KR20210015697A - Method and apparatus for object re-identification - Google Patents

Abstract

The present invention provides a method and an apparatus for object re-recognition which can detect and match objects from various images photographed from various viewpoints. According to one embodiment of the present invention, the method for object re-recognition performed by an apparatus for object re-recognition comprises: a step of detecting objects in a plurality of images; a step of inferring object information including a property for the detect objects; a step of selecting objects having the same property as an object to be recognized in the inferred object information as objects to be compared; a step of inferring a photographing viewpoint angle of the selected objects to be compared; a step of selecting a recognition candidate object among the objects to be compared in accordance with whether the inferred photographing viewpoint angle is included in a preset angle range in response to the object to be recognized; and a step of identifying whether the selected recognition candidate object matches the object to be recognized.

Description

Object re-recognition device and method {METHOD AND APPARATUS FOR OBJECT RE-IDENTIFICATION}

The present invention relates to an apparatus for re-recognizing an object and a method for re-recognizing an object.

Smart city applications have emerged as a way to solve various problems in the modern society, and among smart city applications, the demand for surveillance systems occupies a large proportion. In order to provide a video surveillance service in such a surveillance system, it is necessary to obtain information on each object in the collected video and track a target object. For this, re-recognition of an object to be found by matching a target object in an image is required.

On the other hand, in terms of information sharing, crowdsourcing can receive data from multiple participants and share the reported data. Therefore, crowdsourcing can utilize a wide range of data in various situations.

When this crowdsourcing is applied to a surveillance system, data can be reported from participants with mobility. Therefore, there is no limit to the range that can be analyzed without the need to install infrastructure such as surveillance cameras. In addition, since various images captured at various viewpoints can be secured, objects that are covered by other objects at a specific photographing time can also be checked in the tip image of other participants.

However, when a crowdsourcing environment is used for a surveillance system, objects must be detected and re-recognized from various images taken from various viewpoints.

US registered patent US 10,127,668. Registration date November 13, 2018.

According to an embodiment, an object re-recognition method and apparatus capable of detecting and matching objects from various images captured at various viewpoints are provided.

The problem to be solved of the present invention is not limited to those mentioned above, and another problem to be solved that is not mentioned will be clearly understood by those of ordinary skill in the art from the following description.

The object re-recognition method performed by the object re-recognition apparatus according to the first aspect includes the steps of detecting an object within a plurality of images, inferring object information including attributes for the detected object, and the inferred Selecting an object having the same property as the object to be recognized from among object information as a target object to be compared, inferring a shooting angle of the selected target object to be compared, and the inferred shooting angle to the object to be recognized And selecting a recognition candidate object from among the objects to be compared according to whether the corresponding object falls within a predetermined angular range, and determining whether the selected recognition candidate object matches the recognition target object.

An object re-recognition apparatus according to a second aspect includes an input unit for receiving a plurality of images, a processor unit for processing the images, and an output unit for outputting a result of processing by the processor unit, and the processor unit , Detecting an object in the plurality of images input to the input unit, inferring object information including attributes for the detected object, and comparing objects having the same attributes as the object to be recognized among the inferred object information Selected as a target object, infers the shooting angle of the selected comparison target object, and a recognition candidate among the comparison target objects according to whether the inferred shooting viewpoint angle falls within a preset angle range corresponding to the recognition target object An object is selected, and it is determined whether the selected recognition candidate object matches the recognition target object.

A computer-readable recording medium storing a computer program according to the third aspect, when the computer program is executed by a processor, includes the steps of detecting an object in a plurality of images, and attributes of the detected object. Inferring the object information to be compared, selecting an object having the same property as the object to be recognized among the inferred object information as a target object to be compared, and inferring an angle of a photographing point of the selected object to be compared; Selecting a recognition candidate object among the comparison target objects according to whether the inferred photographing viewpoint angle falls within a preset angle range corresponding to the recognition target object, and the selected recognition candidate object matches the recognition target object And instructions for causing the processor to perform a method including determining whether or not.

According to the fourth aspect, a computer program stored in a computer-readable recording medium, when executed by a processor, detects an object in a plurality of images, and infers object information including attributes for the detected object. And selecting an object having the same property as the object to be recognized among the inferred object information as a target object to be compared; inferring a photographing angle of the selected object to be compared; and the inferred photographing time Selecting a recognition candidate object among the comparison target objects according to whether an angle is included in a preset angle range corresponding to the recognition target object, and determining whether the selected recognition candidate object matches the recognition target object. And instructions for causing the processor to perform the containing method.

According to an embodiment, an object may be detected and re-recognized from various images captured at various viewpoints. For example, when crowdsourcing is applied to a monitoring system to provide an image monitoring service, an object re-recognition may be performed based on the inferred photographing viewpoint angle after inferring a photographing viewpoint angle. Therefore, since it is possible to consider the spatiotemporal change according to the mobility of the participant that may occur in the crowdsourcing environment, there is an effect of improving the re-recognition performance of the object.

1 is a block diagram of a monitoring system to which an object re-recognition device according to an embodiment of the present invention can be applied.
2 is a block diagram of an object re-recognition apparatus according to an embodiment of the present invention.
3 is a block diagram of a processor unit illustrated in FIG. 2.
4 to 6 are flowcharts illustrating an object re-recognition method performed by an object re-recognition apparatus according to an embodiment of the present invention.

Advantages and features of the present invention, and a method of achieving them will become apparent with reference to the embodiments described later together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in a variety of different forms, and only these embodiments make the disclosure of the present invention complete, and are common knowledge in the technical field to which the present invention pertains. It is provided to completely inform the scope of the invention to those who have, and the invention is only defined by the scope of the claims.

The terms used in the present specification will be briefly described, and the present invention will be described in detail.

The terms used in the present invention have been selected from general terms that are currently widely used while considering functions in the present invention, but this may vary depending on the intention or precedent of a technician working in the field, the emergence of new technologies, and the like. In addition, in certain cases, there are terms arbitrarily selected by the applicant, and in this case, the meaning of the terms will be described in detail in the description of the corresponding invention. Therefore, the terms used in the present invention should be defined based on the meaning of the term and the overall contents of the present invention, not a simple name of the term.

When it is said that a part in the specification'includes' a certain element, this means that other elements may be further included instead of excluding other elements unless specifically stated to the contrary.

In addition, the term'unit' used in the specification means software or hardware components such as FPGA or ASIC, and'unit' performs certain roles. However,'part' is not limited to software or hardware. The'unit' may be configured to be in an addressable storage medium, or may be configured to reproduce one or more processors. Thus, as an example,'unit' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, procedures, Includes subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, database, data structures, tables, arrays and variables. The functions provided in the components and'units' may be combined into a smaller number of components and'units', or may be further divided into additional components and'units'.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art may easily implement the present invention. In addition, in the drawings, parts not related to the description are omitted in order to clearly describe the present invention.

1 is a block diagram of a monitoring system to which an object re-recognition device according to an embodiment of the present invention can be applied.

Referring to FIG. 1, the monitoring system 1 includes an object re-recognition device 100, and the object re-recognition device 100 is connected to a communication network 10 to operate in a crowdsourcing environment. For example, the object re-recognition apparatus 100 may interwork with an edge server or a crowd server, or may be included in an edge server or a crowd server. In addition, the operation environment of the object re-recognition apparatus 100 is not particularly limited, and may operate in various environments in which an image including an object to be recognized may be provided from a plurality of terminals.

The object re-recognition apparatus 100 may receive various images captured at various viewpoints through the communication network 10 and detect and match objects included in the received images.

2 is a block diagram of an object re-recognition apparatus 100 according to an embodiment of the present invention.

Referring to FIG. 2, the object re-recognition apparatus 100 includes an input unit 110 and a processor unit 120. In addition, the object re-recognition device 100 may further include an output unit 130 and/or a storage unit 140.

The input unit 110 receives a plurality of images that may include an object that may be selected as a recognition candidate object, and provides the received plurality of images to the processor unit 120. For example, the input unit 110 may include a communication module capable of receiving image data through the communication network 10 of FIG. 1 or may include an interface capable of directly receiving image data.

The processor unit 120 may process the image provided from the input unit 110 and control the output unit 130 to output a result of the processing.

The processor unit 120 detects an object in a plurality of images input to the input unit 110 and infers object information including attributes for the detected object. In addition, the processor unit 120 selects an object having the same properties as the object to be recognized among the inferred object information as the object to be compared, and infers the angle of the photographing point of the selected object to be compared. In addition, the processor unit 120 selects a recognition candidate object from among the comparison target objects according to whether the inferred shooting point angle is included in a preset angle range corresponding to the recognition target object. In addition, the processor unit 120 determines whether the selected recognition candidate object matches the recognition target object.

Here, the processor 120 may infer a photographing viewpoint angle based on a result of comparing the selected object to be compared with reference shape information preset for each attribute of the object.

The processor unit 102 may use a deep learning model based on a convolutional neural network when inferring object information or when selecting an object to be compared.

The processor unit 120 adds a fully connected layer to the last layer of the deep learning model based on the convolutional neural network when inferring the shooting angle, and then inputs the inferred object information into the fully coupled layer. Thus, it is possible to obtain a photographing viewpoint angle as an output of the pre-combined layer.

When selecting a recognition candidate object, the processor unit 120 classifies a region of interest (ROI) based on a preset angular range with respect to the image of the object to be compared, and can express it using feature values extracted from the classified region of interest. A plurality of recognition candidate objects may be selected based on the feature vector. Here, the feature vector may be expressed as a feature vector by performing pixel-based feature value extraction and convolutional-based feature value extraction for the classified ROI, and then performing reconstruction fixing to a dimension of a specific size. For example, the convolutional-based feature value may use an intermediate convolutional layer output matrix of a deep learning model based on a convolutional neural network.

When determining whether the selected plurality of recognition candidate objects match the recognition target object, the processor unit 120 performs clustering on the selected plurality of recognition candidate objects based on their attributes, and performs a recognition target object for each clustered cluster. A distance average is calculated by calculating an Euclidean distance with and, and a recognition candidate object belonging to a cluster having the smallest calculated distance average may be recognized as an object matching the object to be recognized. Here, "attribute" may be a feature vector that can be expressed using feature values extracted from the region of interest.

The output unit 130 may output a result of processing by the processor unit 120 according to the control of the processor unit 120. For example, the output unit 130 may include a communication module capable of transmitting processing result data by the processor unit 120 or an interface capable of transmitting processing result data to another electronic device. In addition, the output unit 130 may include a display device or a printing device capable of visually identifying the result data of processing by the processor unit 120.

The storage unit 140 may store a result of processing by the processor unit 120 under the control of the processor unit 120. For example, the storage unit 140 includes magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, and floptical disks. It may be a computer-readable recording medium such as a hardware device specially configured to store and execute program instructions such as magneto-optical media or flash memory.

3 is a block diagram of a processor unit illustrated in FIG. 2.

Referring to FIG. 3, the processor unit 120 includes an object detection unit 121, an information inference unit 122, an attribute selection unit 123, an angle inference unit 124, a candidate selection unit 125, and an object matching unit ( 126).

The object detection unit 121 detects an object in a plurality of input images.

The information inferring unit 122 infers object information including attributes for the object detected by the object detecting unit 121.

The attribute selection unit 123 selects an object having the same attribute as the object to be recognized among the object information inferred by the information inference unit 122 as the object to be compared.

The angle inference unit 124 infers the photographing viewpoint angle of the object to be compared selected by the attribute selection unit 123.

The candidate selection unit 125 selects a recognition candidate object from among the comparison target objects according to whether the shooting point angle inferred by the angle inference unit 124 is included in a preset angle range corresponding to the recognition target object.

The object matching unit 126 determines whether the recognition candidate object selected by the candidate selection unit 125 matches the recognition target object.

4 to 6 are flowcharts illustrating an object re-recognition method performed by an object re-recognition apparatus according to an embodiment of the present invention.

Hereinafter, a method of re-recognizing an object in a monitoring system by an object re-recognition apparatus according to an embodiment of the present invention will be described in detail with reference to FIGS. 1 to 6.

First, the input unit 110 of the object re-recognition apparatus 100 acquires a plurality of images that may include an object that may be selected as a recognition candidate object through the communication network 10 and provides them to the processor unit 120. For example, crowdsourcing participants can shoot various images at various viewpoints using a communication device equipped with a camera, such as a smartphone, and the object re-recognition device 100 through the communication network 10 to capture various images from various viewpoints. ) Can be uploaded.

Then, the object detection unit 121 of the processor unit 120 detects an object in a plurality of images input through the input unit 110 (S410), and the information inference unit 122 of the processor unit 120 is an object detection unit. Object information including attributes for the object detected by (121) is inferred (S420).

Here, when the information inference unit 122 infers object information, a deep learning model based on a convolutional neural network may be used. For example, a plurality of image data by the input unit 110 may be input to a pre-learned deep learning model, and such a deep learning model includes attributes (c) of each object detected from each image data, and Object information such as center coordinates (cx, cy), width (w) and height (h) of the boundary area may be inferred and output. For example, the attribute (c) of an object may be information indicating whether it is a car, a person, or an object.

Subsequently, the attribute selection unit 123 of the processor unit 120 selects an object having the same attribute as the object to be recognized among the object information inferred by the information inference unit 122 as the object to be compared (S430). That is, the attribute selection unit 123 may select only an object having the same attribute as the object to be recognized among the objects detected by the object detection unit 121 as the object to be compared.

Next, the angle inferring unit 124 of the processor unit 120 infers the shooting point angle of the object to be compared selected by the attribute selection unit 123. Here, the photographing viewpoint angle may be inferred based on a result of comparing the reference shape information preset for each object property and the selected object to be compared. The photographing viewpoint angle may include an azimuth angle, an upward angle, a plane rotation angle, and the like, and among them, the azimuth angle may be used as a representative value of the photographing viewpoint angle. The azimuth angle is the coordinate representing the position of the object in the horizontal coordinate, the upward angle is the angle between the imaginary line connecting the object and the shooting point and the surface, and the plane rotation angle is the left and right angle on the surface. For example, the angle inference unit 124 adds a fully connected layer to the last layer of the deep learning model based on the convolutional neural network when inferring the angle of the photographing point of view, and then the information inference unit 122 The object information inferred by) may be input to the pre-combined layer to obtain a photographing viewpoint angle as an output of the pre-combined layer (S440).

In addition, the candidate selection unit 125 of the processor unit 120 is a recognition candidate among the object to be compared according to whether the angle of the shooting point inferred by the angle inference unit 124 is included in a preset angle range corresponding to the object to be recognized An object is selected (S450).

In step S450, when selecting a recognition candidate object, the candidate selection unit 125 classifies the ROI based on a preset angle range with respect to the image of the object to be compared (S451), and from the classified ROI. A recognition candidate object may be selected based on a feature vector that can be expressed by using the extracted feature value (S452). For example, the angular range for classification of the region of interest is defined to have four regions of interest such as front, front side, side, rear side, rear, etc. You can, and you can set each angle range according to the characteristics of each object.

Here, the feature vector may be expressed as a feature vector by performing pixel-based feature value extraction and convolutional-based feature value extraction for the classified ROI, and then performing reconstruction fixing to a dimension of a specific size. For example, in order to extract a feature value in units of pixels, a Scale Invariant Feature Transform (SIFT) technique may be applied. In addition, as the convolutional-based feature value, an intermediate convolutional layer output matrix of a deep learning model based on a convolutional neural network may be used. Here, in the case of extracting a pixel-by-pixel feature value using the SIFT technique, since the dimensional size of the output feature value changes according to the input, it is necessary to fix it to a dimension of a specific size. For example, by applying VLAD (Vector of Locally Aggregated Descriptors) pooling to the extracted feature values for each pixel, it may be reconstructed into a dimension of a specific size. And, in order to match the dimensional size with the convolutional-based feature value, it can be reduced to the same dimensional size through Principal Component Analysis (PCA). Thereafter, a feature vector may be finally obtained by combining the pixel-based feature value and the convolutional feature value.

Next, the object matching unit 126 of the processor unit 120 determines whether the recognition candidate object selected by the candidate selection unit 125 matches the recognition target object (S460).

In step S460, when determining whether the selected recognition candidate object matches the recognition target object, the object matching unit 126 may perform clustering on the selected recognition candidate object based on the attribute. For example, K-means clustering may be used (S461).

For each clustered cluster, a distance average is calculated by calculating an Euclidean distance from the object to be recognized (S462), and a recognition candidate object belonging to the cluster having the smallest calculated distance average is identified as the object to be recognized. It can be recognized as an object matching with (S463).

Meanwhile, a result of the processing by the processor unit 120, that is, information on the recognition candidate object determined to match the recognition target object is output by the output unit 130 under the control of the processor unit 120. For example, the output unit 130 may transmit result data of processing by the processor unit 120 through a communication module or may transmit result data of the processing to another electronic device through an interface. Alternatively, the output unit 130 may output data resulting from the processing by the processor unit 120 to enable visual identification through a display device or a printing device.

Also, the storage unit 140 may store a result of processing by the processor unit 120 under the control of the processor unit 120.

Each step included in the object re-recognition method according to the above-described embodiment may be implemented in a computer-readable recording medium that records a computer program including instructions for performing these steps.

In addition, each step included in the object re-recognition method according to the above-described embodiment may be implemented in the form of a computer program stored in a computer-readable recording medium programmed to include an instruction for performing such a step.

As described so far, according to an exemplary embodiment of the present invention, objects can be detected and re-recognized from various images captured at various viewpoints. For example, when crowdsourcing is applied to a monitoring system to provide an image monitoring service, an object recognition may be performed based on the inferred photographing viewpoint angle after inferring a photographing viewpoint angle. Therefore, since it is possible to consider the spatiotemporal change according to the mobility of the participant that may occur in the crowdsourcing environment, there is an effect of improving the recognition performance of the object.

Combinations of each step of each flowchart attached to the present invention may be performed by computer program instructions. Since these computer program instructions can be mounted on the processor of a general purpose computer, special purpose computer or other programmable data processing equipment, the instructions executed by the processor of the computer or other programmable data processing equipment are the functions described in each step of the flowchart. Will create a means of doing things. These computer program instructions can also be stored on a computer-usable or computer-readable recording medium that can be directed to a computer or other programmable data processing equipment to implement a function in a specific manner, so that the computer-readable or computer-readable medium. It is also possible to produce an article of manufacture that includes instruction means for performing the functions described in each step of the flow chart with instructions stored on the recording medium. Computer program instructions can also be mounted on a computer or other programmable data processing equipment, so that a series of operating steps are performed on a computer or other programmable data processing equipment to create a computer-executable process to create a computer or other programmable data processing equipment. It is also possible for instructions to perform processing equipment to provide steps for executing the functions described in each step of the flowchart.

In addition, each step may represent a module, segment, or part of code that contains one or more executable instructions for executing the specified logical function(s). Further, it should be noted that in some alternative embodiments, the functions mentioned in the steps may occur out of order. For example, two steps shown in succession may in fact be performed substantially simultaneously, or the steps may sometimes be performed in the reverse order depending on the corresponding function.

The above description is merely illustrative of the technical idea of the present invention, and those of ordinary skill in the art to which the present invention pertains will be able to make various modifications and variations without departing from the essential quality of the present invention. Accordingly, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain the technical idea, and the scope of the technical idea of the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present invention.

1: Surveillance System 100: Object Recognition Device
110: input unit 120: processor unit
130: output unit 140: storage unit

Claims (18)

As an object re-recognition method performed by an object re-recognition device,
Detecting an object within a plurality of images; and
Inferring object information including attributes for the detected object; and
Selecting an object having the same property as the object to be recognized among the inferred object information as a target object to be compared,
Inferring a photographing point angle of the selected object to be compared;
Selecting a recognition candidate object among the comparison target objects according to whether the inferred photographing viewpoint angle is included in a preset angle range corresponding to the recognition target object; and
And determining whether the selected recognition candidate object matches the recognition target object.
Object re-recognition method. The method of claim 1,
The photographing viewpoint angle is inferred based on a result of comparing the selected object to be compared with reference shape information preset for each object property.
Object re-recognition method. The method of claim 1,
The step of detecting the object,
Acquiring the plurality of images through crowdsourcing
Object re-recognition method. The method of claim 3,
The step of inferring the angle of the photographing viewpoint,
After adding a fully connected layer to the last layer of a deep learning model based on a convolutional neural network, the inferred object information is input to the pre-combined layer, and the photographing time as an output of the pre-combined layer To acquire the angle
Object re-recognition method. The method of claim 1,
The step of selecting the recognition candidate object,
Classifying a region of interest (ROI) based on the preset angle range with respect to the image of the object to be compared;
And selecting the recognition candidate object based on a feature vector that can be expressed using feature values extracted from the classified region of interest.
Object re-recognition method. The method of claim 5,
The feature vector is expressed as the feature vector by performing pixel-by-pixel feature value extraction and convolutional-based feature value extraction for the classified ROI, and then performing reconstruction fixing to a dimension of a specific size.
Object re-recognition method. The method of claim 6,
The convolutional-based feature value is
Using the intermediate convolutional layer output matrix of a deep learning model based on a convolutional neural network
Object re-recognition method. The method of claim 1,
The step of determining whether the match is performed,
Clustering the selected recognition candidate object based on its attribute,
For each of the clustered clusters, calculating a distance average by calculating an Euclidean distance with the object to be recognized;
Recognizing a recognition candidate object belonging to a cluster having the smallest calculated distance average as an object matching the recognition target object
Object re-recognition method. An input unit that receives a plurality of images;
A processor unit that processes the image;
And an output unit for outputting a result of the processing by the processor unit,
The processor unit,
Detecting an object in the plurality of images input to the input unit,
Infer object information including attributes for the detected object,
Among the inferred object information, an object having the same property as the object to be recognized is selected as the object to be compared,
Infer the shooting angle of the selected object to be compared,
Selecting a recognition candidate object among the comparison target objects according to whether the inferred photographing viewpoint angle is included in a preset angle range corresponding to the recognition target object,
To determine whether the selected recognition candidate object matches the recognition target object
Object re-recognition device. The method of claim 9,
The photographing viewpoint angle is inferred based on a result of comparing the selected object to be compared with reference shape information preset for each object property.
Object re-recognition device. The method of claim 9,
The input unit,
Acquiring the plurality of images through crowdsourcing
Object re-recognition device. The method of claim 9,
The processor unit,
When inferring the shooting angle,
After adding a fully connected layer to the last layer of a deep learning model based on a convolutional neural network, the inferred object information is input to the pre-combined layer, and the photographing time as an output of the pre-combined layer To acquire the angle
Object re-recognition device. The method of claim 12,
The processor unit,
When selecting the recognition candidate object,
Classify a region of interest (ROI) based on the preset angle range with respect to the image of the object to be compared,
Selecting the recognition candidate object based on a feature vector that can be expressed using feature values extracted from the classified ROI
Object re-recognition device. The method of claim 13,
The feature vector is expressed as the feature vector by performing pixel-by-pixel feature value extraction and convolutional-based feature value extraction for the classified ROI, and then performing reconstruction fixing to a dimension of a specific size.
Object re-recognition device. The method of claim 14,
The convolutional-based feature value,
Using the intermediate convolutional layer output matrix of a deep learning model based on a convolutional neural network
Object re-recognition device. The method of claim 9,
The processor unit,
When determining whether the above matches,
Clustering is performed on the selected recognition candidate object based on its attribute,
For each clustered cluster, a distance average is calculated by calculating an Euclidean distance to the object to be recognized,
Recognizing a recognition candidate object belonging to a cluster having the smallest calculated distance average as an object matching the recognition target object
Object re-recognition device. As a computer-readable recording medium storing a computer program,
The computer program, when executed by a processor,
Detecting an object within a plurality of images, inferring object information including attributes for the detected object, and selecting an object having the same attribute as the object to be recognized among the inferred object information as a comparison target object Recognizing among the comparison target objects according to the step of selecting, inferring the shooting angle of the selected comparison target object, and whether the inferred shooting viewing angle falls within a preset angle range corresponding to the recognition target object A computer-readable recording medium comprising instructions for causing the processor to perform a method including selecting a candidate object and determining whether the selected recognition candidate object matches the recognition target object. As a computer program stored in a computer-readable recording medium,
The computer program, when executed by a processor,
Detecting an object within a plurality of images, inferring object information including attributes for the detected object, and selecting an object having the same attribute as the object to be recognized among the inferred object information as a comparison target object Recognizing among the comparison target objects according to the step of selecting, inferring the shooting angle of the selected comparison target object, and whether the inferred shooting viewing angle falls within a preset angle range corresponding to the recognition target object A computer program comprising instructions for causing the processor to perform a method including selecting a candidate object and determining whether the selected recognition candidate object matches the recognition target object.

Images