical elements can be parts of one device. Alternatively, the alarm system , the one or more thermal cameras and the one or more analytical elements can be parts of different devices.

In a preferred embodiment, the security system comprises a behaviour recognition system configured to analyse, learn and recognise behaviour in the (thermal) images.

The security system can further comprise one or more sensors. The one or more sensors can be motion sensors.

In what follows, the invention is described by way of non-limiting examples or figures illustrating the invention, and which are not intended to and should not be interpreted as limiting the scope of the invention. EXAMPLES

The invention will now be further explained on the basis of the following example, without however being limited to this.

EXAMPLE 1

FIG. 1 relates to a flow chart schematically representing an embodiment of the method according to the present invention.

An owner of the area under surveillance can indicate that the area is under surveillance, in other words: enable the surveillance (1 ). Once the surveillance has been activated, one or more thermal cameras will capture (thermal) images (2). The one or more cameras can also capture (thermal) images and optionally save them if surveillance is disabled. The captured (thermal) images are processed, whereby the presence of people in the (thermal) images is evaluated (3). If the (thermal) images comprise one or more persons, the (thermal) images are further processed, whereby the one or more identified persons are classified as being authorised or unauthorised (4). If one of the one or more persons is classified as unauthorised, the (thermal) images are sent to a community (6). The community will evaluate the (thermal) images as being an alarm situation or not an alarm situation (7). Preferably, if the (thermal) images are evaluated as being an alarm situation by at least a predetermined number and/or percentage of users, an alarm is initiated (8). The alarm may comprise a silent alarm (for example, automatically alerting the police, neighbourhood watch and/or an owner of the area under surveillance). The alarm may also comprise an audible alarm in the area under surveillance.

EXAMPLE 2

This example concerns a comparison of the present invention when using thermal imaging and thermal cameras, with other known security systems.

The protection according to the present invention was compared with ten other known security systems. The quality of night vision, the reliability, the accuracy, the possibility of problems with privacy legislation and the ability to detect an alarm situation before an alarm situation occurred was tested. The results are displayed in Table 1 . Night vision quality, reliability and accuracy were evaluated using ‘+ ’; the more '+ ’ the better the night vision quality, reliability or accuracy. The potential for privacy legislation issues and the ability to detect an alarm situation before an alarm situation occurred was evaluated as present (‘+ ') or absent

It is clear from Table 1 that, compared to the ten other known security systems, the method and security system according to the present invention resulted in a general improvement of known security systems. The present invention ensures that the quality of night vision (and visibility in obstructing weather conditions, such as rain or fog) is improved, while the chance of privacy legislation issues is eliminated. In addition, reliability and accuracy are also improved. Furthermore, it is also ensured that an alarm situation is possibly registered before it actually occurs, which is not possible with the other known security systems. Existing systems typically work in a way that only recognises alarm situations when or after they occur (i.e. if an intrusion has already occurred, and someone has entered a site without permission, has broken a window or forced a lock). By correctly interpreting these situations, and having them evaluated in a particularly fast way - which is only possible with the community system of the invention - it is possible to prevent problems, for example by triggering an alarm , by immediately sending personnel to a location, by switching on lights and thus chasing away a possible burglar. Existing systems wait until an alarm situation has already progressed too far and certain actions have already been taken, while the aim of the present invention is precisely to prevent them .

Table 1

EXAMPLE 3

Example 3 relates to a comparison between different security systems based on thermal imaging.

Four different security systems based on thermal imaging were compared. The first security system comprises capturing thermal images. The second security system comprises capturing thermal images and using a community to analyse the thermal images. The third security system comprises capturing thermal images and processing the thermal images, preferably via artificial intelligence. The last security system comprises capturing thermal images, processing the thermal images, preferably via artificial intelligence, and evaluating the thermal images by a community.

The different security systems were compared on the basis of reliability, accuracy and timely initiation of an alarm (timely notification of an authorised person and/or timely initiation of an alarm present in the area under surveillance). The results of the comparison are displayed in Table 2. It is clear from the results that a community mainly ensures improved reliability and accuracy, while processing the thermal images mainly ensures that an alarm is initiated in time in case of an evaluated alarm situation. Moreover, it becomes clear from the results that the combination of processing the thermal images and evaluating the thermal images by a community provides an improvement over the other three security systems.

Table 2

Claims

CLAI MS

1 . A method for securing an area, the method comprising

- digitally capturing images, preferably thermal images;

- processing the captured images for detecting persons, comprising detecting shapes, preferably thermal shapes, in the captured images, wherein the processing is performed on the basis of shape and pattern, preferably thermal pattern, of the detected images, said processing comprising identifying the detected shapes as one or more persons;

- further processing the images if in the first processing one or more persons were identified in the detected forms, the further processing comprising classifying the detected persons as authorised or as unauthorised based on behaviour;

- sending the images upon detection of one or more unauthorised persons to a community comprising a plurality of users, the users each individually evaluating received images as being an alarm situation or not an alarm situation; wherein an alarm is initiated upon evaluation of the images when the received images are evaluated by at least a predetermined number of users within a predetermined time span as an alarm situation, the predetermined number of users being at least 3, and the predetermined time span being at most 1 minute, preferably at most 30 seconds, and still more preferably at most 10 seconds.

2. Method according to the preceding claim 1 , wherein the images are sent via a mobile application to a wearable electronic device, preferably a smartphone or tablet, of the users, and wherein the users evaluate the images via the application as being an alarm situation or not an alarm situation.

3. Method according to any of the preceding claims 1 or 2, the images being thermal images, and the shapes being thermal shapes, and the pattern being a thermal pattern.

4. Method according to any of the preceding claims 1 to 3, wherein the alarm is initiated on evaluation of the images when the received images are evaluated by at least 80%, preferably by at least 90% , of the users as being an alarm situation.

5. Method according to the preceding claims 1 to 3, wherein the alarm is initiated on evaluation of the received images if an alarm situation is evaluated by at least three users within 10 seconds at most.

6. Method according to at least one of the preceding claims, wherein each user has a user score, wherein the user score is adjusted on the basis of correct or incorrect evaluation.

7. Method according to at least one of the preceding claims, the method further comprising the step of confirming whether there is an alarm situation or not by a user from whom the alarm situation originated, the step of confirming occurring later than the initiation of the alarm.

8. Method according to at least one of the preceding claims, wherein the community is at least partially in a different time zone than the area under protection.

9. Method according to the preceding claim 8, wherein the time difference between at least a portion of the community and the area under protection is at least 4 hours.

10. Method according to at least one of the preceding claims, wherein the images are sent to the users via a mobile application, and wherein the users evaluate the images via the application as being an alarm situation or not an alarm situation.

1 1. Method according to at least one of the preceding claims, wherein the processing of the observed images comprises defining a zone in the images and evaluating movement in the zone, preferably evaluating movement comprises evaluating whether shapes move in a particular direction, shapes enter the zone and/or shapes have been in the zone for a certain period of time.

12. Method according to at least one of the preceding claims, wherein the further processing comprises evaluating the distance travelled by one or more shapes in the images within a certain time, the trajectory travelled by one or more shapes in the images, the posture of one or more shapes in the images, the number of detected shapes in the images and/or the time that one or more shapes are recorded in the images.

13. Method according to any of the preceding claims, wherein the images are processed and/or further processed by means of artificial intelligence, preferably by means of Deep Learning.

14. Method according to any of the preceding claims, wherein the method comprises evaluating whether one or more shapes show behaviour analogous to one or more authorised persons.

15. A security system comprising

- an alarm system ;

- one or more cameras, preferably thermal cameras, configured to digitally capture images, preferably thermal images;

- one or more analytical elements, wherein at least one element is configured to process the captured images, comprising detecting shapes, preferably thermal shapes in the captured images, for detecting persons, wherein the at least one analytical element performs said processing based on shape and pattern, preferably thermal pattern, of the detected shapes in the images, said processing comprising identifying the detected shapes as one or more persons; and wherein at least one element is configured to further process the observed images upon detection of one or more persons, thereby classifying the detected individuals as authorised or unauthorised based on behaviour, at least one analytical element processing the images using artificial intelligence, preferably using Deep Learning;

- a community comprising a plurality of users, each user receiving the images upon detection of one or more unauthorised persons by the one or more analytical elements, the users each individually evaluating the received images as being an alarm situation or not an alarm situation; wherein an alarm is initiated upon evaluation of the images when the received images are evaluated by at least a predetermined number of users within a predetermined time span as an alarm situation, the predetermined number of users being at least 3, and the predetermined time span being at most 1 minute, preferably at most 30 seconds, and still more preferably at most 10 seconds.

16. Security system according to the preceding claim 15, wherein the cameras comprise thermal cameras, and are preferably thermal cameras, and wherein the images comprise thermal images, and are preferably thermal images.

17. Security system according to any of the preceding claims 15 or 16, wherein the one or more analytical elements comprise a Machine Learning Platform .

18. Security system according to at least one of the preceding claims 15 to 17, wherein the alarm system comprises an audible alarm .

19. The security system according to at least one of the preceding claims 15 to 18, wherein the analytical elements are adapted to send the images to a mobile application on a wearable electronic device of the users, and wherein the users evaluate the images via the application as being an alarm situation or not an alarm situation.

20. Security system according to at least one of the preceding claims 15 to 19, wherein the security system is adapted to perform the method according to at least one of the preceding claims 1 to 14.