JP2010056720A - Network camera, and network camera system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a network camera which stably tracks one and the same person even when an imaging direction of the person to be tracked considerably changes among different cameras. <P>SOLUTION: The network camera has: an imaging part for imaging an image of an imaging space; a face detection part for detecting a face area from an imaged image; a face direction correction part for correcting the direction of the detected face to compose it with a front face image; a characteristic extraction part for extracting characteristic data from the corrected face image; a receiving part for receiving the characteristic data transmitted by another network camera; a character data buffer for accumulating received characteristic data; a characteristic collation part for collating the characteristic data extracted by the characteristic extraction part with characteristic data accumulated in the characteristic data buffer; and a transmission part for transmitting the characteristic data to another network camera on the basis of the collation result in the characteristic collation part. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to an apparatus and system for tracking a person using a plurality of network cameras.

Conventionally, there has been known an apparatus for detecting and tracking a person by template matching or the like from an image taken by a camera. Also, Patent Document 1 and the like have proposed a technique for tracking a person using images taken by a plurality of cameras. In the invention described in Patent Document 1, when an intruder is detected from an image captured by the first camera, the detection target is registered as a template, and the next detected intruder image using the template is registered. Perform matching. This template is sequentially updated based on the latest detection data. The first camera transmits the registered template to the second camera. The second camera tracks the intruder detected by the first camera by matching the person detected from the image with the registered template.
Japanese Patent No. 3814797

  Since the change in the shape of the detection target is gentle in the image of one camera, it is possible to track a person by sequentially updating the template. However, when a template is transferred between cameras and tracking is performed by template matching, the shape of the object changes discontinuously with respect to the template shape due to the discontinuous movement of the object between the cameras. There may be cases. If the cameras are fixedly installed so that the imaging spaces of the respective cameras are completely continuous, the tracking operation between the cameras may be possible even with the method described in Patent Document 1. It is not realistic to provide a restriction, and the camera can generally change the imaging direction even after installation. Therefore, since the template matching described in Patent Document 1 cannot sufficiently cope with a change in the direction of imaging, it is difficult to perform stable tracking.

  In view of the above background, an object of the present invention is to provide a network camera that can stably track the same person even when the imaging direction of the person to be tracked changes greatly between different cameras. .

  The network camera of the present invention includes an imaging unit that captures an imaging space, a face detection unit that detects a face area from the captured image, and a face orientation correction that synthesizes a face image facing the front by correcting the orientation of the detected face. A feature extraction unit that extracts feature data from the corrected face image, a reception unit that receives feature data transmitted from another network camera, a feature data buffer that stores received feature data, and the feature A feature matching unit that matches the feature data extracted by the extraction unit with the feature data stored in the feature data buffer, and transmits the feature data to another network camera based on the matching result in the feature matching unit. And a transmission unit. Note that the network camera of the present invention is used in connection with another network camera having the same configuration as that described above.

  In this way, the face orientation correction unit corrects the face direction to the front direction, and the feature extraction unit extracts the feature data of the face image. Therefore, feature data suitable for matching is acquired regardless of the imaging direction. can do. Since the feature data obtained in this way is received from another network camera and used for face matching, it is possible to perform face matching appropriately based on the feature data of the face image acquired by the other network camera. The person can be tracked stably between a plurality of cameras. It should be noted that a pattern recognition method described in Japanese Patent Application No. 11-295058 can be used for feature extraction and collation of face images. Further, since the feature data of the face image is transmitted to another network camera based on the matching result in the feature matching unit, the detected face data of the person's face can be taken over by another network camera. The feature data transmitted to other network cameras may be data extracted by the feature extraction unit or feature data stored in the feature data buffer.

  A network camera system according to the present invention includes a plurality of the above-described network cameras, and uses the facial image feature data detected by one network camera to collate the facial image feature data detected by another network camera. Thus, a person detected by one network is tracked by another network camera.

  With this configuration, it is possible to appropriately track a person even when the imaging direction changes between a plurality of network cameras.

  The tracking method by the network camera system of the present invention is a method of tracking a person by a network camera system to which a plurality of network cameras are connected, wherein the first network camera images an imaging space, and the first Detecting a face area from an image captured by the network camera; correcting a face direction detected by the first network camera to synthesize a front face image; and correcting by the first network camera. Extracting feature data from a later face image; and feature data extracted from the face image that the first network camera has captured with the second network camera and the face orientation has been corrected; Receiving from the first network camera and the feature received by the first network camera Has a step of storing a over data in the feature data buffer, a configuration including a step of collating the first stored feature data and the extracted feature data in the feature data buffer in the network camera.

  With this configuration, similarly to the network camera of the present invention, it is possible to appropriately track a person even when the imaging direction changes between a plurality of network cameras.

  According to the present invention, the face orientation correction unit corrects the face direction to the front direction, and the feature extraction unit extracts the feature data of the face image. Therefore, the feature is suitable for matching regardless of the imaging direction. Data can be acquired, and feature data obtained in this way can be received from other network cameras and used for face matching, so that tracking of a person can be performed stably between multiple cameras. It has an excellent effect of being able to.

  Hereinafter, a network camera and a network camera system according to embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram showing a configuration of a network camera system 1 according to the present invention. The network camera system 1 is configured by connecting a plurality of network cameras 10, 10a, and 10b. In FIG. 1, for convenience of explanation, the network cameras 10, 10a, and 10b are distinguished from each other, but the network cameras 10, 10a, and 10b have the same configuration. In addition, although FIG. 1 illustrates an example having three network cameras 10, 10a, and 10b, the number of network cameras may be any number.

  The network camera 10 is configured by an imaging device such as a CCD, and the like. An imaging unit 12 that captures an imaging space to obtain an image, a face detection unit 14 that detects a face region from the image captured by the imaging unit 12, and a face A face orientation correcting unit 16 that corrects the orientation of the face detected by the detecting unit 14, a feature extracting unit 18 that extracts facial features from the face image corrected by the face orientation correcting unit 16, and another network camera 10a. Received from the facial feature data and the receiving unit 22, the feature data buffer 24 storing the feature data received by the receiving unit 22, the feature data extracted by the feature extracting unit 18 and the feature data buffer 24 A feature matching unit 20 that matches feature data, and a transmission unit 26 that transmits the feature data to another network camera 10b according to the matching result in the feature matching unit 20 are provided.

  FIG. 2 is a diagram illustrating the operation of the network camera 10 having the above-described configuration. The operation of the network camera 10 will be described, and the functions of the above components will be described in detail.

  The imaging unit 12 acquires the captured image as digital data (S10). The face detection unit 14 detects a human face region from the digital image data captured by the imaging unit 12 (S12).

  The face orientation correction unit 16 estimates the orientation of the face image detected by the face detection unit 14 and synthesizes the face image facing the front (S14). For example, “J.Xiao, S. Baker, I. Matthews, and T. Kanade,“ Real-time combined 2D + 3D active appearance models ”, InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition, volume II, pp. 535-542, 2004. ”can be applied. With this configuration, it is possible to solve the problems that the face image collation accuracy is reduced in a face orientation other than the front face, and the imaging direction cannot be specified by another camera.

  The feature extraction unit 18 extracts feature data for performing face matching from the face image corrected in the front direction by the face direction correction unit 16 (S16). The feature data is data indicating the position and shape of facial parts such as eyes, nose, mouth, and eyebrows. By using such feature data, there is an effect that it is difficult to be affected by the difference in illumination conditions.

  On the other hand, the other network camera 10a transmits facial feature data and incidental data such as camera data and tracking data. Here, the “tracking data” is history data in which a certain person is tracked among a plurality of network cameras, and includes identification numbers of the respective network cameras, data on the positions of detected persons in images, time data, and the like. In addition to “feature data”, “accompanying data” is data passed between network cameras for person tracking. In addition to “tracking data”, text data indicating the name of each camera, Including various camera parameters, data necessary for the final application can be attached. The receiving unit 22 receives the facial feature data and the incidental data transmitted from the other network camera 10a together (S18). The receiving unit 22 temporarily stores the feature data and incidental data received from the other network camera 10 a in the feature data buffer 24.

  The feature collation unit 20 collates the feature data extracted by the feature extraction unit 18 with all the feature data stored in the feature data buffer 24 (S20). The feature matching unit 20 uses the face detection unit 14 to detect the data in the feature data buffer 24 that is equal to or higher than a preset threshold value and has the highest similarity with respect to the feature data extracted by the feature extraction unit 18. It is determined that the data of the same person as the selected person.

  When there is data of the same person, the data related to the network camera 10 is added to the incidental data stored in the feature data buffer 24 as tracking data, and the own data is added to the tracking data, and together with the feature data, Transmission is performed from the transmission unit 26 to another network camera 10b (S22).

  The feature data transmitted at this time may be the feature data extracted by the feature extraction unit 18, or the face orientation correction is small among the data accumulated in the feature data buffer 24 and the feature data extracted by the feature extraction unit 18. In other words, the image may be taken in a direction close to the front, or may be comprehensively determined to be suitable for collation in consideration of information such as the size of the imaged face. Further, instead of sending either one of the feature data, a feature data group may be sent with a plurality of feature data determined to be the same person.

  In the above description, the operation after the person is detected by one network camera 10, particularly the operation in the case of taking over the tracking operation among the plurality of network cameras 10 has been mainly described. The operation when the network camera 10 first detects a person will be described. When the feature data of the detected face image does not match any feature data stored in the feature data buffer 24, the network camera 10 determines that the face image is a new person, and the face image is Store in the feature data buffer 24. An operation when feature data is not received from another network camera 10 will also be described. Also in this case, the network camera 10 performs tracking within one network camera 10 by collating the detected feature data of the face image with the feature data stored in the feature data buffer 24. The network camera 10 and the network camera system 1 according to the present embodiment have been described above.

  In the network camera 10 according to the present embodiment, the face image detected by the face detection unit 14 is corrected in the front direction by the face direction correction unit 16, and then the facial feature is extracted by the feature extraction unit 18. Accordingly, each network camera 10 can always perform face matching processing based on the feature data of the face image facing forward regardless of the imaging direction of the person to be tracked. Can be tracked at.

  The network camera system and the network camera according to the embodiment of the present invention have been described in detail with reference to the embodiment. However, the present invention is not limited to the above-described embodiment.

  In the above-described embodiment, the destination of the feature data from the transmission unit 26 is the other network camera 10b. However, depending on the system configuration, other than cameras such as a server device, a data storage device, and a data display device may be used. The device may be the transmission destination.

  In the above-described embodiment, the example in which the feature data is received from the network camera 10a and the feature data is transmitted to the network camera 10b has been described. However, the destination of the feature data may be the network camera 10a. Thereby, even when the tracking target returns to the imaging space of the network camera 10a, the tracking can be performed.

  As described above, the present invention has an excellent effect of stably tracking a person between a plurality of cameras, and is useful as a system for tracking a person using a plurality of network cameras. It is.

The figure which shows the structure of the network camera system in embodiment The figure which shows operation | movement of the network camera in embodiment

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Network camera system 10, 10a, 10b Network camera 12 Imaging part 14 Face detection part 16 Face direction correction part 18 Feature extraction part 20 Feature collation part 22 Reception part 24 Feature data buffer 26 Transmission part

Claims (3)

An imaging unit for imaging the imaging space;
A face detection unit for detecting a face area from the captured image;
A face orientation correction unit that corrects the detected face orientation and synthesizes a front face image;
A feature extraction unit that extracts feature data from the corrected face image;
A receiving unit for receiving feature data transmitted from another network camera;
A feature data buffer for storing received feature data;
A feature matching unit for matching feature data extracted by the feature extraction unit with feature data stored in the feature data buffer;
Based on the collation result in the feature collation unit, a transmission unit that transmits feature data to another network camera;
Network camera equipped with. A plurality of network cameras according to claim 1;
Using the face image feature data detected by one network camera, the face image feature data detected by another network camera is collated, so that the person detected by one network can be transferred to another network camera. Tracking network camera system. A method of tracking a person by a network camera system to which a plurality of network cameras are connected,
A first network camera imaging the imaging space;
Detecting a face region from an image captured by the first network camera;
Correcting the face direction detected by the first network camera to synthesize a front-facing face image;
The first network camera extracting feature data from the corrected face image;
The first network camera receiving from the second network camera feature data extracted from a face image captured by the second network camera and corrected in face orientation;
Storing the received feature data in a feature data buffer in the first network camera;
Collating the feature data extracted by the first network camera with the feature data stored in the feature data buffer;
Tracking method with network camera system equipped with

Images