WO2006075352A1

WO2006075352A1 - Surveillance method and surveillance device operating with said method

Info

Publication number: WO2006075352A1
Application number: PCT/IT2006/000015
Authority: WO
Inventors: Franco Valentini; Renato Grassi
Original assignee: Franco Valentini; Renato Grassi
Priority date: 2005-01-14
Filing date: 2006-01-12
Publication date: 2006-07-20
Also published as: ITMN20050003A1

Abstract

The invention relates to a surveillance method, wherein video and audio data are acquired through sensor means, converted to digital form and compared to pre-stored reference data for identifying an anomalous situation or emergency; an alarm is launched when acquired data match the reference data. A surveillance device operating with said method is also disclosed.

Description

Surveillance method and surveillance device operating with said method

The present invention relates to a surveillance method for detecting and reporting an emergency or alarm situation such as, for example, an attempted robbery. The invention furthermore relates to a surveillance device adapted to operate in accordance with said method.

As non-limitative examples, the invention applies to the surveillance of shops, banks, shopping malls, etc ...

According to prior art, video surveillance is largely based on cameras connected to video-recording systems, or connected to monitors with real-time human surveillance. In this latter case, the police or other authorities can be alerted by pressing an emergence button.

These systems are commonly used in supermarkets, banks and so on, but are not fully satisfactory. Video recording systems do not ensure that alarm is given promptly and generally require additional room for external devices to be connected to the camera (such as video recorder, monitor or PC); manned systems on the other side are very expensive due to the constant presence of an operator.

The object of the present invention is to eliminate the above disadvantages of known surveillance methods and systems. The object is achieved with a surveillance method, for detecting or reporting hazard or alarm situations, comprising the following steps: a) acquiring or recording images or audio through sensor means; b) converting audio/video data from the above step into digital format; c) comparing digital data resulting from above step with pre-stored audio/video reference data, by means of a microprocessor. d) sending an alarm signal to the outside world via a suitable communication interface, when a match occurs between said digital data and said reference data.

The invention, as better explained hereinbelow, makes use of techniques for machine-recognition of some features of objects, e.g. the shape, colour or direction; an image sample (pattern), a chromatic or light variation. Such techniques are known, but generally used in conjunction with a computer and for other and more simple purposes, for example in the field of production lines, to detect manufacturing defects and products to be discarded.

Comparison between acquired data and reference data may be performed in real-time, that is at the same time of acquisition, for better security and promptness, or even subsequently. The acquired data can, if necessary, be stored in a storing device like a hard disk or equivalent.

Said comparison between acquired data and reference data, according to preferred embodiments, provides for recognition of shapes, colours, dynamic situations and proportions between objects.

The occurrence of a match is determined by a similarity between acquired data and recorded data. According to one aspect of the method, two or more events may be considered to identify whether a situation is to be deemed anomalous or not, in order to reduce false alarms. The invention also provides a dedicated device adapted to operate with the above method, preferably a surveillance camera wherein hardware and software means adapted to implement the above method are embedded.

More in detail, one aspect of the invention consist in a surveillance camera equipped with processing means, memory means and an interface for connecting said camera with the outside world to provide, in the event of necessity, an alarm signal.

The camera is able to process the audio/video components of an event in progress, compare them with reference data stored into said memory means and activate the alarm signal in the event of a match between acquired data referring to real-life situation and said reference data.

The camera may use one or more of the following techniques: speech recognition; recognition of a predetermined keyword, e.g. a keyword indicating that an alarm is to be issued; image recognition; recognition of posture or gestures of a person.

As stated above, two or more events may determine an alarm. For example, the alarm can be issued on the basis of a sequence of two or more events recognized by said camera, or recognising either the voice and/or the gestures/posture through the mechanism of the succession of preset logical events. The above described method can also be implemented through a digital video recorder (DVR), or other devices (e.g. camera, microphones, ...) connected to external processing means like a suitable hardware or a computer running suitable program code; the embedded solution however is preferred for its simple and quick installation.

The main advantage of the invention is the automatic and unmanned surveillance operable 24 hours a day without excessive costs. The integration in a camera as an embedded system gives the further advantage of avoiding the use of a personal computer or other external devices connected to the surveillance camera.

The invention enables an alarm to be activated simply through recognising a correct temporal sequence of gestures or a combination of speech and gesture, thus being effective even in noisy environments.

A more detailed disclosure is set out below with the help of the attached drawings that show some applicational examples thereof by means of non- limitative example, in which:

Fig. 1 illustrates, in a schematic manner, an application of the invention for the surveillance of premises such as, for example, a bank;

Fig. 2 shows an essential block diagram of an alarm camera according to the invention.

With reference to the figures, the invention relates to a method for detecting and reporting an emergency or alarm situation, that substantially comprises the following steps: detecting the image and audio of an event in progress; converting the detected data into digital form; comparing the digital data thus obtained by means of a microprocessor with further prestored audio/video reference data; sending an alarm signal to the outside world via a suitable communication interface, when detected data match the reference data. Referring with greater detail to Fig. 1, there is illustrated an implementation of the method by means of a camera 100 connected to a police station by means of an interface IP, e.g. the Internet.

Other suitable interfaces are UMTS, GSM, GPRS, EDGE, WLFI, 4G, satellite, etc.; in general any type of known type of communication interface can be used, including wireless systems.

The camera 100 (Fig. 2) essentially comprises sensor means 1, processing means 2, memory means 3 and an interface 4 of known type.

In essential terms, the camera 100 operates according to the following method: the camera 100 acquires video and audio data through sensor means 1; said data are converted into digital format and then compared with reference data by processing means 2. The processing and operating procedures (software) and said reference data are stored in memory means 3.

The processing of the acquired data preferably comprises recognition of shapes, colours, dynamic situations and proportions between the objects. According to particularly preferred implementations, the method may comprise one or more of the features explained hereinbelow.

A first aspect of the invention provides for recognising colours and in particular the colour of the skin, which enables a better detection of positions of human beings and/or of certain encoded actions. As an example, the method may provide for detecting the hands and the face of a human being in relation to other objects or parts of the body, thus detecting a situation wherein a man is lying on the ground, through recognising the colour of the hair, the hands, a given garment or overalls.

Another aspect of the invention provides morphological recognition, that is a recognition based on tracing outlines and recognising shapes from such tracing. This essentially consists of identifying a shape that matches a reference shape contained in said reference data (database). This morphological recognition is particularly useful when applied to the shape of the hands and face of a person. Distinguishing the hands and the face from other objects significantly reduces false alarms that may arise from moving objects with a colour similar to that of skin (such as packaging cardboard, for example). Similarly, the shape of a person on the ground, which is obviously different from that of a person standing up, can be distinguished.

A further aspect consists of the dynamic analysis of the objects, which analyses the nature of the movement and enables a certain movement to be distinguished from another. Said analysis is used to distinguish an encoded action (which is always dynamic), from another one, but also to reduce the incidence of incorrect readings deriving from the aforementioned means.

The dynamic analysis substantially takes account of the movement of a recognised shape (hand, human body, etc..) on the basis of parameters such as direction (e.g. from top to bottom, or vice versa), speed, and path. Some examples are: when a person raises his or her hands, the "hand" shape is moving from bottom to top and not vice versa; when a person falls or lies down on the ground, the "human body" shape moves from top to bottom. The recognition of this dynamic events makes possible to detect a certain situation (e.g. panic, robbery, ...) and to launch alarm. The assessment of speed makes recognising the dynamics even more accurate, because the speed of a gesture can in fact be very different, according to the situation.

The dimensional parameter-setting of the objects is another useful function for avoiding false alarms. It essentially consists of checking how big recognized objects appear.

A standing person, for example, appears to have a greater dimension than a sitting person; a person raising his or her hands in the distance does not have the same dimensions as a person who makes the same gesture from nearby. For example, this function can be used in a supermarket for detecting what occurs near the till, where a robbery is more likely to occur.

A further aspect of the invention is recognising the proportions existing between shapes and/or objects: for example, if the method detects a body that is almost as big as a hand between two hands, said body is most probably a face, and most likely there is a person with his or her arms up (robbery). Preferably, the method also provides the function of recognising people (face recognition) through known algorithms adapted to this purpose. This function can be used for distinguishing who can place the system in alarm status. For example, if the camera 100 is installed in a shop, it can be provided for that the alarm will be given only when the camera recognises the owner of the shop who raises his hands (encoded gesture) to indicate a robbery in progress.

According to a further aspect of the invention, a dynamic analysis of the scene or situation observed by the camera can be performed, in order to discover dangerous behaviours. hi a particularly preferred embodiment, the camera 100 is made from a stainless-steel shell; the sensor means 1 comprise a CCD optic sensor together with a condenser microphone capsule; the processing means 2 consist of a Texas Instruments microprocessor model DM642 DSP; the memory means 3 are made of a RAM integrated circuit and a Flash memory of known type; the interface 4 comprise discrete components of known type and of a Fujitsu MB91101 microprocessor.

The camera 100 operates as follows: the image and audio of an event are captured through the sensor means 1 and converted into digital form. The microprocessor 2 processes these components and the audio/video and at the same time makes a comparison with the components prestored inside the memory 3.

When the processed and stored components are similar, the camera 100 enters into alarm status, sending a signal to the outside world through the interface 4.

The following two alarm situations are, for example, possible. A first situation occurs when two different conditions occur almost simultaneously (in quite a short period of time). This is the case in the non- limitative example of a manager of a petrol station who is cornered on the forecourt by two malefactors who wish to rob him, he raises his hands and almost simultaneously shouts out a key word that is incomprehensible to the malefactors but is significant to the camera 100. The information received by the sensor 1 is subjected to processing by the microprocessor 2. This processing concerns some parameters preset by the firmware and everything passes through a comparative analysis process of the parameters processed by the microprocessor 2 and those contained in the memory 3. The condition is said to be anomalous when the two parameters are similar or identical.

If two events are similar, the camera 100 enters alarm status (recognising an anomalous situation) following which, in addition to continuing its task of transmitting images, it also activates a process of reporting to external users or systems or devices.

The second situation occurs when a person gesticulates or changes posture continuously in discrete intervals of time following a prepackaged sequence. This is the case, for example, when a group of persons (for example an entire class of pupils) is taken hostage by malefactors. Through a series of sequential gestures, which are recognised by the microprocessor 2, the teacher has the chance to activate an alarm key and alert the authorities. The key that activates the alarm condition (anomalous condition) consists of a very precise and preset sequence of gestures/postures that were stored during the configuration phase inside the memory 3 of the camera 100. When the microprocessor 2 recognises that a sequence of gestures corresponds to a sequence of prestored gestures the camera 100 alarm condition is activated immediately. The latter, in addition to transmitting images and alarms to the outside world, also activates a special type of communication that is encoded by gestures. In fact, once the alarm key has been activated, each single gesture/posture that the teacher adopts will transmit to the external centre information corresponding to the meaning of that gesture.

The invention is clearly usable for activating processes/alarms of any type, both in the field of security and in other sectors. For example, the invention can be used to provide anti-robbery surveillance but also to monitor potentially dangerous workplaces, public places, and so on, and to report any anomalous or hazardous situation.

As stated above, the invention can be implemented through the described embedded camera or equivalent means, such as digital video recorders.

Claims

1) Surveillance method, for detecting or reporting hazard or alarm situations, comprising the following steps: a) acquiring or recording images or audio through sensor means; b) converting audio/video data from the above step a) into digital format; c) comparing digital data resulting from above step b) with pre-stored audio/video reference data, by means of a microprocessor (2) d) sending an alarm signal to the outside world via a suitable communication interface (4), when a match occurs between said data resulting from above step b) and said reference data.

2) Method according to claim 1, characterised in that step c) provides the recognition of colours.

3) Method according to claim 2, characterised in that step c) provides the recognition of the colour of the human skin.

4) Method according to anyone of preceding claims, characterised in that the step c) provides a morphological recognition, based on identifying shapes that appear similar to reference shapes contained in a database.

5) Method according to anyone of preceding claims, characterised in that the step c) provides a dynamic analysis for recognising the movement of a recognised shape. 6) Method according to claim 5, characterised in that said dynamic analysis provides the recognition of movement on the basis of direction, speed and course of said movement.

7) Method according to anyone of preceding claim, characterised in that step c) provides a dynamic analysis the scene or situation to discover dangerous behaviours

8) Method according to anyone of preceding claims, characterised in that the step c) provides a dimensional parameter-setting function that consists of checking the apparent dimensions of objects having a colour, morphology or dynamics recognised by the method.

9) Method according to anyone of preceding claims, characterised in that the step c) provides a recognition of the proportions existing between shapes and/or objects.

10) Method according to anyone of preceding claims, characterised in that the step c) provides a face recognition function, to recognize human beings.

11) Method according to anyone of preceding claims, characterised in that the communication interface (4) with the outside world uses standard types of protocol such as TCP/IP, UMTS, ADSL, ISDN, satellite.

12) Method according to anyone of preceding claims, characterised in that the alarm signal is activated by recognising a sound.

13) Method according to claim 12, characterised in that the alarm signal is activated by recognising a key word. 14) A surveillance device (100) suitable for reporting an emergency or alarm situation and adapted to operate with a method according to any one of claims 1 to 12.

15) A device according to claim 14, said device being a surveillance camera (100) comprising:

- sensor means (1) for acquiring image and audio data;

- processing means (2) for processing data from said sensor means (1);

- memory means (3) for storing the processing and operating procedures and for storing reference data;

- interface means (4) for connecting the camera (100) with the outside world.

16) A device according to claim 15, characterised in that: said sensor means (1) provides a CCD optical sensor together with a condenser microphone capsule; said processing means (2) are implemented with a DSP device; the memory means (3) comprise a RAM integrated circuit and Flash memory; the interface means (4) comprise discrete components of known type and a microprocessor.