KR101541384B1 - Device for Recognition of Object and method - Google Patents

Abstract

The present invention relates to image recognition, which includes a camera for performing image acquisition, a storage unit for storing images for image recognition of the acquired image, a color histogram for the acquired image, and a color histogram similarity And a controller for performing image recognition on the acquired image based on the comparison.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device,

The present invention relates to image recognition, and more particularly, to a method and apparatus for recognizing an image using a color histogram.

Since image recognition technology is intuitive and widely applicable to various fields of real life, much research has been done for a long time. Especially, in recent years, a portable terminal equipped with a camera has been popularized, and various services through image recognition have been required to be developed. The image recognition process may include an image acquisition process, an analysis process on the acquired image, a process of image comparison with previously stored information, and a recognition process. Since the image recognition process requires a lot of data comparison in the image comparison process, high-performance hardware and software supporting the same must be provided. In the conventional image recognition process, there is a problem that the image processing speed can not have a satisfactory value due to a very large amount of data comparison process. In addition, there is a problem that image recognition results for the same image are derived differently depending on the degree of light included in the image and the angle at which light is incident.

It is therefore an object of the present invention to provide an image recognition method and apparatus capable of having robust image recognition characteristics with respect to optical characteristics.

It is another object of the present invention to provide an image recognition method and apparatus capable of providing a faster image processing speed in an image recognition process.

According to another aspect of the present invention, there is provided an image processing apparatus including a camera for performing image acquisition, a storage unit for storing images for image recognition of the acquired image, a color histogram for the acquired image, And a controller for performing image recognition on the acquired image based on the similarity comparison.

In the image recognition apparatus according to the present invention, the controller extracts minutiae from the acquired image, performs minutia matching with the stored images, and stores the stored image having a matching ratio of more than a certain ratio or the storage having the best matching value And then the selected object in the acquired image is divided using the selected stored image.

In the image recognition apparatus according to the present invention, the controller divides an object by providing coordinates of the selected storage image to the acquired image using ORSA (Optimized Random Sampling Algorithm).

In the image recognition apparatus according to the present invention, the controller compares the color histogram of the specific object with the color histogram similarity of the stored image in the acquired image.

In the image recognition apparatus according to the present invention, the controller derives a storage image having the best combination of the feature point matching value and the color histogram similarity as the image recognition result.

In the image recognition apparatus according to the present invention, the controller performs feature point matching between the acquired image and the stored image based on a K-NN algorithm.

In the image recognition apparatus according to the present invention, the controller performs feature point matching between the acquired image and the stored image based on LSH (Local Sensitive Hashing) and FLANN algorithm.

In the image recognition apparatus according to the present invention, the controller calculates a color histogram in which a low bit value is ignored among pixel values of the acquired image.

In the image recognition apparatus according to the present invention, the control unit may perform classification display of a specific object in the acquired image, display of a stored image corresponding to the acquired image, display of information related to the recognized image, , A connection to a server device related to the image-recognized stored image, and a specific object tracking in the acquired image.

The present invention also provides an image processing method including the steps of acquiring an image, extracting a feature point of the obtained acquired image, performing a feature point matching of feature points of previously acquired stored images and feature points of the acquired image, Selecting a stored image having an ideal or a highest ratio; dividing a specific object of the acquired image using the stored image; calculating a color histogram of the divided object in the acquired image; Calculating a degree of similarity between a color histogram of an object of the acquired image and a color histogram of the stored image, and deriving an image recognition result of the acquired image based on the similarity value of the color histogram, A method of recognizing an image is disclosed.

In the image recognition method according to the present invention, the feature point matching process may include performing feature point matching between the acquired image and the stored image based on a K-NN algorithm, based on LSH (Local Sensitive Hashing) and FLANN algorithm And performing a feature point matching between the acquired image and the stored image.

In the image recognition method according to the present invention, the object segmentation process may include dividing an object by providing coordinates of the selected storage image to the acquired image using ORSA (Optimized Random Sampling Algorithm) .

In the image recognition method according to the present invention, the color histogram calculation process may be performed by dividing a high bit and a low bit into a predetermined standard defined for each pixel value represented by a certain bit in the object region of the acquired image, And calculating a color histogram by selecting the color histogram.

In the image recognition method according to the present invention, the derivation process is a process of deriving a stored image having the best combination of the color histogram similarity and the feature matching value as an image recognition result.

In the image recognition method according to the present invention, after the projecting process, a specific object in the acquired image is classified and displayed according to the image recognition result, a process of displaying a stored image corresponding to the acquired image, Displaying the information related to the recognized stored image, performing a server device connection with respect to the image-recognized stored image, and tracking a specific object in the acquired image. .

The image recognition apparatus according to the present invention can provide a faster image processing speed and improve the image recognition speed.

In addition, the image recognition apparatus according to the present invention can minimize image recognition errors according to optical characteristics and perform accurate image recognition based on robust image recognition performance with respect to optical characteristics.

1 is a block diagram illustrating a configuration of an image recognition apparatus according to an embodiment of the present invention.
2 is a detailed block diagram of a control unit of the image recognition apparatus according to an embodiment of the present invention.
FIG. 3 is a view for explaining feature points extraction among image recognition functions according to an embodiment of the present invention.
4 is a diagram for explaining object detection among image recognition functions according to an embodiment of the present invention.
5 is a diagram for explaining color space conversion among image recognition methods according to an embodiment of the present invention.
6 is a diagram illustrating an object recognition result in the image recognition method according to an embodiment of the present invention.
7 is a view for explaining an image recognition method according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description, only parts necessary for understanding the operation according to the embodiment of the present invention will be described, and the description of other parts will be omitted so as not to disturb the gist of the present invention.

The terms and words used in the present specification and claims should not be construed as limited to ordinary or dictionary meanings and the inventor is not limited to the meaning of the terms in order to describe his invention in the best way. It should be interpreted as meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in the present specification and the configurations shown in the drawings are merely the most preferred embodiments of the present invention and are not intended to represent all of the technical ideas of the present invention, so that various equivalents It should be understood that water and variations may be present.

1 is a diagram schematically showing a configuration of an image recognition apparatus according to an embodiment of the present invention. The image recognition apparatus 100 illustrated in FIG. 1 may include a camera 110 and may be a portable terminal device such as a smart phone.

Referring to FIG. 1, the image recognition apparatus 100 of the present invention may include a camera 110, an input unit 120, a display unit 140, a storage unit 150, and a controller 160.

The image recognition apparatus 100 having such a configuration performs feature point extraction on an image acquired by the camera 110 and performs comparison with a plurality of stored images previously stored in the storage unit 150 to obtain a specific storage image And supports image recognition function by comparing similarity of color histogram. In this process, the image recognition apparatus 100 of the present invention can minimize the recognition error of the optical characteristics by using the color histogram of the acquired image and the stored image.

The camera 110 is a device for acquiring an image. The camera 110 may be a CCD camera or the like. Description of the Related Art [0002] A CCD camera (charge-coupled device camera, CCD camera) is a digital camera that converts an image into an electrical signal using a charge coupled device (CCD), and stores the digital data in a storage medium such as a flash memory. CCD is an optical sensor semiconductor that converts light into electrical signals from a digital camera. The light transmitted through the lens and diaphragm is converted into an electrical signal by the intensity of the light by the CCD. The signal is converted into an image file through an ADC converter which converts the analog signal into a digital signal of 0 and 1 And stored in memory. In particular, the intensity of light incident through the lens and diaphragm is recorded in the CCD, separated into different colors by the RGB color filters disposed adjacent to the CCD, and then converted into electrical signals in a number of photosensitive elements constituting the CCD, . The control unit 160 may control the CCD image stored in the memory disposed in the CCD camera to be collected in real time for real time image processing. The camera 110 may transmit a moving image or a preview image to the control unit 160. [ The image recognition function of the present invention can support object recognition in a moving image or a preview image. In the following description, a preview image, moving image, still image, or the like acquired by the camera 110 may be an acquired image compared with a stored image.

The input unit 120 may generate various input signals for controlling the image recognition apparatus 100. For this, the input unit 120 may include various keys such as a key button, a side key, a home key, and a touch key. The input unit 120 may generate an input event for entering the image recognition mode of the present invention. Alternatively, the input unit 120 may generate an input event for setting an image recognition mode according to a user operation. Meanwhile, the input unit 120 may generate an input signal for controlling the camera 110. For example, the input unit 120 may generate an input signal for requesting activation of the camera 110, an input signal for requesting a specific image capturing using the activated camera 110, and the like.

The display unit 140 may output various screens related to the operation of the image recognition apparatus 100. For example, the display unit 140 may output an idle screen, a menu screen, or the like of the image recognition apparatus 100. Meanwhile, the display unit 140 may output an icon or a menu item for entering the image recognition mode. The display unit 140 may output a preview image or a moving image obtained according to the camera 110 operation. In addition, the display unit 140 may output an image recognition result for the image obtained according to the image recognition mode. For example, the display unit 140 may display an object classification display that is compared with a stored image in an acquired image, and display information related to the object. Also, the display unit 140 may output a stored image as a result of image recognition of the acquired image. In addition, the display unit 140 may output a screen corresponding to a specific function performed corresponding to the image recognition result. For example, when the specific text information or the image information is included as the image recognition result of the acquired image, the display unit 140 may output the corresponding text and image information.

The storage unit 150 may store various programs for operating the image recognition apparatus 100 and data according to program operations. For example, the storage unit 150 may store an operating system or the like for operating the image recognition apparatus 100. In particular, the storage unit 150 may store an object recognition algorithm 151, an object recognition program 153, and an image database 155 for performing the image recognition mode of the present invention. The storage unit 150 may store the minutia matching result value and the color histogram similarity comparison value between the acquired image and the stored images. The feature point matching value and the color histogram similarity comparison value can be used to derive the image recognition result.

The object recognition algorithm 151 may include various algorithms for performing the image recognition mode of the present invention. For example, the object recognition algorithm 151 may include an algorithm for feature point extraction. In addition, the object recognition algorithm 151 may include a K-NN matching algorithm, a FLANN algorithm, a Local Sensitive Hashing (LSH) algorithm, or an ORSA (Optimized Random Sampling Algorithm). The object recognition algorithm 151 may be called during the operation of the object recognition program 153 and used to perform processes such as minutiae extraction, object extraction, and background removal in an image.

The object recognition program 153 may include various program routines for performing the image recognition mode of the present invention. For example, the object recognition program 153 may include a routine for activating the camera 110 when an event for activating the image recognition mode occurs, and a routine for performing feature point extraction on the image obtained from the camera 110. [ In addition, the object recognition program 153 may include a routine for performing feature point matching between the extracted image and the previously stored image, a routine for selecting a specific stored image having a good or best feature point matching result, And may include routines that remove the background of the image and perform object extraction. Then, the object recognition program 153 outputs the image recognition result based on at least one of the routine for comparing the color histogram of the object extracted from the acquired image and the color histogram of the selected image, the feature point matching result, and the similarity of the color histogram And < / RTI >

The image database 155 may include at least one stored image or an original image to be compared with an object included in the specific image. The image database 155 may include the result of applying the feature point extraction algorithm to the stored images. The image database 155 may also include color histogram values for stored images. For example, if the image database 155 includes three stored images, the image database 155 may store the feature point extraction values and the color histogram values for the respective stored images. The minutiae extraction values and the color histogram values stored in the image database 155 can be used in the comparison process with respect to the acquired image.

The control unit 160 may control various control signals and data processing for performing the image recognition mode of the present invention. For this, the controller 160 may include a configuration as shown in FIG.

2 is a detailed block diagram of a control unit of an image recognition apparatus according to an embodiment of the present invention.

2, the controller 160 may include a feature extraction unit 161, an image selection unit 163, an object division unit 165, and an image recognition unit 167.

The feature point extracting unit 161 can extract feature points in an image acquired by the camera 110. [ The feature point extracting unit 161 may extract the feature points of the stored image stored in the storage unit 150. For this, the feature point extraction unit 161 may use at least one algorithm stored in the object recognition algorithm 151. [ For example, the feature point extracting unit 161 may perform filtering based on Fourier transform or the like on an image to detect boundary line regions of the object, and extract at least a part of the detected boundary line regions as feature points. The acquired image including the minutiae points extracted by the minutiae point extracting unit 161 may be transmitted to the image selecting unit 163. [ The feature points extracted by the feature extraction unit 161 may be stored in the image database 155 in association with the stored images.

FIG. 3 is a view showing a result of extracting feature points of an acquired image and a stored image according to an embodiment of the present invention. The feature point extraction unit 161 may extract a plurality of feature points as shown in FIG. 3 through boundary line region detection, filtering, or the like. The feature point extraction of the stored image may be performed in the process of being stored in the image database 155 or when an image recognition process of the acquired image is requested. In order to improve the image processing speed, the feature point extraction of the stored images may be performed in advance in the process of configuring the image database 155.

The image selection unit 163 can select a storage image in the image database 155 having similar minutiae to the minutiae of the acquired image. For this, the image selector 163 may perform a matching process of the minutiae points between the acquired image and the stored images. That is, the image selector 163 can select a stored image having a distribution of the feature points included in the acquired image and a distribution of the feature points having the matching ratio equal to or greater than a predetermined value. Or a stored image having the largest number of matching points among the stored images may be selected. The image selection unit 163 temporarily stores the result of matching the feature point between the acquired image and the stored images in the storage unit 150 and may transmit the result to the image recognition unit 167. [

In this process, the image selector 163 may use a K-NN matching algorithm for calculating and ranking the vector distances of all the feature points included in the image. Alternatively, the image selector 163 may use an LSH algorithm and an FLANN scheme for grouping feature points having similar directional components to improve the image processing speed, and performing vector distance calculation within the grouped feature points. The threshold of the matching points for image selection may vary according to the designer's intention or the algorithm applied.

The object division unit 165 may divide the object by removing a specific object from the background in the acquired image. To this end, the object division unit 165 may use ORSA (Optimized Random Sampling Algorithm) to remove erroneous mapping points. 4 is a diagram for explaining object detection according to an embodiment of the present invention. The object division unit 165 applies the coordinates of the trained data image set (the object previously divided from the background) of the selected storage image shown in the area 401 of FIG. 4 to the acquired image to obtain the object of the acquired image shown in the area 403 of FIG. It is possible to estimate an outline point and use it to separate a specific object from the background. In this process, the object division unit 165 calculates the epipolar geometry between the acquired image and the selected storage image. The object division unit 165 sets points at random and estimates a homography matrix. The object divider 165 then calculates the quality of the homography matrix and selects the best value among them. The object division unit 165 may transmit information about the divided object to the image recognition unit 167. [

The image recognition unit 167 can perform image recognition based on a color histogram. The digital image can be represented by an 8-bit value divided into 0 to 255 values. Color images such as RGB (Red, Green, Blue) and HSV (Hue, Saturation, Value) require 24-bit space and 8-bit can be used for each channel. The above-mentioned color image space "d (Intensity)" can be defined as the following equation (1).

Where i is the serial number of the pixel and b _i is a value of 0 or 1. The bit-plane information can be composed of high and low, and the high-bit region can contain more accurate information about the acquired image than the low-bit region. Accordingly, the image recognition unit 167 of the present invention can process the values in the low-bit region to eliminate the noise. FIG. 5 is a diagram showing a result of extracting a significant color value in the HSV space. Referring to FIG. 5, it can be seen that distortion caused by light is reduced according to the result of extracting a significant color value of the left acquired image.

When the significant color extraction is completed, the image recognition unit 167 can create a color histogram of the divided object in the acquired image and a color histogram of the trained image stored in the image database 155. [ The color histogram can be defined as shown in the following equation (2).

Here, _Hc (i) is the color histogram of the object included in the specific image, N (I, C _i ) is the number of pixels in the image, C _i is the HSV color space, I is the value of C _i Can mean number. When the color histogram is detected, the image recognition unit 167 can perform similarity check of the detected color histogram data using Equation (3).

Here, H (I) is the color histogram data of the acquired image, and H (J) is the color histogram data of the stored image, for example, have. In particular, d (H (I), H (J)) may be the Euclidean distance of the color histogram of the acquired image and the color histogram of the stored image. According to the formula, the higher the value, the higher the probability.

When the similarity value of the color histogram is detected, the image recognition unit 167 can derive the image recognition result using the following Equation (4).

는 특징점들의 매칭 결과 값이다. 특징점들의 매칭 결과 값에서 분모에 위치한 max_of_vm_matches는 획득 영상과 저장 영상들 간 매칭 포인트가 가장 많이 검색된 수치를 의미할 수 있으며, vm-matches(I,J)는 선택 영상에서의 매칭 결과 값일 수 있다. 영상 인식부(167)는 특징점들의 매칭 결과 값과 컬러 히스토그램의 유사도 값의 합산을 통하여 획득 영상에 대한 영상 인식 결과를 도출할 수 있다. 즉 영상 인식부(167)는 획득 영상과 저장 영상들 간의 컬러 히스토그램 유사도가 가장 양호한 저장 영상과, 획득 영상과 저장 영상들 간의 특징점 매칭 결과 값 중 가장 양호한 매칭 결과 값을 가지는 저장 영상을 영상 인식 결과로서 도출할 수 있다. 이 과정에서 영상 인식부(167)는 특징점 매칭 결과 값을 이용하지 않고 컬러 히스토그램 유사도가 가장 양호한 저장 영상을 영상 인식 결과로서 도출할 수도 있다.Where d (H (I), H (J)) is the similarity value of the color histogram. Meanwhile,

Is the matching result value of the minutiae points. Max_of_vm_matches located in the denominator in the matching result value of the minutiae points may be a value in which the matching point is most searched between the acquired image and the stored images, and vm-matches (I, J) may be a matching result value in the selected image. The image recognition unit 167 can derive the image recognition result of the acquired image by summing the matching result value of the feature points and the similarity value of the color histogram. That is, the image recognition unit 167 recognizes the stored image having the best color histogram similarity between the acquired image and the stored images, and the stored image having the best matching result value among the acquired image matching results between the acquired image and the stored images, As shown in FIG. In this process, the image recognition unit 167 may derive the stored image having the best color histogram similarity without using the feature point matching result as the image recognition result.

FIG. 6 is a diagram illustrating a result of image matching of an acquired image through a feature point matching result and color histogram similarity summation according to an exemplary embodiment of the present invention. Referring to FIG. 6, the image recognition apparatus 100 can correctly detect a stored image matching the specific object included in the acquired image in the image database 155.

On the other hand, the image recognition unit 167 can support performing the predetermined function when the object recognition of the acquired image is completed. For example, the image recognition unit 167 may support outputting the object recognition result to a predetermined area of the display unit 140. [ At this time, the image recognizing unit 167 can output the corresponding information if at least a part of the specific text information and the image information is included or linked to the stored image. Alternatively, the image recognition unit 167 may perform a specific server connection according to the image recognition result. The mobile terminal may further include a communication module for forming a communication channel for connecting to the server.

As described above, in the image recognition apparatus 100 according to the embodiment of the present invention, in the image recognition process for an acquired image from which feature points are extracted, You can select images. Accordingly, the image recognition apparatus 100 of the present invention can provide a more reliable image recognition result by minimizing the distortion caused by light, and it is possible to compare all the feature points by ignoring the feature points having a relatively low bit value It is possible to support a relatively faster image processing speed. Also, the present invention can support a faster image processing speed by applying a method of performing vector distance calculation processing of feature points having similar directional components in the feature point extraction and matching process.

7 is a view for explaining an image recognition method according to an embodiment of the present invention.

Referring to FIG. 7, in step 701, the controller 160 determines whether the image recognition mode is entered. For entering the image recognition mode, the image recognition apparatus 100 may output at least one of an icon and a menu item for executing the image recognition mode. If the image recognition mode is set to be automatically entered when the camera 110 is activated, step 701 may be replaced with checking whether the camera 110 is activated.

If the input event is not an input event for entering the image recognition mode in step 701, the control unit 160 may control the execution of the corresponding function according to the input event in step 703. For example, the control unit 160 may play back the file stored in the storage unit 150 or support the communication service according to the input event type. Alternatively, the control unit 160 may receive and store the stored image from the server apparatus that provides the stored image according to the user operation or the set schedule information.

On the other hand, if there is a setting for entering the image recognition mode or an input event occurs in step 701, the controller 160 branches to step 705 to check whether the image is collected. If the image collection is performed in step 705, the controller 160 may perform the feature point extraction in step 707. FIG. The image collection at this stage may include various image acquisition processes such as a preview image, a moving image, and a still image. The image collection process may also include a process of receiving an image from another external electronic device.

In step 709, the controller 160 may perform image selection. In this process, the controller 160 can select a stored image having matching points of a predetermined reference or more with the minutiae points of the specific object included in the acquired image, using the specific algorithm included in the object recognition algorithm 151. [ For example, the control unit 160 compares the acquired image and the matching result of the image stored in the image database 155 using the Local Sensitive Hashing (LSH) and the FLANN algorithm to perform image selection with the highest feature point matching result can do. For this, the controller 160 may calculate the matching point between the acquired image and the stored image, and compare the result of the matching point. The controller 160 may store the matching point result of the selected stored image and the matching point result of all the stored images. The stored matching result values may be used to derive an image recognition result.

Meanwhile, the controller 160 may perform object background division in step 711. FIG. In this process, the control unit 160 estimates the epipolar geometry between two images using ORSA (Optimized Random Sampling Algorithm), and detects four outline points of the object based on the epipolar geometry. At this time, the controller 160 provides coordinates of the trained data set image of the selected image in the image database 155 (pre-divided object from the background) to the acquired image, Point can be detected.

If object detection is performed, the controller 160 may control the image recognition and the performance of the function using the color histogram in step 713. In more detail, the controller 160 may calculate a color histogram of the acquired image and a color histogram of the stored images prestored in the image database 155. In this process, the controller 160 may calculate the color histogram of the divided object of the acquired image. In order to calculate the color histogram, the controller 160 calculates the pixel values of the images and constructs bit-planes. Then, the controller 160 selects a pixel value (High-bit) Can be calculated. That is, the control unit 160 may prepare a color histogram based on a high bit value, which is relatively robust in light distortion, by ignoring bit values below a predetermined reference level, where distortion is relatively high with respect to light.

When the color histogram calculation is completed, the controller 160 may calculate a similarity value between the color histogram of the divided object and the color histogram of the stored images in the acquired image. The controller 160 may sum up the matching result values of the feature points of the acquired image and the stored images and the color histogram similarity value to derive the stored image having the best value as the image recognition result of the acquired image. The control unit 160 may control the performance of the function according to the predefined schedule information when the image recognition result is derived. For example, the control unit 160 can perform classification display of the detected object in the acquired image, display information of the stored image recognized for the object, reproduce the file according to the stored image recognition, or access the server device. In addition, the controller 160 may perform an object tracking function for the object recognized in the acquired image.

In step 715, the controller 160 determines whether there is an input event for ending the function. If there is no input event related to function termination, the control unit 160 branches to the step before step S701 and can support the following process again.

As described above, the image recognition apparatus and method according to the embodiment of the present invention minimizes image recognition problems due to optical distortion by using the result of the feature point matching with the stored images and the similarity of the color histogram for object recognition in the acquired image can do.

In the above description, the minutia matching result and the similarity value of the color histogram are used to derive the recognition result for the stored image, but the present invention is not limited thereto. That is, the present invention can perform the storage image selection for the acquired image by only comparing the similarity value of the color histogram, and may also apply the result of matching the feature point.

While the present invention has been described with reference to several preferred embodiments, these embodiments are illustrative and not restrictive. It will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit of the invention and the scope of the appended claims.

100: image recognition device 110: camera
120: input unit 140:
150: storage unit 160:

Claims (15)

A camera for performing image acquisition;
A storage unit for storing stored images for image recognition of the acquired image;
A controller for performing image recognition on the acquired image based on a color histogram of the acquired image and a comparison of color histogram similarities of the stored images; Lt; / RTI >
Wherein,
Extracting feature points from the acquired image, performing feature point matching with the stored images, selecting a stored image having a matching ratio of a predetermined ratio or more, or a stored image having the best matching value, A specific object in an acquired image is divided,
Wherein,
And a storage image having the best combination of the feature point matching value and the color histogram similarity is derived as an image recognition result. delete The method according to claim 1,
The control unit
Wherein an object is divided by providing coordinates of the selected storage image to the acquired image by using ORSA (Optimized Random Sampling Algorithm). The method according to claim 1,
The control unit
Wherein the comparison unit compares the color histogram of the specific object with the color histogram similarity of the stored image in the acquired image. delete The method according to claim 1,
The control unit
And performs feature point matching between the acquired image and the stored image based on the K-NN algorithm. The method according to claim 1,
The control unit
Wherein the feature point matching is performed between the acquired image and the stored image based on LSH (Local Sensitive Hashing) and FLANN algorithm. The method according to claim 1,
The control unit
And calculates a color histogram by ignoring a low bit value among pixel values of the acquired image. The method according to claim 1,
The control unit
And displaying the stored image corresponding to the acquired image, displaying information related to the image-recognized stored image, accessing a server device related to the image-recognized stored image, And controlling at least one of the specific object tracking in the acquired image. A process of acquiring an image;
Extracting feature points of the acquired image;
A feature point matching step of performing feature point matching of feature points of the acquired image and stored images stored in advance;
Selecting a stored image in which the feature point matching is a matching value or a matching ratio or a best matching value;
A dividing step of dividing a specific object of the acquired image using the stored image;
A color histogram calculating step of detecting a color histogram of the divided object in the acquired image;
Calculating a degree of similarity between a color histogram of the object of the acquired image and a color histogram of the stored image;
A derivation step of deriving an image recognition result for the acquired image based on the similarity value of the color histogram; / RTI >
In the derivation process,
Wherein the step of deriving a storage image having the best combination of the color histogram similarity and the feature matching value as an image recognition result. 11. The method of claim 10,
The feature point matching process
Performing feature point matching between the acquired image and the stored image based on a K-NN algorithm;
Performing feature point matching between the acquired image and the stored image based on LSH (Local Sensitive Hashing) and FLANN algorithm;
The method comprising the steps of: 11. The method of claim 10,
The object segmentation process
Dividing an object by providing coordinates of the selected storage image to the acquired image using ORSA (Optimized Random Sampling Algorithm);
And an image processing unit for processing the image. 11. The method of claim 10,
The color histogram calculation process
Dividing a high bit and a low bit by a predefined constant reference for each pixel value represented by a certain bit in an object region of the acquired image and calculating a color histogram by selecting only a high bit;
And an image processing unit for processing the image. delete 11. The method of claim 10,
After the derivation process,
Dividing and displaying a specific object in the acquired image according to the image recognition result;
Displaying a stored image corresponding to the acquired image;
Displaying information related to the image that is recognized as an image;
Performing a server device connection with respect to the stored video image;
Tracking a specific object in the acquired image;
The method of claim 1, further comprising:

Images