JPH11283036A - Object detector and object detection method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve a detection rate by considering individual differences of the feature area position of an object to be identified, fluctuations of a size and of a photographing environment and detecting the object and to perform a simple processing by eliminating need of verification of the position relation of a feature area. SOLUTION: Images including a face are inputted from an image input part 1 and stored as gradation images, and the gradation images are differentiated by an image processing part 2 so that edge images are generated and stored. Then, in a position detection part 3, and area model for which plural judgement element obtaining areas are set corresponding to the characteristic areas such as the eye, the nose and the mouth, etc., of face images in used for the gradation images and the edge images, the area model is fitted to the image while successively specifying the position, a feature amount is extracted from the respective judgement element obtaining areas and the face image is detected from the images based on the extracted feature amounted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an object detecting apparatus and an object detecting method for detecting an object to be identified, such as a face or an article.

[0002]

2. Description of the Related Art As an apparatus for detecting an object to be identified from an image including a certain object to be identified and a background, for example, Japanese Patent Application Laid-Open No. 7-129770 is known. For example, in an assembling process or an inspection process in a factory, an object to be identified is photographed by a television camera, and a correlation is made between a grayscale image of the photographed input image and a reference template image, thereby obtaining an image of the object to be identified. The position of an object is detected. Specifically, a local region having the same size as the template image is set in the input image as a search image, and the correlation value between the template image and the search image is calculated while shifting the local region in order, and the position where the correlation value is the highest is calculated. Is specified as the position where the template exists.

As a technique for detecting a face image from an input image by using a template matching method, for example, Japanese Patent Application Laid-Open Nos. 9-251534 and 9-446.
No. 76 is known. In Japanese Patent Application Laid-Open No. 9-251534, in order to extract a face image region from an input image, a correlation value is calculated while moving a pre-registered standard face image (template) over the entire screen, and the highest correlation value is obtained. A region having a value is extracted as a face region. Japanese Patent Application Laid-Open No. 9-44676 scans an image including a face image with a template image in which eyes are represented by density information, performs a correlation operation between the density information of the target area and the density information of the template image, An area having a high similarity is extracted as an eye candidate. Similarly, a nose candidate is extracted by scanning an image including a face image with a template image representing the nose in shade information, and an image including a face image is scanned in a template image representing the mouth in the shade information. Extract candidates. And
When the extraction is completed, the face area is extracted.
The face image is extracted from the image by comparing and verifying the combination of the eye, nose, and mouth candidates with the positional relationship of the prepared eyes, nose, and mouth.

Further, as an example utilizing color information of a face image, Japanese Patent Application Laid-Open No. 9-50528 is known. This involves extracting a skin color region from the RGB values of an input image, automatically determining a mosaic size for this region, converting the candidate region into a mosaic, comparing the candidate region with a person face dictionary to determine the presence or absence of a person face. This is to cut out a human face.

[0005]

By the way, faces vary in position and size of nose and eyes for each individual, and the application location of the face detection of a person is different from that of a factory automation device in a factory. Is complicated and the external light environment fluctuates greatly in many cases. In addition, since a person moves in front of the camera, the face size changes due to a slight shift in the front-back position of the person, and the inclination of the face also varies.

In view of the above, Japanese Patent Application Laid-Open No.
No. 70 and Japanese Patent Application Laid-Open No. 9-251534 use a template matching method, in which the face of a person is used as it is as a template image of the entire face. Was not so high. That is,
Since individual differences and the like are not taken into account, a situation occurs in which a certain person can be detected but another person cannot.

In the method of extracting individual parts as disclosed in Japanese Patent Application Laid-Open No. 9-44676, the procedure is complicated, such as changing the extraction method for each individual part and verifying the positional relationship between candidate regions. However, there is a problem that stable detection is difficult, and in particular, when the background is complicated, there is a possibility that the number of component candidates may become large, and there is a problem that it becomes difficult to verify the positional relationship. . Further, Japanese Patent Application Laid-Open
The method of narrowing down the candidate areas in advance by using color information for extracting a skin color area as in Japanese Patent No. -50528 has a problem that it is easily affected by lighting conditions and stable detection becomes difficult.

Therefore, the inventions according to claims 1 to 7 are based on individual differences in the position of the characteristic region of the object to be identified, fluctuations in the size,
Object detection that can detect an object to be identified in consideration of fluctuations in the shooting environment, improve the detection rate, and realizes simple processing without the need to verify the positional relationship between characteristic regions. Provide equipment. Claims 8 to 1
According to the invention described in (1), the object to be identified can be detected in consideration of individual differences in the position of the characteristic region of the object to be identified, variation in size, and variation in the imaging environment, and the detection rate can be improved. Provided is a method for detecting an object that can perform simple processing without having to verify the positional relationship between characteristic regions.

[0009]

According to the first aspect of the present invention,
Image input means for inputting an image, storage means for storing an area model in which a plurality of determination element acquisition areas are set corresponding to characteristic areas of the image of the object to be detected, and input input by the image input means A position designating means for sequentially designating a position where the area model stored in the storage means is applied to one or both of an image and an image obtained by performing image processing on the input image; and a position designated by the position designating means. Each time the region model is sequentially applied, a determination element acquisition unit that acquires a determination element from each determination element acquisition region of the region model, and whether or not the image is a target object image based on the determination element acquired by the determination element acquisition unit And a determination unit for determining whether the target is an object to be identified based on the determination result of the determination unit.

According to a second aspect of the present invention, there is provided an image input means for inputting an image, and a plurality of determination element acquisition areas corresponding to characteristic areas such as eyes, mouth, nose, and cheeks of a face image to be detected. Means for storing the area model stored in the storage means, and a position at which the area model stored in the storage means is applied to one or both of an input image input by the image input means and an image obtained by performing image processing on the input image. A position specifying means for specifying, a determination element obtaining means for obtaining a determination element from each determination element obtaining area of the area model each time the area model is sequentially applied to the position specified by the position specification means, The object detection device includes a determination unit that determines whether or not the image is a face image based on the determination element acquired by the acquisition unit, and detects a face based on the determination result of the determination unit.

According to a third aspect of the present invention, in the object detecting device according to the first or second aspect, the determination element obtaining means obtains a feature amount as a determination element from each determination element obtaining area of the area model. . According to a fourth aspect of the present invention, in the object detection apparatus according to the third aspect, the determination element obtaining means obtains a feature amount using a template image for at least one determination element obtaining area of the area model. It is in. According to a fifth aspect of the present invention, in the object detection device according to the third or fourth aspect, the determination element obtaining means includes one of an eye, a mouth, and a nose as a determination element obtaining area of the area model.
Alternatively, when a plurality of regions are set, a feature amount is obtained for at least one of the regions by using a grayscale value or a differential value of the grayscale value in the region.

According to a sixth aspect of the present invention, in the object detecting device according to the third or fifth aspect, when the area of the cheek is set as the determination element acquisition area of the area model, On the other hand, a feature amount is obtained by using a variance value of gray values in a region. According to a seventh aspect of the present invention, in the object detection device according to the third or fifth aspect, when the cheek area is set as the determination element acquisition area of the area model, the determination element acquiring unit determines a template for the area. It is to acquire a feature amount using an image.

The invention according to claim 8 corresponds to a characteristic region of an image of an object to be detected with respect to one or both of an input image and an image obtained by performing image processing on the input image. A plurality of determination element acquisition areas are set by sequentially fitting the area model while designating the position, and each time the area model is applied, a determination element is obtained from each determination element acquisition area of the area model, and the obtained determination element is obtained. In the object detection method, a determination is made as to whether or not the image is an object to be identified based on the determination result, and the object to be identified is detected based on the determination result.

According to a ninth aspect of the present invention, the input image and one or both of the images obtained by performing image processing on the input image are characterized by the characteristic face, mouth, nose, etc. of the detected face image. A region model in which a plurality of determination element acquisition regions are set corresponding to the region is applied while sequentially specifying the position, and each time the region model is applied, a determination element is obtained from each determination element acquisition region of the region model. An object detection method is to determine whether or not the image is a face image based on the obtained determination element and to detect a face based on the determination result.

The invention according to claim 10 is the invention according to claim 8 or 9.
In the described object detection method, a feature amount is obtained as a determination element from each determination element acquisition area of the area model. According to an eleventh aspect of the present invention, in the object detecting method according to the tenth aspect, a feature amount is acquired using a template image for at least one determination element acquisition region of the region model.

[0016]

Embodiments of the present invention will be described with reference to the drawings. (First Embodiment) This embodiment describes a case where face detection is used as an example of detection of an object to be identified.

FIG. 1 is a block diagram showing the overall configuration of an object detection apparatus. An image input unit 1 serving as an image input unit for inputting an image including a human face. An image processing unit 2 for performing a predetermined process, image information input by the image input unit 1 and the image processing unit 2
And a position detector 3 for detecting the position of the face image from the processed image information.

The image input unit 1, as shown in FIG.
A CCD camera 11 for photographing the face of a person and outputting digital gray image information including the face, and an image input board 12 for receiving digital gray image information from the CCD camera 11
And an image memory 13 for storing digital grayscale image information captured by the image input board 12. The input image may be a color image.

The image processing unit 2 includes the image memory 13
, Digital gray image information is read from the image data, differentiated and converted into edge image information, and stored in the image memory 13. That is, the grayscale image shown in FIG. 3A is differentiated to generate an edge image shown in FIG. 3B and stored in the image memory 13. For the edge extraction, a well-known Laplacian filter or Sobel filter is used.

Since the eyes, nose, mouth, etc. have considerable edges, it is effective to use them as feature values.
When using a feature amount other than an edge, another image processing means may be added. When only the image gray level is used as the feature value, the image processing unit may not be provided. Further, the image processing may not be performed on the entire image but may be performed on each local region to be compared.

As shown in FIG. 4, the position detecting section 3
A model generation unit 31 that generates a region model of the face, a model storage unit 32 that stores the region model generated by the model generation unit 31, and a region model stored in the model storage unit 32 is stored in the image memory 13. A position specifying unit 33 that specifies a position to be applied to the digital grayscale image and the edge image; and a determination element acquisition unit that extracts feature amounts of a plurality of determination element acquisition regions in the region model at the position specified by the position specification unit 33. And determining whether or not the image is a face image using the determination information stored in the determination storage unit 35 from the characteristic amount of each determination element acquisition area extracted by the characteristic amount extraction unit 34 And a determination unit 36 as determination means for performing the determination.

The model generating section 31 generates an area model in which a plurality of determination element acquisition areas are set corresponding to characteristic areas such as eyes, nose, mouth, and cheeks of a face image as shown in FIG. It has become. In the region model shown in FIG. 5A, the characteristic regions are the eye, nose, mouth, and cheek regions, and the judgment element acquisition regions 41 and 42 for the eyes, the judgment element acquisition region 43 for the nose, and the judgment element acquisition for the mouth are obtained. An area 44 and a determination element acquisition area 45 for the remaining cheeks are set. In the region model shown in FIG. 5B, the characteristic regions are the eye, mouth, and cheek regions, and the judgment element acquisition regions 41 and 42 for the eyes, the judgment element acquisition region 44 for the mouth, and the judgment element acquisition region 45 for the cheek. Is set.

In the region model shown in FIG. 5C, a characteristic region is defined as an eye or cheek region, and a determination element acquisition region 4 for the eye.
1, 42, and a determination element acquisition area 45 for the cheek is set. In the region model shown in FIG. 5D, the characteristic regions are the eye, nose, and cheek regions, and the judgment element acquisition regions 41 and 42 for the eyes, the judgment element acquisition region 43 for the nose, and the judgment element acquisition region 45 for the cheek. Is set.

In the region model shown in FIG. 5E, a characteristic region is defined as an eye, a nose, a mouth, and a cheek region, a judgment element acquisition region 46 in which the left and right eyes are set to one, and a judgment element acquisition for the nose. An area 43, a determination element acquisition area 44 for the mouth, and a determination element acquisition area 45 for the cheek are set. (F) of FIG.
In the region model, the characteristic regions are the regions of eyes, nose, mouth, cheeks, and hair, and the judgment element acquisition regions 41 and 42 for the eyes, the judgment element acquisition region 43 for the nose, the judgment element acquisition region 44 for the mouth, and the cheek And a determination element acquisition area 47 for hair are set. These region models are called window models.

The area model shown in FIG. 6 is called a mask model, and each area is represented by a mask. Numerical values 1 and 2 are determination element obtaining regions for the eyes, numerical value 3 is a determining element obtaining region for the nose, numerical value 4 is a determining element obtaining region for the mouth, and numerical value 5 is a determining element obtaining region for the cheek. In addition, numerical value 0
Is a region to be excluded. With such a mask expression, a more detailed region model can be generated.

Regardless of which region model is used, a single model can absorb a certain degree of difference in face size and individual differences in the position of face parts, but in order to cope with a wider variety of face sizes. May use region models of different sizes. If a necessary area model is created in advance and stored in the model storage unit 32, the model generation unit 3
1 can be omitted.

The model storage section 32 stores a face area model as shown in FIG. In the case of the window model, the coordinates, width, height, and the like of each window representing each determination element acquisition area are stored with the upper left of the window as the origin.
In the case of a mask model, the mask value shown in FIG. 6 is stored as it is. There are various other methods for storing the model, and the method is not limited to these.

The area model stored in the model storage unit 32 is applied to the digital grayscale image from the image input unit 1 and the edge image processed by the image processing unit 2 to detect a face image. By applying such an area model M, it becomes possible to detect the faces of various persons as shown in FIG. 7, a person with a slightly inclined face, and a person wearing glasses. It should be noted that it is possible to specify each area in the area model as a parameter without taking the form of storing a face model in a memory on a computer program. Regard it.

The position specifying section 33 and the feature quantity extracting section 34
Designates a position where the area model M is applied to both the digital grayscale image and the edge image stored in the image memory 13 as shown in FIG.
The feature amount of each determination element acquisition region in the region model M is extracted from the two images. That is, the position specifying unit 33 basically specifies the position of the region model sequentially for the entire two images, and the feature amount extracting unit 34 obtains each determination element in the region model M at each of the specified positions. The feature of the region is extracted. Then, when the area model M is specified at the position of M ', the most face-like feature amount is extracted.

When using other feature values, the image processing unit 2 generates an image including other feature values from the input image, stores the image in the image memory 13, and uses the stored image. Here, when the input image is a color image, an area that is likely to be skin color may be extracted from this color input image, and the position of the area model may be specified only for that area.

As described above, it is also possible to specify the position of the area model after narrowing down the detection area by the preprocessing. Also,
It is also possible to use the feature of skin color using a color image. Here, as a method of fitting the region model, the entire region model is sequentially fitted to the input image while moving the designated position. However, the present invention is not limited to this. The fitting may be performed sequentially while moving the designated position one by one.

At the position designated by the position designating section 33, as a feature quantity to be extracted from both the digital grayscale image and the edge image, for example, a feature quantity expressed by the following equation is used. The area model of the face to be used is shown in FIG.
The model shown in Further, the determination element acquisition area 41 of the right eye in the area model is RE, and the determination element acquisition area 42 of the left eye is
Is LE, the nose determination element acquisition area 43 is N, the mouth determination element acquisition area 44 is M, and the remaining cheek determination element acquisition area 45 is C. In addition, the digital grayscale image and the edge image in the image memory 13 are both images having values of pixels 0 to 255. Further, P (i) is a gray value of the digital gray image at the position i, and E (i) is a gray value of the edge image at the position i.

The sum of the low-luminance features of the right-eye determination element acquisition area RE is

(Equation 1)

The sum of the edge feature amounts of the determination element acquisition area RE of the right eye is

(Equation 2)

The sum of the low-luminance features of the left-eye determination element acquisition area LE is

(Equation 3)

The sum of the edge feature amounts of the left-eye determination element acquisition area LE is

(Equation 4)

The sum of the low-luminance features of the nose determination element acquisition area N is

(Equation 5)

The sum of the edge feature amounts of the nose determination element acquisition region N is

(Equation 6)

The sum of the low luminance feature amounts of the mouth determination element acquisition area M is

(Equation 7)

The sum of the edge feature amounts of the mouth determination element acquisition area M is

(Equation 8)

The average brightness value of the cheek determination element acquisition area C is

(Equation 9)

The variance of the luminance value of the determination element acquisition area C of the cheek is

(Equation 10)

Is as follows. In addition, regarding the luminance value variance of the determination element acquisition region C of the cheek, the remaining cheek region C excluding the eyes, the nose, and the mouth has a characteristic that the luminance variance is small.

For the purpose of determination, each feature value represented by each of the above equations is extracted and stored in a memory. FIG. 9 shows an example of the feature amounts extracted in this way. The feature amount of each determination element acquisition area varies depending on the person. That is, there are individual differences. This data also includes a difference in the shooting environment, a slight difference in the face size, a change in the inclination of the face, and the like. The determination unit 36 determines whether the position of the area model specified by the determination unit 36 is a face based on the feature amount of each area of the area model specified by the position specification unit 33 and extracted by the feature amount extraction unit 34.

For the determination, for example, a fuzzy membership function is used. For example, a fuzzy membership function is defined as shown in FIG. 10 by setting μ _RE-d as the correctness of the low-luminance pixel amount of the right eye. This function obtains the determination element acquisition area RE for the right eye in the area model at a specified position.
Input the low-luminance pixel amount RE-d of
The value of ~ 1.0 is output.

In FIG. 10, RE-d is 1,200 to 10,000.
If measured as 0, μ _RE-d outputs 1.0. This range is determined with reference to individual differences and variations in feature amounts obtained from a sample image as shown in FIG. For values in other ranges, the output of a linear function forming the left and right sides of the trapezoid is the value of μ _RE-d . The closer μ _RE-d is to 1.0, the more accurate the right-eye low-luminance pixel amount is.

The trapezoidal membership function is determined by
As shown in FIG. 9, the feature amount of the face is measured in advance from various learning images, and is determined based on the measured feature amount.
Similarly, RE-e, LE-d, LE-e, Nd, Ne are calculated from the feature amounts of the sample images of the respective determination element acquisition areas as shown in FIG.
, Md, Me, Ca, and Cv are defined as fuzzy membership functions, and their outputs μ _RE-e , μ _LE-d ,
μ _LE-e , μ _Nd , μ _Ne , μ _Md , μ _Me , μ _Ca , μ
Calculate _Cv .

The respective membership functions may be stored in the storage unit 35 for determination. Then, it is finally determined whether or not the position is a face. As the faciality F, for example,

[Equation 11]

The following function is used. In this case,
The closer the output of each membership function is to 1.0, the more F
Is close to 1.0. However, in the case of this formula, all are 1.
If it is not 0, F does not become 1.0.

Various methods can be considered for the formula of F.
It can be obtained by using fuzzy rules or by algebraic sum. If the part having the largest value of F is set as the position of the face and a circle is drawn at that position on the input image, an image as shown in FIG. 11 is obtained, for example. For the above determination, it is not always necessary to use fuzzy logic. In addition, in the case of performing more reliable detection, it is further verified whether the face is a true face at the detected position.

As described above, the face to be identified is
Create an area model of the face divided into characteristic determination element acquisition areas such as eyes, nose, mouth, cheek, etc., and apply this area model maintaining the positional relationship of each determination element acquisition area to an input image and an input image. Since the position to be applied to both of the image-processed images is specified, the feature amount of each determination element acquisition region in the region model at that position is measured, and it is determined whether or not the face is the entire region model. There is no need to verify the positional relationship between parts such as the eyes and nose of the face, and simple processing can be realized.
In addition, it is simple and stable because it can absorb positional shifts and differences in size of parts due to individual differences in the face, and can also absorb variations in face size and slight inclination of the face at the front and back positions of the subject. Face can be detected.

In this embodiment, the face model is determined by applying the area model to both the input image and the image obtained by processing the input image, and measuring the characteristic amount of each determination element acquisition area. It is not necessarily limited to this,
A face model may be determined by applying a region model to one of an input image and an image obtained by performing image processing on the input image and measuring the feature amount of each determination element acquisition region.

(Second Embodiment) This embodiment also relates to face detection and is basically the same as the first embodiment, except that a cheek determination element acquisition area in an area model is used as a template. The point is that it has been replaced with an image.

That is, as shown in FIG. 12A, the cheek determination element acquisition region in the mask-represented face region model 51 is replaced with the template image 52. This template image has a configuration as shown in FIG. Note that the area of numerical value 1 is the determination element acquisition area 53 of the right eye, and the area of numerical value 2 is the determination element acquisition area 5 of the left eye.
The area of numerical value 3 is a nose determination element acquisition area 55, and the area of numerical value 4 is a mouth determination element acquisition area 56. This template image 52 may be a gray-scale image cut out from an actual face image. For example, since the cheek is almost flat, a grayscale image such as a grayscale value of 1,1,1,1,1,1,... Is used.
The similarity C-similar of the cheek region when template matching is performed with these images is obtained.

Let the template image of this cheek region be g,
Assuming that an image obtained by cutting out the cheek region at the designated position from the input grayscale image with the same size as g is f, the similarity C-s
Is

(Equation 12)

## EQU5 ## The similarity C-s takes a value between 0 and 1. This similarity is obtained and learned from various persons in the same manner as in the first embodiment, is stored in the memory as determination data, and is used together with the feature amounts of other areas as shown in FIG. . For the actual face determination, the similarity may be considered as one of the feature amounts, and the same processing as in the first embodiment may be performed.

As described above, the face detection similar to that of the first embodiment can be performed by combining the template matching of the cheek region and the feature amount of the other region in the region model instead of the template matching of the entire face. it can. In this embodiment, the similarity between the cheek region and the template is obtained. However, the present invention is not limited to this, and a distance from the template image may be used. There are various other similarity determination methods, and any of them can be applied.

In each of the above-described embodiments, the face detection is performed by sequentially designating the position where the region model is applied to the entire input image to extract the feature amount and perform face detection. Instead, as shown in FIG. 14, if the background is known in advance, a known area 57 is removed from the entire image to generate a candidate area 58, and the candidate area 58 In the same manner as in the above embodiment, the position where the region model is applied may be sequentially designated to extract the feature amount and perform face detection. By doing so, the extraction processing of the feature amount becomes faster.

(Third Embodiment) This embodiment is similar to the third embodiment.
An example in which an object other than a face is detected as an object to be identified is described. Specifically, an example of detecting a one-way road sign from a scene will be described. Although color features may be used using a color image, here, only a gray image is used. The target is a one-way sign in a scene image arranged horizontally in front. As an area model, an area model 6 including two areas 61 and 62 as shown in FIG.
Use 0.

The area 61 indicated by the arrow is denoted by A1, and the other areas 6
2 is A2. Next, the feature amount used in each of the regions A1 and A2 is determined. Generally, in a one-way sign, the luminance of the area A1 is higher than that of the other area A2, and the variance of the luminance is very small in both the areas A1 and A2. Therefore, assuming that the two regions A1 and A2 use the average luminance and the luminance variance value as feature amounts, respectively, A1-a,
A1-v, A2-a, and A2-v. Then, each feature value is measured from sample images of various one-way road signs,
A measurement value as shown in FIG. 16 is obtained. Then, as a correctness of each of the regions A1 and A2 from the learning result, for example, a fuzzy membership function shown in FIG. 17 is defined. Here, the correctness of the average luminance of the area A1 is defined as μ _A1-a . Similarly, the correctness of the luminance variance value of the area A1 is defined as μ _A1-v , and the correctness of the average luminance of the area A2 is μ _{μ1−v.
A2-a} , and the correctness of the luminance variance value of the area A2 is μ
Defined as _A2-v .

In the actual detection, first, an image of a scene including the marker 65 as shown in FIG. 18 is input, and the area model 60 is sequentially moved on the scene image.
1, the average luminance and the luminance variance of A2 are measured.
Calculate the likelihood F that the marker 65 is present at that location in a calculation formula such as = _A1-a x _A1-v x _A2-a x _A2-v ,
It is detected that the marker 65 is located at a position where F is larger than a certain threshold. In this manner, the area model 60 of the face in which the marker 65, which is the object to be identified, is divided into the characteristic determination element acquisition areas A1 and A2 of the marker image is created, and the positional relationship between the respective determination element acquisition areas is determined. Since the feature amount of each determination element acquisition area in the area model is measured while designating the position where the maintained area model 60 is applied to the scene image, and it is determined whether or not the entire area model is a sign, Simple processing can be realized, and simple and stable detection of a label can be achieved.

[0062]

According to the first to seventh aspects of the present invention, individual differences in the position of the characteristic region of the object to be identified, variations in size,
Object detection that can detect an object to be identified in consideration of fluctuations in the shooting environment, improve the detection rate, and realizes simple processing without the need to verify the positional relationship between characteristic regions. Equipment can be provided.

Further, according to the present invention, a face area model is created by dividing a face image into a characteristic area such as an eye, a nose, a mouth, and a cheek. And
A face is determined by sequentially specifying a position at which the region model maintaining the positional relationship of the region is applied to the image and acquiring a determination element such as a feature amount from each determination element acquisition region of the region model. There is no need to verify the positional relationship of parts such as the eyes, nose, mouth, etc., and simple processing can be performed. In addition, differences in the position and size of parts due to individual differences in the face can be absorbed, and the subject It is possible to provide an object detection device that can absorb a change in the size of the face and a slight inclination of the face at the front and rear positions, and can detect the face simply and stably.

According to the present invention, the object to be identified is detected in consideration of the individual difference in the position of the characteristic region of the object to be identified, the variation in size, and the variation in the imaging environment. As a result, it is possible to provide an object detection method that can improve the detection rate and can realize simple processing without having to verify the positional relationship between the characteristic regions.

According to the ninth to eleventh aspects of the present invention, an area model of a face which is expressed by dividing a characteristic image such as an eye, a nose, a mouth, and a cheek into an image including a face is prepared. And
A face is determined by sequentially specifying a position at which the region model maintaining the positional relationship of the region is applied to the image and acquiring a determination element such as a feature amount from each determination element acquisition region of the region model. There is no need to verify the positional relationship of parts such as the eyes, nose, mouth, etc., and simple processing can be performed. In addition, differences in the position and size of parts due to individual differences in the face can be absorbed, and the subject It is possible to provide an object detection method capable of absorbing a change in the size of the face and a slight inclination of the face at the front and rear positions, and enabling a simple and stable face detection.

[Brief description of the drawings]

FIG. 1 is a block diagram of an overall configuration showing a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of an image input unit according to the embodiment.

FIG. 3 is a diagram showing a grayscale image input and an edge image subjected to image processing in the embodiment;

FIG. 4 is a block diagram showing a configuration of a position detection unit according to the embodiment.

FIG. 5 is a view showing an example of a window model type area model which can be used in the embodiment;

FIG. 6 is a view showing an example of a mask model type area model that can be used in the embodiment;

FIG. 7 is an exemplary view showing various examples of fitting an area model to a face in the embodiment;

FIG. 8 is a view for explaining position designation and feature extraction to which the region model is applied in the embodiment.

FIG. 9 is a diagram showing an example of a feature amount of each region extracted from a sample face image in the embodiment.

FIG. 10 is a diagram showing an example of a fuzzy membership function of the low-luminance pixel amount RE-d in the right-eye determination element acquisition area in the embodiment.

FIG. 11 is a diagram showing an example of a result of face position detection according to the embodiment;

FIG. 12 is a view showing the second embodiment of the present invention and shows the use of a template image in a determination element acquisition area of a cheek in an area model of a face represented by a mask.

FIG. 13 is a view showing an example of a feature amount of each region extracted from a sample face image in the embodiment.

FIG. 14 is a view for explaining another example of position designation and feature amount extraction to which an area model is applied.

FIG. 15 is a diagram showing an example of a region model according to the third embodiment of the present invention.

FIG. 16 is a view showing an example of a feature amount of each region extracted from a sample image in the embodiment.

FIG. 17 shows an average luminance A1 of an area A1 in the embodiment.
The figure which shows the fuzzy membership function example of -a.

FIG. 18 is a view showing an example of a scene image in the embodiment.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 image input unit 2 image processing unit 3 position detection unit 13 image memory 32 model storage unit 33 position designation unit 34 feature extraction unit 36 determination unit

Claims (11)

[Claims] An image input unit for inputting an image; a storage unit for storing an area model in which a plurality of determination element acquisition areas are set corresponding to characteristic areas of an image of an object to be detected; Position designation means for sequentially designating a position at which the area model stored in the storage means is applied to one or both of an input image input by the input means and an image obtained by performing image processing on the input image; Each time an area model is sequentially applied to the position specified by the means, a determination element obtaining means for obtaining a determination element from each determination element obtaining area of the area model, and a determination element obtained by the determination element obtaining means. An object detection apparatus, comprising: a determination unit configured to determine whether an image is an object to be identified, and detecting an object to be identified based on a determination result of the determination unit. 2. An image input means for inputting an image, and an area model in which a plurality of determination element acquisition areas are set corresponding to characteristic areas such as eyes, mouth, nose, and cheek of a face image to be detected are stored. Storage means, and a position designation for sequentially designating a position at which the area model stored in the storage means is applied to one or both of an input image input by the image input means and an image obtained by performing image processing on the input image. Means, a determination element obtaining means for obtaining a determination element from each determination element obtaining area of the area model each time the area model is sequentially applied to the position designated by the position specification means, and An object detection apparatus, comprising: determination means for determining whether or not the image is a face image based on the determined determination element, and detecting a face based on the determination result of the determination means. 3. The object detecting apparatus according to claim 1, wherein the determination element obtaining unit obtains a feature amount as a determination element from each determination element obtaining area of the area model. 4. The method according to claim 3, wherein the determination element obtaining means obtains a feature amount of at least one determination element obtaining area of the area model using a template image.
An object detection device according to any one of the preceding claims. 5. When at least one of an eye, a mouth, and a nose region is set as a determination element acquisition region of a region model, the determination element acquisition unit determines a gray level value in the region for at least one of the regions. 5. The object detecting apparatus according to claim 3, wherein the feature amount is obtained using a differential value of the gray value. 6. When a cheek area is set as a determination element acquisition area of an area model, the determination element obtaining means obtains a feature amount for the area by using a variance value of gray levels in the area. The target object detection device according to claim 3 or 5, wherein: 7. The method according to claim 7, wherein when a cheek region is set as the determination element acquisition region of the region model, the determination element acquisition unit acquires a feature amount of the region using a template image. 6. The object detection device according to 3 or 5. 8. Acquisition of a plurality of determination elements for one or both of an input image and an image obtained by performing image processing on the input image, in accordance with a characteristic region of a target object image to be detected. The area model in which the area is set is applied while sequentially specifying the position, and each time the area model is applied, a determination element is obtained from each determination element acquisition area of the area model, and the identification target is determined based on the obtained determination element. An object detection method, comprising: determining whether an image is an object image; and detecting an object to be identified based on the determination result. 9. A method according to claim 1, wherein one or both of the input image and an image obtained by subjecting the input image to image processing correspond to characteristic regions such as eyes, mouth, and nose of a face image to be detected. The area model in which the determination element acquisition area is set is applied while sequentially specifying the position, and each time the area model is applied, a determination element is obtained from each determination element acquisition area of this area model, and based on the obtained determination element, A method for determining whether or not the image is a face image, and detecting a face based on a result of the determination. 10. The object detection method according to claim 8, wherein a feature amount is acquired as a determination element from each determination element acquisition area of the area model. 11. The object detection method according to claim 10, wherein a feature amount is acquired for at least one determination element acquisition area of the area model using a template image.