KR20020013190A - Apparatus for preventing surreptitious use of image or video communication devices by image recognition - Google Patents

Abstract

PURPOSE: An apparatus for preventing the misappropriation of a video communication machine by image recognition is provided to prevent the misappropriation of other people by discriminating the lawfulness of a speaker through the image recognition of the speaker and informing the image data of the speaker in the case of unlawfulness. CONSTITUTION: An image data collecting unit(11) collects moving picture data of a speaker photographed through a camera. A preprocessing unit(12) adjusts the collected image data to be suitable for an object. A face region detecting unit(13) detects a face region in a preprocessed image data. A face feature extracting unit(14) extracts the face feature of the speaker in the detected face region. A face recognition and lawfulness judging unit(15) recognizes a face from the extracted face feature and judges the lawfulness of the recognized face. A communication control unit(16) controls the progress and stop of telephone conversation for preventing the misappropriation of an image communication machine of other people and controls communication for transmitting image data of an unlawful speaker.

Description

Apparatus for preventing surreptitious use of image or video communication devices by image recognition}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a locking device for a video communication device such as a video phone, and more particularly, to determine the legality of the speaker through the recognition of the speaker and to proceed or interrupt the call. The present invention relates to an apparatus for preventing theft of an imager by image recognition, which enables the user to track and report and to prevent theft and unauthorized use of the imager and to help the collection.

Recently, as the multimedia communication service is popularized, multimedia communication based on video communication, which can face and face conversations of far-flung opponents, is developing. Examples include video telephones and video conferencing systems. Video telephones exchange two images and voices through communication lines between two remote points, allowing people in remote locations to talk to each other while looking at each other's expressions as if they were in one place.

On the other hand, due to the rapid development of personal mobile communication, the number of mobile phone holders such as cell phones and PCS phones that can be called anytime and anywhere is increasing rapidly. Starting with IMT2000, these mobile phones have the ability to talk face to face with each other in earnest. Will be provided.

However, such a video communication device has a risk of theft or unauthorized use when lost or stolen.

Accordingly, an object of the present invention is to solve the above problems, to determine the legality of the speaker through the video recognition of the speaker to proceed or stop the call and to report the image data when it is determined to be illegal to prevent the theft of others The present invention provides an apparatus for preventing theft of an image communication apparatus using image recognition, which can help the recovery.

BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a block diagram showing an apparatus for preventing theft of an image communication apparatus by image recognition according to the present invention.

<Description of the symbols for the main parts of the drawings>

11: image data collecting unit 12: preprocessing unit

13: face area detection unit 14: face feature extraction unit

15: Face Recognition and Legality Determination 16: Communication Control Unit

17: storage unit

An apparatus for preventing theft of an image communicator by image recognition according to the present invention for achieving the above object is a face feature for extracting features of a speaker's face from input image data of a speaker in an apparatus for preventing theft of an image communicator. A face recognition and legal judgment for determining whether a legal speaker is recognized by recognizing a face from the facial feature extracted by the face feature extraction means, and when the face recognition and legal judgment are judged as legal speakers, the call proceeds. And a communication control unit for controlling to stop the call if it is determined to be an illegal speaker.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a block diagram illustrating an apparatus for preventing theft of an image communication apparatus based on image recognition according to an exemplary embodiment of the present invention, which illustrates a case where a face is a recognition target among various parts that can be recognized first. The anti-theft device shown in FIG. 1 is, first, an image data collecting unit 11 for collecting moving picture data of a speaker photographed through a camera or the like, and a preprocessing unit for adjusting the collected image data to suit a purpose ( 12), a face region detector 13 for detecting a face region from the preprocessed image data, a face feature extractor 14 for extracting a feature of the speaker face from the detected face region, and an extracted face Face recognition and legality determination (15) for recognizing a face from a feature and determining whether the recognized face is legitimate, and controlling the progress and interruption of a call to prevent the illegal use of a video communication device by another person according to the determination result. And a communication control unit 16 for controlling the transmission of the image data. The preprocessor 12 adjusts the image data to suit a specific purpose by noise reduction, histogram adjustment, and the like. The face area detection unit 13 first segmentes the image data for each subject by contour detection, labeling, structural analysis, and the like, and then extracts the features of each region from the region-divided image data for each subject and writes them. The face region is detected by comparing and analyzing the stored facial feature vectors. In general, in order to detect a face region, a feature vector generated by extracting and processing features such as color, shape, and texture by using mathematical, structural, and transformational methods for each subject region is used.

The face feature extractor 14 generates a face feature vector that can distinguish individual face differences from the detected face regions, and in the face recognition and legality determination unit 15, the legitimate caller pre-stored in the storage unit 17 ( And facial feature vectors. The facial feature vector here is different from that of the face region detection section 13 for distinguishing it from other classes of objects. The storage unit 17 also stores the image data of the illegal speaker for reporting. The facial feature vector for the facial region detector 13 is also stored in the storage 17. The operation of the apparatus for preventing the theft of an image communication apparatus by the image recognition of FIG. 1 having such a configuration will be described in detail.

In FIG. 1, the image data collecting unit 11 collects moving picture data of a speaker photographed through a camera and outputs the same to the preprocessing unit 12. The preprocessing unit 12 performs image preprocessing, such as improving the image quality or adjusting the image to suit a specific purpose, including adjusting the contrast, removing noise, smoothing, sharpening, and the like. Manipulation is included. This allows the preprocessor 12 to quickly and easily extract the feature of each subject from the image data. The face area detector 13 detects a face area by separating areas of each subject from preprocessed image data and analyzing colors, shapes, and textures of each area.

The facial feature extracting unit 14 is a technique for extracting unique features for each individual based on different facial features and the like for each person. For example, almost every aspect of a person is characterized by eye size or spacing, eye color, ear size or position, length, width, lip shape and size, chin, facial contour, and hairstyle. Among these features, identifying the most distinctive features of a particular person can make identification with others faster and easier.

In the feature extraction, the mathematical model may generate a feature vector by directly using numerical values obtained by measuring the intervals of the entire face or a specific part or a specific part, or by processing by statistical processing. You can also use values related to the size, area, and distribution of each part. Transformation techniques include Hotelling (or Eigen vector, principal component, discrete Karhunen-Loeve) transformation, Fourier transformation, DCT transformation, Walsh function transformation, wavelet transformation, etc., and are easily used for texture analysis. The neural network model mainly uses the unsupervised model for feature extraction. A representative model of unsupervised neural networks is Kohonen's Self-organized feature mapping (SOFM) network. This SOFM algorithm creates a vector quantizer by adjusting the connection weight from the input node to the output node. The output nodes are generally arranged in a two-dimensional lattice shape, and are connected locally between neurons during learning. Enlarge the connection. Doing so effectively transforms the high-dimensional input vector into a low-dimensional feature vector. That is, the output itself of the unsupervised artificial neural network can be used to form a feature vector. In the present embodiment, the facial feature extractor 14 may be implemented by a known technique as described above.

The facial feature vectors extracted by the facial feature extractor 14 are input to the face recognition and legality determination unit 15. The face recognition and legality determination unit 15 performs face recognition by comparing and analyzing face feature vectors, and an artificial neural network model is most preferred at present. The neural network used for recognition is mainly a supervised neural network, and its representative model is an artificial neural network using a backporpagation (BP) algorithm. The neural network is preferred for recognition because the input processing speed after learning is close to real time, and it is advantageous over other techniques for nonlinear problem analysis. In the case of using a mathematical model for face recognition, the correlation between the input pattern and the data stored for the legal speaker is mathematically calculated. In the case of using the AI model, the correlation is logically analyzed and recognized by a program based on knowledge. Since the methods used for face recognition are also known techniques, detailed descriptions thereof will be omitted here.

The face recognition and legality determination unit 15 also determines whether the recognized face is the legitimate speaker and outputs the determination result to the communication control unit 16. That is, the face recognition and legality determination unit 15 determines the legality by comparing and analyzing the similarity or correlation between the facial features of the legal speaker and the extracted speaker's facial features. Here, many may be allowed as legal speakers. The communication control unit 16 proceeds the call when it is determined to be a legitimate speaker and stops the call when it is determined to be an illegal speaker according to the face recognition and the determination result of the legality determination unit 15, thereby preventing theft. In addition, the communication control unit 16 stores the image data of the speaker in the storage unit 17 when the face recognition and legality determination unit 15 judges to be an illegal speaker, and automatically dials the designated number to the storage unit ( The video data of illegal speaker stored in 17) can be transmitted to help illegal speaker tracking and terminal collection. Here, the designated number may be a number previously designated by a telecommunication operator or a legal speaker.

Here, the facial feature extractor 14, the facial recognition and legality determination unit 15, and the storage unit 17 may be implemented to perform the function together by one or more artificial neural networks.

In this embodiment, the legitimacy of the speaker through face recognition has been described. However, when face recognition is difficult due to makeup or injury, the speaker is instructed to photograph fingerprints, irises, and the like. Even if the implementation of the possible embodiment is not described based on the present embodiment, it will be apparent to those skilled in the art.

As described above, the apparatus for preventing theft of an image communication apparatus according to the image recognition of the present invention identifies whether a legitimate speaker is identified through image recognition of a speaker input through a camera, and accordingly determines whether to allow a call, thereby preventing loss or theft. It is possible to prevent the illegal use and unauthorized use of other people's video communication device, and has an effect of helping to track and collect the terminal by automatically transmitting the image data to the designated number when illegal use.

Claims (11)

An apparatus for preventing theft of a video communication device, Face feature extraction means for extracting a feature of the speaker face from the input image data of the speaker; Face recognition and legality judging for judging whether a legitimate speaker is recognized by recognizing a face from the face feature extracted by the face feature extracting means; And And a communication control unit configured to control a call to proceed when the face recognition and the legality determination unit determine the legal speaker, and to stop the call when determined as the illegal speaker. 2. The apparatus of claim 1, further comprising a storage unit for storing an image data characteristic of an illegal speaker. 3. The apparatus of claim 2, wherein the communication control unit automatically transmits and reports the image data of the speaker stored in the storage unit to a designated number when it is determined to be an illegal speaker. According to claim 1, wherein the facial feature extraction means A preprocessor for preprocessing the input image data of the speaker; A face region detector for detecting a face region from preprocessed image data; And And a facial feature extracting unit for extracting a feature of the speaker's face in the detected face region. The image communication apparatus according to claim 4, wherein the face region detection unit divides image data into regions to extract features of each region, and compares and analyzes them with previously stored facial feature vectors to detect face regions. Anti-theft device. The apparatus of claim 4, wherein the facial feature extracting unit extracts facial features using mathematical, structural, and transformational means. The apparatus of claim 4, wherein the facial feature extracting unit extracts a feature using a non-supervised artificial neural network model, such as a SOFM network. 6. The apparatus of claim 4, wherein the face recognition and legality determination unit recognizes a face using a mathematical model, artificial intelligence, an artificial neural network, and the like. The apparatus of claim 8, wherein the face recognition and legality determination unit recognizes a face using a supervised artificial neural network model using a backpropagation algorithm or the like. 10. The method of claim 4, wherein the face feature extracting means receives images of all body parts that can be used for security, including a speaker's fingerprint and an iris, instead of face recognition, and extracts the feature. Video communication device theft prevention device. The apparatus of claim 1, wherein a plurality of the legal speakers are allowed.