KR20130125283A - Apparatus and method for body components detection - Google Patents

Abstract

The present invention provides a method for measuring human body parts and an apparatus therefor. The method includes the following steps of: identifying a group sub image related to human body parts in an image to be measured; assigning confidence coefficients about human body parts which exist in each sub image of the group sub image based on the basic visual features of sub images and the expansion features of the peripheral areas of the sub images; and measuring the positions of the human body parts by merging sub images which have higher confidence coefficients than a threshold value. The method and the apparatus according to the present invention are able to improve the function of a human body part measurement more efficiently. [Reference numerals] (310) Identify a sub image;(320) Assign confidence coefficients about human body parts in each sub image;(330) Merge sub images which have higher confidence coefficients than a threshold value

Description

Apparatus and Method for Body Components Detection

The present invention relates to computer vision and mode identification technology, and more particularly, to a method and apparatus for measuring a human body part based on a multi-part context descriptor.

Human posture estimation is a fundamental task of computer vision technology, which is widely used in various areas such as human-machine interaction, gay games, movies, and real role animation, 3D. Recently, due to his technical and commercial value, the study of the human body posture estimation has become the focus of the computer visual field. In particular, the measurement of the human body part is a basic technique in the human body posture estimation technique. Human body measurement techniques can provide important observational information to infer human posture.

In human body measurement, object measurement can be divided into two types. That is, the measurement of the human body part based on the object and the measurement of the human body part based on the pixel. The method of measuring a body part based on an object is regarded as a positive sample using each pixel in the object area, and the relationship between the pixel of the object area and its surrounding pixels is typically used as a feature descriptor. The method of measuring a human body part based on pixels may measure an object using surrounding information, but does not use an internal feature and a contour feature of the object. The method of measuring a body part based on pixels considers the entire subject area as a positive sample and describes the area using some features. The method of measuring the body part based on the object uses the internal features and the contour features of the object but does not use the surrounding information of the object.

In order to improve the accuracy of human body measurement, there is a need for a human body measurement method and apparatus that combines the advantages of an object based measurement and a pixel based measurement method.

According to one aspect of the invention, there is provided a method for measuring a human body part. The method includes identifying a group sub image associated with a body part in the image to be measured; Assigning a confidence coefficient to a body part in each sub-image of the group sub-image based on a basic visual feature of the sub-image and an extended feature of the peripheral region of the sub-image; And merging sub-images whose confidence coefficient is higher than a threshold value to measure the position of the body part.

According to an aspect of the present invention, the step of assigning a confidence coefficient to the human body part in each sub-image of the group sub-image, defines a multi-part context descriptor of each sub-image The site context descriptor includes a base descriptor and an extension descriptor; Assigning a confidence coefficient to each sub-image based on the similarity between the multi-site context descriptor of each sub-image and the multi-site descriptor of a pre-trained human body part. The basic descriptor describes basic visual features of the human body part of the sub-image, and the extension descriptor describes the spatial structure relationship between the human body part and the surrounding human body part of the sub-image.

According to one aspect of the invention, the sub-image is an image area having different geometric shapes, positions, and sizes in the image to be measured.

According to one aspect of the present invention, a plurality of sub-images in the sub-image group overlap each other.

According to one aspect of the invention, the method further comprises preprocessing the image to be measured before identifying the image to be measured.

According to one aspect of the invention, the pretreatment step comprises: performing quantization on the data of the image to be measured; Measuring image 3D information or image depth information on an image to be measured; Dividing the image to be measured; Extracting the foreground of the image to be measured.

According to one aspect of the invention, the basic visual feature is an ordinal (ordinal) feature; Binary features; Haar feature; Histogram of gradient description features; Contour features; At least one of the gray histogram features.

According to one aspect of the present invention, merging the sub-images comprises one of two methods: direct merging and weighting merging, and the threshold value depends on the selection and combination for the multi-site context descriptor.

According to one aspect of the present invention, the direct merging step includes selecting a sub image having a higher confidence coefficient than a predetermined value and merging positions of the selected sub image directly by statistical and geometric methods. .

According to one aspect of the present invention, the weighted merging step includes merging the positions of the sub images based on different confidence coefficient threshold values, or merging the positions of the sub images according to a clustering algorithm.

According to another aspect of the present invention, a human body measuring device is provided. The apparatus includes a sub image identification unit for identifying a group sub image associated with a human body part in an image to be measured; A confidence coefficient is assigned to a human body in each sub-image of the group sub-image based on a basic visual feature of the sub-image and an extended feature of a peripheral region of the sub-image, and the sub-coefficient is a sub-threshold having a threshold value higher than a threshold value. It includes a human body part identification unit for merging the images to measure the position of the human body part.

According to another aspect of the present invention, the apparatus for measuring a human body part further includes a training unit configured to perform training on a sample image to obtain a multi-part context descriptor of the human body part. The multi-site context descriptor includes a base descriptor and an extension descriptor, wherein the base descriptor describes basic visual features of the human body portion of the sub-image, and the extension descriptor is a spatial structure relationship between the human body portion and the surrounding human body portion of the sub-image. Describe. The human body part identification unit defines a multi-site context descriptor of each sub-image and trusts each of the sub-images based on the similarity between the multi-site context descriptor of each sub-image and the multi-site descriptor of the human body trained in the training unit. Assign coefficients.

According to another aspect of the present invention, the human body region measuring device further includes a preprocessing unit that performs preprocessing on the image or sample image to be measured.

Hereinafter, exemplary embodiments will be described with reference to the accompanying drawings. This makes the above and other objects and advantages clearer.
1 is a flowchart illustrating a training process of a method for measuring a human body part according to the present invention.
2A-2C illustrate a multi-part context descriptor in a depth image in accordance with an embodiment of the invention.
3 is a flowchart illustrating a method of measuring a human body part according to an exemplary embodiment of the present invention.
Figure 4 illustrates a human body area measurement system according to an embodiment of the present invention.

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

1 is a flowchart illustrating a training process of a method for measuring a human body part according to the present invention.

As shown in FIG. 1, the training process of the method for measuring a human body part according to the present invention includes: training image data building step 110; Preprocessing step 120; The training sample preparation step 130 and the human body sorter training step 140 include four steps. Hereinafter, each step will be described in detail.

First, in the training image data construction step 110, the training image data may be constructed by synthesizing the image data or the real image data.

Next, the preprocessing step 120 performs quantization and background removal on the training image data. The quantization process is a representation of original image data as quantized depth image data to perform subsequent processing. For example, a gray image is a representative quantization depth image, and the value of its image data is quantized from 0 to 255. By quantization, the original image data is quantized to reduce noise or reduce computation. In general, background removal processing is performed based on the limitation on the depth value, the area extraction involved, the motion capture, and the like.

Specifically, the processing on the training image data may include performing quantization on the training image data; Measuring image 3D information or image depth information on the training image data; Dividing the training image data; The method may include one of extracting a foreground of the training image data.

로 표시할 수 있고,

및 는 부위 영역의 주변 영역의 중심 위치를 의미하고, R는 주변 영역의 사이즈를 의미하고, M는 주변 영역의 분할된 형상을 의미한다. 분할된 형상은 사각형, 원형 등이 될 수 있다. 네거티브 샘플은 측정될 인체 부위를 포함하는 서브 이미지와 반대의 서브 이미지인 것이다. 즉, 측정될 인체 부위를 포함하지 않는 서브 이미지인 것이다. 일반적으로 샘플 트레이닝 단계에서 두 기지 타입의 네거티브 샘플을 사용한다. 하나는 인체가 포함되지 않는 배경 이미지인 것이고, 또 하나는 해당 트레이닝 인체 부위가 이미 추출된 인체 이미지인 것이다. 네거티브 샘플의 트레이닝에 의하여 서브 이미지에 측정될 인체 부위가 포함되는지 구분할 수 있다. Next, in the training sample preparation step 130, three types of training samples are prepared. That is, site samples, site context samples, and negative samples. The site sample focuses on the site object itself. The site context sample focuses on the surrounding area of the site. The division for the area around the site

Can be displayed as

And Denotes the central position of the peripheral region of the site region, R denotes the size of the peripheral region, and M denotes the divided shape of the peripheral region. The divided shape may be a rectangle, a circle, or the like. The negative sample is the sub-image opposite to the sub-image containing the body part to be measured. That is, the sub image does not include the human body part to be measured. Typically, two known types of negative samples are used in the sample training phase. One is a background image that does not include the human body, and the other is a human image from which the training human body part has already been extracted. By training the negative sample, it is possible to distinguish whether the sub-image includes a human body part to be measured.

Next, the human body classifier training step 140 performs a human body classifier training to measure the human body region. Human body classifier training includes describing a sample; Improving training; Steps to build a meter include three steps.

In describing the sample, the sample is described based on the basic feature of each sub-image and the extended feature of the peripheral area of the sub-image. According to an embodiment of the present invention, a basic feature of each sub-image and an extended feature of a peripheral area of the sub-image can be expressed using a multi-site context descriptor. The multi-site context descriptor according to an embodiment of the present invention may not only describe basic visual features of the object of each sub-image, but may also describe context information, that is, extended features, of neighboring regions adjacent to the region. For example, in the embodiment of the present invention, the feature of the multi-size ordinal mode (MSOP) may be used as the basic feature of the multi-site context descriptor.

Based on the multi-site context descriptor of the MSOP, it can be expressed as equation (1) as follows:

Equation (1)

로 표현할 수 있다.

는 MSOP 모드의 파라미터를 의미하고 MSOP 모드의 파라미터는 모드의 위치, 크기 및 유형을 포함할 수 있다. 본 발명의 실시예에 따른 MSOP 모드는 서브 이미지에 포함된 부위의 기본적인 특징을 기술하기 위한 MSOP 모드 및 주변 부위의 확장 특징을 기술하기 위한 MSOP 모드를 포함할 수 있다. 그러나 MSOP 모드에 근거하는 특징 기술자는 본 발명의 실시예에 따른 멀티 부위 콘텍스트 기술자의 한 예시로써 본 발명이 이에 의하여 제한되지 않는다. 본 발명은 기타의 시각 특징 기술 방식으로 멀티 부위 콘텍스트 기술자를 형성할 수 있다. 예를 들어, 이미지의 Haar 특징; 기울기 히스토그램(HOG) 특징; 바이너리(binary) 특징; 윤곽 특징; 그레이(gray) 히스토그램 특징을 이용하여 멀티 부위 콘텍스트 기술자의 기본적인 기술자로 간주한다.In Equation (1), b (x) means a Boolean function, and if x> 0, b (x) = 1, and if x <0, b (x) = 0. gi is the pixel value in the grid (or subimage) in MSOP mode, i is the index of the grid, gc is the average pixel value for the grid in the subimage and the area around the subimage

.

Denotes a parameter of the MSOP mode and the parameter of the MSOP mode may include the position, size, and type of the mode. The MSOP mode according to an embodiment of the present invention may include an MSOP mode for describing basic features of a region included in a sub image, and an MSOP mode for describing extended features of a peripheral region. However, the feature descriptor based on the MSOP mode is an example of a multi-site context descriptor according to an embodiment of the present invention, and the present invention is not limited thereto. The present invention can form multi-site context descriptors in other visual feature description manners. For example, Haar features of the image; Slope histogram (HOG) features; Binary features; Contour features; The gray histogram feature is used as the basic descriptor of a multi-site context descriptor.

Hereinafter, a mode of a multi-site context descriptor according to an embodiment of the present invention will be described with reference to FIGS. 2A-2C.

를 도시한 것이다. 도 2C는 본 발명의 실시예에 따른 멀티 부위 콘텍스 기술자에 있는 부위 주변 영역의 콘텍스트 정보를 기술하는 확장 기술자의 모드 파라미터

를 도시한 것이다. 상기 기술 방식에 있어서, x및y 는 기술자의 위치 파라미터를 의미하고, w및 h는 기술자의 형상 파라미터를 의미하고, t는 기술자의 유형을 의미한다. 상기 기술 방식으로, 주변 영역의 크기 및 형상은 측정될 인체 부위를 포함하는 서브 이미지와 일정 관계를 구비하는 것으로 정의될 수 있다. 예를 들어, 주변 영역의 위치는 서브 이미지의 영역을 중심으로 하는 더 큰 영역이 될 수 있고(도2B의 블록205 및 207을 참조), 서브 이미지 영역의 위쪽, 아래쪽, 왼쪽, 오른쪽 영역이 될 수도 있다. 주변 영역의 크지는 서브 이미지 영역의 2배, 3배, 3.5배등이 될 수 있다. 주변 영역의 형상은 서브 이미지와 같은 형상이 될 수 있고 고정된 사각형, 원형 등도 될 수 있다. 2A-2C illustrate mode parameters of a multi-site context descriptor in accordance with an embodiment of the invention. Referring to FIG. 2B, blocks 201 and 203 refer to sub-images including a human body part. The human body part contained in block 201 is the upper arm of the human body and the human body part contained in block 203 is the head of the human body. Blocks 205 and 207 refer to peripheral regions corresponding to blocks 201 and 203. 2A illustrates a mode parameter of a basic descriptor describing the visual characteristics of a portion in a multi-site context descriptor in accordance with an embodiment of the present invention.

It is shown. 2C is a mode parameter of an extended descriptor describing context information of a region around a region in a multi-site context descriptor according to an embodiment of the present invention.

It is shown. In the above description scheme, x and y denote descriptor positions of the descriptor, w and h denote descriptors of the descriptor, and t denotes the descriptor type. In this manner, the size and shape of the peripheral area can be defined as having a certain relationship with the sub-image containing the human body part to be measured. For example, the position of the peripheral area may be a larger area centered on the area of the sub image (see blocks 205 and 207 of FIG. 2B) and may be the top, bottom, left, and right areas of the sub image area. It may be. The size of the peripheral area may be 2 times, 3 times, 3.5 times, etc. of the sub image area. The shape of the peripheral area may be a shape like a sub image, or may be a fixed rectangle, a circle, or the like.

With respect to the head, the multi-site context descriptor in accordance with an embodiment of the present invention can not only describe the features of the head (eg, oval contour features), but also describe information about the neck or shoulder area around the head. Can be. With respect to the upper arm, the multi-site context technician according to an embodiment of the present invention can not only describe the shape of the upper arm, but also describe information about the upper body around the upper arm. Accordingly, the multi-site context descriptor according to an embodiment of the present invention not only includes the internal features and the contour structure of the site object but also the context information of the surrounding area. Thus, multi-site context descriptors according to embodiments of the present invention have higher stability compared to local descriptors of the current technology.

In addition, as described above, the illustration of the human body by the head and upper arm in Figure 2, but those skilled in the art should understand that the human body can be more precisely distinguished. For example, the human body can be head, left upper arm, left lower arm, left hand, right upper arm, lower right arm, right hand, left thigh, left calf, right thigh, right calf, etc. As a result, more precise multi-site context descriptor training can be performed for each of the sites.

Referring back to FIG. 1, in the step of improving training, all types of classifiers are trained to measure human parts, and each type of classifier represents a feature of the sub-image. For example, we train our classifier through algorithms such as SVM, Forest, and Boosting. In an embodiment of the present invention, classifier training is performed using the AdaBoost algorithm. A multi-site context descriptor based on AdaBoost can be expressed as equation (2) as follows.

Equation (2)

는 부위 대상 자체의 특징을 기술하기 위한 분류기를 의미하고,

는 부위 주변 영역에 대한 콘텍스트 정보의 특징을 기술하기 위한 분류기를 의미하고,

는

분류기의 개수를 의미하고,

는

분류기의 개수를 의미하고, F(x)는 최종의 분류기를 의미한다.In equation (2), x means a sample,

Means a classifier for describing the characteristics of the site object itself,

Means a classifier for describing the characteristics of the context information for the region around the site,

The

Means the number of classifiers,

The

Means the number of classifiers, and F (x) means the final classifier.

Then, the cascade of the various kinds of classifiers trained in the step of constructing the body part measurer improves the performance of the body part measurer.

Hereinafter, a method of measuring a human body part according to an exemplary embodiment of the present invention will be described with reference to FIG. 3.

First, in step 310, a group sub image associated with the human body part in the image to be measured is identified. Specifically, in the identifying process, the sub image may be searched based on the specified position and size of the depth image, and it may be determined whether the sub image is a human body part to be measured.

로 표현할 수 있고,

는 각각 서브 이미지를 검색할 때 사용된 가장 작은 사이즈, 가장 큰 사이즈, 및 step size 를 의미하고,

는 각각 서브 이미지를 검색할 때 사용된 이니셜 검색 포인트 위치, 종점 위치 및 searching step size를 의미한다. 다리 말하면, 서브 이미지는 측정될 이미지에 서로 다른 기하 형상, 위치 및 크기를 구비하는 이미지 영역이 될 수 있고 서브 이미지 그룹에 있는 복수의 서브 이미지가 서로 오버랩이 될 수 있다. 구체적으로 상기 검색 과정은 이미지에서 순환으로 서브 이미지 영역을 추출하는 과정으로 간주될 수 있다. 우선, 이니셜 사이즈 Smin부터 서브 이미지 영역의 크기를 결정하고, 그 다음, 이미지의 이니셜 검색 포인트 위치Pstart로부터 서브 이미지 영역의 중심 위치를 결정하고 종점 위치 Pend에 도작할 때까지 Ps를 step로 하여 순서대로 이미지 영역의 중심 위치를 이동하고, 트래버스(traverse)된 각 위치에서 서브 이미지를 추출한다. 그 다음, 가장 큰 사이즈Smax 가 될 때까지 step Sstep에 의하여 순서대로 서브 이미지 영역의 크기를 증가한다. 각 영역 사이즈마다 한번씩 서브 이미지의 중심으로 Pstart로부터 Pend까지 트래버스하여 검색할 모든 서브 이미지를 획득한다.The search process

Can be expressed as,

Means the smallest, largest, and step size, respectively, used when searching for sub-images,

Denotes initial search point position, endpoint position and searching step size used when searching for sub-images, respectively. In other words, the sub image may be an image area having different geometric shapes, positions and sizes in the image to be measured, and a plurality of sub images in the sub image group may overlap each other. In detail, the searching process may be regarded as a process of extracting a sub image region from an image in a cycle. First, the size of the sub-image area is determined from the initial size Smin. Then, the center position of the sub-image area is determined from the initial search point position Pstart of the image, and then in order, Ps is set in step until it reaches the end position Pend. The center position of the image area is moved and a sub image is extracted at each traversed position. Then, the size of the sub image area is sequentially increased by step Sstep until the largest size Smax is reached. Once for each region size, all the sub images to be searched are acquired by traversing from Pstart to Pend as the center of the sub image.

Preferred is the extraction of the human body in the image to be measured via foreground extraction techniques in the identification process. By extracting the foreground, only the foreground area can be measured and the amount of search for the sub image position can be reduced. In addition, a depth value of the extracted foreground object may be measured to reduce a search range for the sub image. In addition, the sub-image may be identified by image 3D information and image segmentation.

Then, in step 320, a confidence coefficient is assigned to the human body part of each sub-image in the group sub-image based on the basic feature description of each sub-image and the extended feature description of the area around the sub-image. In other words, it is determined based on the trained multi-site context descriptor to determine if the sub-image is a human body part. Specifically, confidence coefficients are calculated for the sub-images in terms of the final classifier F (x) (see equation (2)) finally obtained by human body classifier training. First, we calculate the corresponding classifier f according to the characteristics of the human body classifier training part, overlay the calculated values for each classifier with a combination of F (x) expressions, and overlay the output values of the corresponding sub-images. Considered as a confidence coefficient.

Next, in step 330, the position of the body part is measured by merging sub-images whose confidence coefficient is higher than a threshold value. Here, the threshold value depends on the selection and combination of types of multi-site context descriptors and according to an embodiment of the present invention, the merging step may include both direct merging and weighting merging. In the direct merging step, a sub-image having a higher confidence coefficient than a predetermined value is selected to directly merge the positions of the selected sub-images by statistical and geometric methods to finally obtain the position of the human body part. Specifically, the average center position and the average size size of the sub-image meeting all the reliability coefficient requirements can be calculated as the final human body position by means of the average value. In addition, the region where the sub-image that meets all the reliability coefficient requirements is most concentrated can be used as the final output position of the human body part.

In the weighted merging step, the positions of the sub images are merged based on different confidence coefficient threshold values, or the positions of the sub images are merged according to a clustering algorithm. The clustering algorithm may include mean-shift and k near neighbor. Specifically in weighted merging, the confidence coefficients for sub-images that meet the confidence coefficient requirements can be maintained simultaneously and determine the importance of the sub-images in the merge. For example, the weighted average center position and the weighted average size size of the sub-image may be the position of the final human body part.

Hereinafter, with reference to Figure 4 will be described the operating principle of the human body area measurement system according to an embodiment of the present invention. As shown in FIG. 4, the human body part measuring system according to the exemplary embodiment of the present invention includes an image obtaining apparatus 410 and a body part measuring apparatus 420. Although the image acquisition device 410 and the human body measuring device 420 are separately installed in FIG. 4, the two devices may be realized as one facility.

In addition, devices and facilities, such as a Primesence mapping device, a Time-of-Flight camera, and a multi-view camera, may be regarded as the image acquisition device 410. The image data acquired by the image acquisition device 410 may be used as training image data and may be used as image data to be measured.

The human body region measuring apparatus 420 includes an image preprocessor 421, a training unit 422, a sub image identification unit 423, and a human body region identification unit 424.

The image preprocessor 421 is used to preprocess the image data. Here, the image preprocessor 421 performs preprocessing on the acquired image data or the sample image data stored in the training sample image database. For example, the image preprocessor 421 may perform quantization on the image data and the foreground extracting process and the depth may be performed on the image data for the training unit 422 and the sub image identification unit 423 to perform subsequent processing. Value measurement, image 3D information measurement, image segmentation and the like can be performed.

The training unit 422 prepares a training sample from the training sample image data, and trains a human body classifier (that is, a multi-site context descriptor) that measures a human body part using the training sample, and uses the human body part classifier to Build. The training sample image data may be image data stored in advance in the image database or may be image data acquired by the image acquisition device 410.

The sub image identification unit 423 identifies a group sub image related to the human body part in the image to be measured. The human body part identifying unit 424 assigns a confidence coefficient to the human body part in each sub-image of the group sub-image based on the basic visual feature of each sub-image and the extended feature of the peripheral area of the sub-image. Sub-images higher than the threshold value are merged to measure the position of the human body part. In an exemplary embodiment of the present invention, the human body region identification unit 424 defines a multi-site context descriptor of each sub-image and based on the similarity between the multi-site context descriptor of each sub-image and the multi-site descriptor of the trained human site. Assign confidence coefficients to subimages. In addition, the human body part identification unit 424 may output the measured result to an external device, process it by the external device, and then display the final human body part identification result.

The configuration of each module for the body part measurement device 420 described above may be divided into more modules or merged into a smaller number of modules by way of example.

Compared with the human body measurement technique based on the object currently used and the human body measurement technique based on the pixel, the multi-site context technician according to the present invention can describe the internal and contour features of the human body object itself. In addition, the characteristics of the region surrounding the human body part may be described. Therefore, it is possible to reduce the complexity of the classifier because it is possible to improve the performance of the body region meter using the surrounding information of the human body part and to achieve the purpose of the classifier training more quickly. Therefore, the human body part measuring method according to the present invention can more efficiently improve the performance of the human body part measurement.

As described above, the present invention has been described by way of a limited embodiment, but the present invention is not limited to the above embodiments, and those skilled in the art to which the present invention pertains may make various modifications and variations from this description. Do. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be determined by the equivalents of the claims, as well as the claims.

Claims (10)

Identifying a group sub image associated with the body part in the image to be measured;
Assigning a confidence coefficient to a body part in each sub-image of the group sub-image based on a basic visual feature of the sub-image and an extended feature of the peripheral region of the sub-image;
Measuring the position of the body part by merging sub-images whose confidence coefficient is higher than a threshold value
Human body measurement method comprising a. The method of claim 1,
Assigning a confidence coefficient to the human body part in each sub-image of the group sub-image,
Defining a multi-part context descriptor of each sub-image, wherein the multi-part context descriptor includes a basic descriptor and an extended descriptor;
Assigning a confidence coefficient to each sub-image based on the similarity between the multi-site context descriptor of each sub-image and the multi-site descriptor of a pre-trained human body part,
The basic descriptor describes basic visual features of the human body part of the sub-image, and the extension descriptor describes the spatial structure relationship between the human body part and the surrounding human body part of the sub-image. The method of claim 1,
Preprocessing the image to be measured before identifying the image to be measured
Human body measuring method further comprising. The method of claim 3,
The pretreatment step,
Performing quantization on the data of the image to be measured;
Measuring image 3D information or image depth information on an image to be measured;
Dividing the image to be measured;
And extracting a foreground of the image to be measured. 3. The method of claim 2,
The basic visual feature,
Ordinal features; Binary features; Haar feature; Histogram of gradient description features; Contour features; A method for measuring a body part comprising at least one of gray histogram features. 3. The method of claim 2,
Merging the sub image,
A method for measuring a body part, comprising one of two methods, direct merging and weighting merging, wherein the threshold value depends on the selection and combination for the multi-site context descriptor. The method according to claim 6,
The direct merge step,
Selecting a sub-image having a higher confidence coefficient than a predetermined value and merging the positions of the selected sub-images directly by statistical and geometric methods. The method according to claim 6,
The weighted merge step,
Merging the positions of the sub-images based on different confidence coefficient threshold values or merging the positions of the sub-images according to a clustering algorithm. A sub image identification unit identifying a group sub image related to a human body part in the image to be measured;
A confidence coefficient is assigned to a human body in each sub-image of the group sub-image based on a basic visual feature of the sub-image and an extended feature of a peripheral region of the sub-image, and the sub-coefficient is a sub-threshold having a threshold value higher than a threshold value. Body part identification unit for merging the images to measure the location of the body part
Human body measuring device comprising a. 10. The method of claim 9,
Training unit for training the sample image to obtain a multi-part context descriptor of the human body part
Further comprising:
The multi-site context descriptor includes a base descriptor and an extension descriptor, wherein the base descriptor describes basic visual features of the human body portion of the sub-image, and the extension descriptor is a spatial structure relationship between the human body portion and the surrounding human body portion of the sub-image. Describe it,
The human body part identification unit defines a multi-site context descriptor of each sub-image and trusts each of the sub-images based on the similarity between the multi-site context descriptor of each sub-image and the multi-site descriptor of the human body trained in the training unit. An apparatus for measuring a body part, which assigns a coefficient.

Images