EP4272159A1 - Procédé et système de détection d'un objet statique dans un environnement surveillé par au moins une caméra - Google Patents
Procédé et système de détection d'un objet statique dans un environnement surveillé par au moins une caméraInfo
- Publication number
- EP4272159A1 EP4272159A1 EP21847698.4A EP21847698A EP4272159A1 EP 4272159 A1 EP4272159 A1 EP 4272159A1 EP 21847698 A EP21847698 A EP 21847698A EP 4272159 A1 EP4272159 A1 EP 4272159A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- image
- camera
- alert
- scene
- images
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/254—Analysis of motion involving subtraction of images
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/103—Static body considered as a whole, e.g. static pedestrian or occupant recognition
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING SYSTEMS, e.g. PERSONAL CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
- G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
- G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
- G08B13/19602—Image analysis to detect motion of the intruder, e.g. by frame subtraction
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING SYSTEMS, e.g. PERSONAL CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B25/00—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems
- G08B25/01—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium
- G08B25/08—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium using communication transmission lines
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20212—Image combination
- G06T2207/20224—Image subtraction
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30112—Baggage; Luggage; Suitcase
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30232—Surveillance
Definitions
- TITLE Method and system for detecting a static object in an environment monitored by at least one camera
- the present invention relates to a method and a system for detecting a static object in an environment monitored by at least one camera.
- a suspicious package is luggage or any other object abandoned, either voluntarily with the aim of committing a malicious act, or involuntarily by thoughtlessness of its owner.
- Security procedures aim to protect the public, assess the threat posed by the suspicious package, and possibly destroy it.
- Security procedures may vary from one context to another and between operators, but generally consist of securing a perimeter around the suspicious package, deploying security personnel authorized to intervene and possibly implementing appropriate means to destroy the suspicious package.
- the object of the present invention is to solve this problem by proposing a solution adapted to places welcoming the public.
- the subject of the invention is a method for detecting a static object in an environment monitored by at least one camera, said method comprising: an acquisition step, by a camera, during a detection time window , of an image stream, each image of the image stream being indexed by an acquisition date and corresponding to the observation of an environmental scene; then, for each image of the stream of images, a preprocessing step consisting in: subtracting from the considered image, a background image corresponding to a background of the scene observed, to obtain an image of shapes, which contains shapes present in the observed scene; isolate, in the image considered, a silhouette for each person present in the observed scene, to obtain an image of silhouettes; and combining the image of shapes and the image of silhouettes associated with the image considered to obtain a processed image, which comprises only objects present in the observed scene, and finally, on a temporal succession of processed images corresponding to observation of the same scene of the environment, a step consisting in following from one processed image to another each object present in the environment and in issuing an alert as soon as an object
- This process makes it possible to drastically reduce the detection time of a suspicious package while presenting high reliability, such as to minimize false alerts and therefore to initiate security procedures only when they are justified.
- this method comprises one or more of the following characteristics, taken in isolation or in all technically possible combinations:
- the method further comprises a step of automatic identification of the alerts received from the cameras of the plurality of cameras in order to filter the false alerts.
- the identification step comprises a classification sub-step making it possible to assign, to a static object associated with an alert, a class from among a plurality of predefined classes corresponding to types of suspicious packages.
- the identification step comprises a sub-step making it possible to assign, to a static object associated with an alert, a class among a plurality of dynamic classes, the dynamic classes being defined during the implementation of the method, defined dynamic classes being stored in a whitelist.
- a predefined class corresponds to an infrastructure element belonging to the background of the scene whose appearance changes over time, an alert associated with an object belonging to said predefined class being deleted because it is considered a false alert.
- a predefined class corresponds to a new type of suspicious package, an alert associated with an object belonging to said predefined class being confirmed because considered as a real alert.
- the identification step consists in applying a classification algorithm of the artificial neural network type.
- the invention also relates to a detection system for implementing the above detection method.
- the system comprises at least one camera, the camera being capable of carrying out the steps of acquiring a stream of images, of preprocessing the images of the stream of images and of tracking the objects during a window detection time so as to emit an alert in the event of detection of a static object.
- the system comprises a control center in communication with said at least one camera and capable of filtering the alerts generated by said at least one camera by implementing the automatic identification step.
- This system has the ability to evolve based on operator feedback to better filter false alerts and thus adapt to the scene actually observed by the camera.
- FIG. 1 is a schematic representation of an embodiment of a detection system according to the invention.
- FIG. 2 is a schematic representation of an embodiment of the detection method according to the invention.
- the detection system 1 comprises a plurality of cameras 10-1, 10-2,..., 10-N, a communication network 20 and a control center 30.
- the function of the plurality of cameras is to monitor an environment, in particular a place open to the public, such as an airport for example.
- Each camera observes the environment from a specific point of view. Therefore, the scene observed by a camera is specific to this camera.
- the plurality of cameras comprises N cameras, which are referenced 10-1, 10-2, ... 10-N in Figure 1.
- the cameras are identical to each other.
- the structure of a camera for example the 10-N camera, will now be presented.
- a camera comprises an optical sensor 11, acquisition electronics 12, a computer 13 and a communication interface 17.
- the optical sensor 11 is able to observe a scene of the environment.
- the acquisition electronics 12 is capable of generating an image signal, or more simply an image in what follows, from the light collected by the optical sensor 11.
- the computer 13 comprises calculation means, such as a processor 14, and information storage means, such as a memory 15.
- the memory 15 notably stores computer programs, including the instructions, when they are executed by the processor 14, make it possible to implement certain functionalities.
- the memory 15 stores a pre-processing program 16, the execution of which allows the implementation of part of the detection method according to the invention, as will become apparent below.
- the communication interface 17 connects the camera to a network in order to exchange, in reception, camera parameter information (for example parameters for adjusting the optical properties of the sensor 11, or even program configuration parameters 16) and, in transmission, all or part of raw images or pre-processed images and/or suspicious package detection alerts.
- camera parameter information for example parameters for adjusting the optical properties of the sensor 11, or even program configuration parameters 16
- the plurality of cameras is connected to a communication network 20.
- This is for example a local area network (“Local Area Network—LAN”), using for example the Ethernet protocol.
- a control center 30 is for example a computer. It comprises a communication interface 32, a computer 33, a peripheral interface 37 and a plurality of peripherals 38.
- the communication interface 32 connects the control center 30 to the network 20, in order in particular to allow communication with the plurality of cameras, in reception, of image preprocessing information delivered by the various cameras and, in transmission, of camera control/command information.
- the computer 33 comprises calculation means, such as a processor 34, and information storage means, such as a memory 35.
- the memory 35 notably stores computer programs, including the instructions, when they are executed by the processor 34, make it possible to implement certain functionalities.
- the memory 35 stores an alert filtering program, the execution of which allows the implementation of part of the detection method according to the invention, as will become apparent below.
- the memory 35 also stores a white list 39 of the objects having led to the transmission of a false alarm, but which are not suspicious, belonging for example to the background of the scene observed although changing in appearance over time.
- the plurality of peripherals 38 allow an operator to interact with the control center 30.
- the plurality of peripherals 38 comprises for example a screen which can allow the viewing of the images acquired by such or such a camera, a mouse which can allow select such or such an area of an image displayed on the screen, and a keyboard that can allow the white list 39 to be updated, for example by entering a descriptor of the area selected on the screen.
- the detection system 1 is part of a distributed architecture in which certain tasks are remoted so as to be carried out in the cameras or at least as close as possible to the cameras (these are in particular the tasks of pre-processing of the images acquired by each camera individually) while other tasks are centralized in the control center (these are the tasks centralizing all the alerts reported by the cameras).
- the detection method 100 begins with a step 110 of acquisition of a stream of images by a camera (for example the camera 10-1 of FIG. 1), more precisely by the optical sensor and the electronics of associated acquisition.
- a camera for example the camera 10-1 of FIG. 1
- Each image of the stream is indexed with the instant t at which it was acquired.
- the image at time t is thus denoted l(t).
- the camera operating with a sampling time At the image which follows the image l(t) is the image l(t+At).
- An image corresponds to the observation of a scene, defined as the observation of an area of the environment according to a certain pose of the camera, that is to say according to an orientation of the optical axis of the camera and taking into account the optical properties of the latter (in particular the aperture angle of the camera).
- the second step 120 of the method 100 consists in applying pre-processing to each image l(t) acquired.
- it is the computer 13 which performs this step by executing the preprocessing program 16.
- a background image I0 is subtracted from the image l(t).
- an image of shapes is obtained at time t, denoted lf(t).
- the background image I0 is an image of the background of the scene observed by the camera 10-1. It is advantageously obtained by means of the camera 10-1 having enabled the acquisition of the image l(t) considered.
- the background image I0 corresponds to the observation of the scene when the latter is empty, i.e. when it contains neither person nor object (other than the objects belonging to the background of the scene) .
- the background image I0 is for example obtained by acquiring images at night, when the place is not frequented. It can also be obtained by associating portions of images which remain invariant over a long period of observation of the scene.
- the image of shapes lf(t) is a binary image, a pixel taking for example the value zero when the corresponding pixel of the image l(t) is found in the background image I0, and the value unit otherwise.
- the image of shapes if(t) comprises a priori different shapes, which correspond to people and objects, static or mobile, present in the scene observed at time t.
- This first sub-step is not sufficient because, in public environments where the density of people is high, the immobility of a user, present next to baggage, makes it difficult to distinguish the object from the user: in the image of forms there will be no border between these two forms, which will be merged into one and the same form.
- the pre-processing step 120 thus comprises a second sub-step 124, carried out in parallel with the first sub-step 122 and consisting in isolating, in the image l(t), a silhouette for each person present in the observed scene. An image of silhouettes at time t, denoted ls(t), is obtained as output.
- an algorithm of the people segmentation type is applied to the image l(t).
- Algorithms of the people segmentation type are known.
- An example of such an algorithm is presented in the article by Cao, Z., Martinez, G. H., Simon, T., Wei, S. E., & Sheikh, Y. A., “OpenPose: realtime multi-person 2D pose estimation using Part Affinity Fields”, 2019, IEEE transactions on pattern analysis and machine intelligence.
- a person segmentation type algorithm identifies whether a person is present in the observed scene. If so, a silhouette is associated with that person.
- the image of silhouettes groups together the silhouettes associated with each of the people present in the image I (t).
- the image of silhouettes ls(t) is a binary image, a pixel taking for example the value zero when the corresponding pixel of the image ls(t) is found inside the silhouette of a person present in the observed scene, and the unit value otherwise.
- the third sub-step 126 of the second step 120 consists in combining the images obtained at the end of the first and second sub-steps 122 and 124 from the same image l(t) of the stream in order to obtain an image d 'objects at time t, denoted lo(t).
- the image of objects lo(t) only includes the shapes present in the observed scene which correspond to objects, the shapes corresponding to people having been eliminated.
- the image of shapes and the image of silhouettes at time t, lf(t) and I s(t), are for example combined pixel by pixel, by application of an “AND” logical operator.
- a pixel of the image of objects taking for example the unit value when the corresponding pixel of the shape image If (t) is inside a shape and when the corresponding pixel of the silhouette image ls(t) is outside a silhouette, and the value zero otherwise.
- This combination step is particularly interesting since it limits the pixels of interest to a small number of pixels, which lightens the processing carried out downstream, in particular the tracking of objects.
- a fourth sub-step 128 the pixels of the image of objects I o(t) that can be associated with the same object are brought together in a region delimited by an outline (“blob” in English).
- a region detection algorithm is applied to the image lo(t) to define areas whose pixels have substantially constant properties, but which differ from the pixels of the surrounding areas. This is a classic approach based on edge detection and illustrated for example by the article by S. Suzuki, “Topological structural analysis of digitized binary images by border following”, Computer vision, graphies, and image processing, 30(1), 32-46, 1985.
- the output is an image processed at time t, denoted l*(t).
- the third step 130 of the method 100 consists in analyzing a temporal succession of processed images l*(t). For example, the images belonging to a detection time window, of a few seconds for example, are taken into account.
- a tracking or tracking algorithm is implemented here. This processing step is known as such. It is explained for example in the book by Maggio, E., & Cavallaro, A, “Video tracking: theory and practice. John Wiley & Sons” published in 2011.
- An alarm Ai is emitted for each object Oi among the objects present in the observed scene which is identified as immobile throughout the detection time window.
- a vignette Zi is isolated in the image l(t) around the object Oi. Only this vignette Zi is transmitted, with the alert Ai, to the control center 30 for the rest of the process 100.
- the detection method 100 comprises a fourth step 140 of automatically identifying the alerts intended to filter the alerts in order to limit the number of false alerts and confirm the real alerts among all the alerts Ai received from the plurality of cameras.
- Step 140 includes a first sub-step 142 of classification. It consists of running a classification algorithm on a Zi thumbnail to determine the class to which the corresponding Oi object belongs.
- the classification algorithm is configured to assign a class among a plurality of predefined classes.
- the plurality of predefined classes includes for example the classes “suitcases”, “handbag”, “backpack”, etc.
- the classification algorithm used is for example a suitably trained deep neural network.
- this classification algorithm is the one presented in the article by He, K., Zhang, X., Ren, S., & Sun, J., “Deep residual learning for image recognition”, in Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770-778, 2016.
- the output of sub-step 142 is a score S1 of membership in one of the plurality of predefined classes that the classifier can distinguish
- step 140 comprises a second sub-step 144 of determining membership of the white list 29.
- This step implements a whitelist membership determination algorithm such as the one presented in the article Chen, Y., Zhou, X. S., & Huang, T. S., “One-class SVM for learning in image retrieval”, in Proceedings 2001 International Conference on Image Processing, Vol. 1, p. 34-37, October 2001, IEEE.
- the white list 29 includes infrastructure elements that belong to the background of the scene but whose appearance changes over time: shadows of static equipment, automatic doors that open and close, advertising posters that vary, moving escalator, etc.
- infrastructure elements that belong to the background of the scene but whose appearance changes over time: shadows of static equipment, automatic doors that open and close, advertising posters that vary, moving escalator, etc.
- Such evolving infrastructure elements cannot be subtracted from the image during pre-processing, since it is not possible, when installing the camera, to know what the objects in the background of the scene are. actually observed which may lead to artefacts in the processed image, and consequently to false alarms.
- the white list then includes the infrastructure elements that belong to the background of the scene as well as their characteristics. It is dynamically defined during step 150, as detailed below.
- the sub-step 144 consists in determining whether the thumbnail Zi associated with an alert Ai in fact corresponds to one of the elements of the white list. More precisely, firstly, a neural network makes it possible to extract characteristic elements from the thumbnail, then, secondly, a comparator compares, using a machine learning algorithm, these characteristics with those of characteristic elements of the white list. If it is similar enough to an infrastructure item in the list whitelisted, the thumbnail is considered to match that whitelisted infrastructure element. This processing will therefore make it possible to identify the artefacts of the image which are due to known modifications of the background of the scene and which therefore should not raise an alert.
- the sub-step 144 is in fact also a classification step, but on a set of classes (the infrastructure elements of the white list) which evolves dynamically according to the history of the use of the detection system.
- Sub-step 144 calculates a second score S2.
- step 140 includes a third voting sub-step 146 making it possible to compare the first and second thresholds S1 and S2 with each other. If the second threshold is greater than the first for alert Ai, the alert is suppressed. Otherwise the alert Ai is validated and the validated alert Ai* is displayed on the screen so that the operator becomes aware of it and triggers the appropriate response.
- the method 100 includes a configuration step 150 making it possible to change the white list 29 during the use of the detection system and the implementation of the method, making it possible to increase the rate of rejection of false alarms.
- a sub-step 152 the operator verifies the exact nature of the validated alert Ai*.
- the images at the origin of the alert are presented to him.
- a sub-step 154 If it considers that it is a false alert, in a sub-step 154, it updates the white list. The operator adds this false positive to the whitelist of uninteresting objects so that objects similar to those present in this whitelist no longer return alerts thereafter.
- the method 100 identifies in a succession of images of the same scene any static object that does not correspond to an object of a type deemed uninteresting.
- the white list is shared at the central level and thus benefits all of the cameras 10-1 to 10-N, and not only the camera by means of which the so-called “uninteresting” object was observed. .
- a whitelist mechanism is implemented for infrastructure elements leading to a rejection of alerts classified as such, and a whitelist mechanism for new threats leading to the confirmation of alerts classified as such.
- each alert not discarded by the white list of infrastructure elements is submitted to an operator, who then determines whether or not the corresponding event constitutes a threat of a new kind, which is not part of of the set of predefined classes from step 142. If so, it then indicates what the observed object corresponds to.
- This object is included in an intermediate list, either as a new occurrence of an already existing threat category, or as the first occurrence of a new threat category. Once a category in this intermediate list has a sufficient number of samples, a new class is automatically created in the whitelist of new threats for automatic detection by implementing an algorithm similar to that of step 144.
- the prior art also does not make it possible to recognize non-invariant objects from the background of the observed scene as irrelevant.
- classification algorithms have a very high computational complexity.
- the preprocessing steps according to the invention make it possible to limit the computational complexity.
- a dynamic consideration of the real situation through the use of a white list makes it possible to reduce the number of predefined classes to follow.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Business, Economics & Management (AREA)
- Emergency Management (AREA)
- Human Computer Interaction (AREA)
- Image Analysis (AREA)
- Alarm Systems (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| FR2014290A FR3118522B1 (fr) | 2020-12-31 | 2020-12-31 | Procede et systeme de detection d'un objet statique dans un environnement surveille par au moins une camera |
| PCT/EP2021/087706 WO2022144346A1 (fr) | 2020-12-31 | 2021-12-28 | Procédé et système de détection d'un objet statique dans un environnement surveillé par au moins une caméra |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP4272159A1 true EP4272159A1 (fr) | 2023-11-08 |
Family
ID=75953940
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP21847698.4A Pending EP4272159A1 (fr) | 2020-12-31 | 2021-12-28 | Procédé et système de détection d'un objet statique dans un environnement surveillé par au moins une caméra |
Country Status (3)
| Country | Link |
|---|---|
| EP (1) | EP4272159A1 (fr) |
| FR (1) | FR3118522B1 (fr) |
| WO (1) | WO2022144346A1 (fr) |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8284249B2 (en) * | 2008-03-25 | 2012-10-09 | International Business Machines Corporation | Real time processing of video frames for triggering an alert |
| JP6241796B2 (ja) * | 2015-12-25 | 2017-12-06 | パナソニックIpマネジメント株式会社 | 置去り物監視装置およびこれを備えた置去り物監視システムならびに置去り物監視方法 |
-
2020
- 2020-12-31 FR FR2014290A patent/FR3118522B1/fr active Active
-
2021
- 2021-12-28 EP EP21847698.4A patent/EP4272159A1/fr active Pending
- 2021-12-28 WO PCT/EP2021/087706 patent/WO2022144346A1/fr not_active Ceased
Also Published As
| Publication number | Publication date |
|---|---|
| WO2022144346A1 (fr) | 2022-07-07 |
| FR3118522A1 (fr) | 2022-07-01 |
| FR3118522B1 (fr) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10289823B2 (en) | Camera system for traffic enforcement | |
| EP2013816B1 (fr) | Procede d'identification d'une personne par analyse de caracteristiques de ses cils | |
| EP3410413B1 (fr) | Génération améliorée d'événements d'alerte sur la base d'une détection d'objets à partir d'images de caméra | |
| Filonenko et al. | Unattended object identification for intelligent surveillance systems using sequence of dual background difference | |
| EP3285209B1 (fr) | Procede de surveillance au moyen d'un systeme multi-capteur | |
| EP2965287A1 (fr) | Procédé de suivi d'une cible dans une séquence d'images avec prise en compte de la dynamique de la cible | |
| WO2021110789A1 (fr) | Système et procédé de détection de fraude | |
| Tripathi et al. | Robust abnormal event recognition via motion and shape analysis at ATM installations | |
| FR3111452A1 (fr) | Procédé de protection automatique d’un objet, d’une personne ou d’une information ou œuvre visuelle vis-à-vis d’un risque d’observation non désirée | |
| EP2483590A1 (fr) | Système et procédé pour détecter la présence d'un sujet en zone sensible | |
| Saketh et al. | Intelligent surveillance support system | |
| EP4272159A1 (fr) | Procédé et système de détection d'un objet statique dans un environnement surveillé par au moins une caméra | |
| Chang et al. | Localized detection of abandoned luggage | |
| Tay et al. | Application of computer vision in the construction industry | |
| FR2997212A1 (fr) | Appairage d'un objet et d'un detenteur de cet objet | |
| Joshi et al. | Suspicious object detection | |
| Madhavan et al. | Violence Detection from CCTV Footage Using Optical Flow and Deep Learning in Inconsistent Weather and Lighting Conditions | |
| Pham et al. | Two-stage unattended object detection method with proposals | |
| EP2926294A1 (fr) | Procédé de détection de personnes et/ou d'objets dans un espace | |
| Raju et al. | Efficient abandoned luggage detection in complex surveillance videos | |
| Abirami et al. | A comparative analysis on ATM theft detection based on human behaviour | |
| FR3112467A1 (fr) | Magasin autonome de vente au détail en libre-service | |
| EP3067709A1 (fr) | Procédé et dispositif de suivi d'individus dans un lieu équipé de moyens répartis de détection | |
| Yang et al. | Surveillance system using abandoned object detection | |
| WO2015092040A1 (fr) | Procédé d'estimation du nombre de personnes et/ou d'objets dans un espace |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
| 17P | Request for examination filed |
Effective date: 20230630 |
|
| AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
| DAV | Request for validation of the european patent (deleted) | ||
| DAX | Request for extension of the european patent (deleted) | ||
| RAP1 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: HITACHI RAIL GTS FRANCE SAS |