KR20070118806A - Method of detecting face for embedded system - Google Patents

Abstract

A method for detecting a face in an embedded system is provided to require no high specification to surrounding systems while providing a high face recognition speed by providing an SoC for detecting and recognizing the face. An input buffer receives image data of a user and previously stores face feature of the registered users. An image processor controls lighting and processes the image data such as reduction/enlargement, rotation, and color conversion. A face detection controller stores a feature filter for discriminating a face part from the image data by applying a mask to the image data so as to find the face from the image data with a detection filter. A face recognition controller comprises an extractor extracting the feature needed for recognizing the face and a recognizer recognizing the face with the extracted feature. An output buffer stores face position and recognition information, and the feature extracted from the face.

Description

Facial detection method for embedded system {METHOD OF DETECTING FACE FOR EMBEDDED SYSTEM}

1 is a block diagram showing a schematic configuration according to a face detection method for an embedded system.

The present invention relates to a face detection method for an embedded system.

Traditionally, in order to recognize the identity of a specific user for security, video conferencing, and unattended surveillance in mobile, videophone, and intelligent robot fields, a password input system or a smart card is frequently used. However, this method has a disadvantage that a user may forget a password or a smart card, and a malicious user may predict or use a password or steal a smart card.

The biometric technology actively used to overcome these problems is a technology for recognizing the unique characteristics inherent in the user itself, such as iris, fingerprint, palm, face, signature, behavior, and voice. It refers to the technology of recognizing the physical or behavioral characteristics of the user himself or not, or not having it. This unique feature of the user is not forgotten like a password or lost like a smart card, and is difficult to steal and be used by other users.

1 is an explanatory diagram schematically showing a field of use of a face recognition system.

Among the various features of biometrics, the facial feature provides a more familiar and comfortable method using only the input image of the camera without the user's direct input, compared to other features requiring direct input using a specific device. As shown in 1, it is being utilized and applied in various fields.

Algorithms or systems for face recognition have been steadily developed based on various technologies and theories in order to be reliably applied to various changes occurring in the user's external environment or face changes such as lighting, poses, facial expressions, jewelry, and age. Currently developed face recognition methods have strong advantages of "two-dimensional recognition" (general face recognition data) and "three-dimensional recognition" (face pose change) depending on the face data input method, but the recognition rate due to lack of training / test data It can be classified into 'degradation problem', and according to the characteristics of the data, the 'heat pattern image' of the face using the infrared camera (which is robust to light changes, but the application area is limited) and 'color' (modeled by the skin color of the face, Although the processing speed is fast, it can be divided into lighting, race skin color variation and sensitivity) and 'black and white' data (general face recognition data).

The known face recognition system receives a face image from a peripheral device such as a scanner or a camera capable of reading a face image, detects a face portion by analyzing image information, and then normalizes the detected face image to characterize the face portion. It is configured to finally recognize whether the face is obtained by comparing the facial data registered by the specific user with the result.

The face recognition solution / system implemented so far is a Pentium-based 'software'. In order to apply such a software system, a high specification system must be built. These software-based systems are very limited in use, less convenient, and less expensive.

Summarizing this conventional problem,

1) Existing face detection, recognition, and feature information extraction methods are designed to require a lot of resources and calculations. Therefore, the implementation of actual chips requires a lot of light weight and chip application, and also the performance There was a problem that the actual chip implementation was difficult because of the disadvantage that the degradation is accompanied,

2) In the existing face-related algorithms, in order to achieve high recognition performance, additional normalization, a lot of lighting, poses, etc. are required. Costly to design, implement,

3) By using a method such as Neural Network, which is a complex pattern recognition, to distinguish face and non-face, detection performance is high, but the computation required for calculation is exponentially enlarged, making it difficult for real-time processing and necessary for computation. The disadvantage was that the size of the data was large,

4) In addition, a method of detecting a face has been developed by extracting a skin color distribution from a color distribution of an input image by modeling a skin color of a face in a color image for real-time processing. There was a problem that it was difficult to obtain the level of detection performance.

Therefore, it is possible to perform excellent recognition processing by minimizing errors through more various distance calculation analysis, and propose memory and hardware structure for detecting and recognizing face in real time, and this is based on FPGA and system on chip (SoC). The hardware structure designed and implemented can be applied as the core of embedded system, which leads to the necessity of a new system-based detection method that can show superior speed than the software processing method.

The present invention has been made in order to overcome the above problems, it is operated by the existing software method to overcome the need for high performance of the system to implement the face detection and recognition processing in 'SoC' has excellent recognition processing speed and peripheral Its purpose is to provide a system based facial detection method that does not require high specification of the system.

In order to achieve the above object, the embedded system for face recognition according to the present invention is characterized by comprising an input buffer, an output buffer, an image processing unit, a face detection control unit, a face recognition control unit.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. The accompanying drawings are not drawn to scale, and like reference numerals in each of the drawings refer to like elements.

The facial recognition chip for an embedded system according to the present invention receives an image, detects and recognizes a face, and provides facial position, size information, eye position, rotation information, registered person or member information in the image. Do this.

Referring to FIG. 1, a face recognition chip for implementing an embedded system according to the present invention includes an input, an output buffer, an image processor, a face detection controller, and a face recognition controller. The input buffer accepts image information from cameras, videos, and images and facial feature information of people registered in advance, and the image processing unit performs basic image processing such as basic image size, rotation, and color conversion for face detection and recognition. Consists of a function for light processing to perform the operation in a general environment. The face detection control unit is configured to store a feature filter for distinguishing a face from a non-facial image, and to use a detection filter to detect a face in an image and to determine whether a face is present. Finally, the face recognition control unit is composed of a part for extracting the feature information required for recognition and a part for performing the recognition using the same, and finally, the output buffer of the face position information, recognition information, extracted facial feature information Has a result.

  The function of each component is as follows.

A) Input buffer

(1) Receive user's input image such as camera or image, video and facial feature information data of previously registered people, and perform intermediate step through image processing for speed and operation optimization through intermediate processing

(2) Storage and processing of input image and integral image

(3) Save registered facial feature information

B) Image processing

(1) It is in charge of basic image processing, filtering, lighting processing to perform face detection, recognition

(2) Resize image

(3) image rotation

(4) mask filtering

(5) illumination correction

Face detection

(1) Detect the face from the input image information entered into the input buffer, and verify that the detected image is the actual face

(2) Face detection: Built-in filter that can distinguish face and non-facial image irrespective of various environment, lighting, and people's change, and integrates it into the chip to process the position of face Function to get size information

(3) face verification: function to verify the actual frontal face by constantly normalizing the position and size of the obtained face and obtaining facial features such as eyes, nose and mouth

D) Face recognition

(1) Facial feature information extraction (face extraction): the function to correct facial illumination, pose, facial expression changes, and extract facial feature information for recognition

(2) Face identification: Using the facial recognition feature information of the pre-stored people as input, the personal identification information is obtained by comparing the relationship with the facial feature information extracted from the currently detected face and detected in the image. The ability to find out who a person is

E) Output buffer

(1) face location: A part that stores face position and size information obtained as a result of face detection

(2) face ID information: a portion that stores facial recognition feature information extracted from the person who is currently detected and recognized

As described so far, the configuration and operation of the facial detection method for an embedded system according to the present invention have been expressed in the above description and the drawings, which are merely examples, and the spirit of the present invention is not limited to the above description and the drawings. Of course, various changes and modifications are possible without departing from the spirit of the present invention.

As described above, the facial detection method for an embedded system according to the present invention has an effect of speeding up processing and performing more accurate face recognition by embedding the system in a soc state for face recognition.

Claims (1)

Facial detection method for embedded system, And a step of implementing an input buffer, a step of implementing an output buffer and an image processing unit, a step of implementing a face detection control unit, and a step of implementing face detection.

Images