KR102060370B1 - System for recognizing and tracking an object based on divisional feature information and method for recognizing and tracking using the same - Google Patents

Abstract

In the partition appearance-based object recognition and tracking system and the object recognition and tracking method using the same, the object recognition and tracking system includes a plurality of tracking units, a plurality of storage units, a discriminating unit, and an integrated unit. The tracking units track a plurality of objects in each tracking area, and extract location and appearance information of each object. The storage units store location and appearance information of the objects tracked by the tracking units, respectively. When the object is destroyed in any one of the tracking units and the new object is tracked in any one of the tracking units, the determination unit determines the similarity between the destroyed object and the new object. The integration unit allocates the new object as the destroyed object when it is determined that the new object is similar to the destroyed object, and integrates the tracked objects by allocating the new object as a new object when it is determined that the new object is similar to the destroyed object.

Description

Object Recognition and Tracking System based on Segmental Appearance Information and Object Recognition and Tracking Method Using the System {SYSTEM FOR RECOGNIZING AND TRACKING AN OBJECT BASED ON DIVISIONAL FEATURE INFORMATION AND METHOD FOR RECOGNIZING AND TRACKING USING THE SAME}

The present invention relates to an object recognition and tracking system and an object recognition and tracking method using the same. More particularly, the present invention uses segmented appearance information of the object to recognize or track multiple objects such as a plurality of people through multiple cameras. The present invention relates to a segmented appearance information based object recognition and tracking system and an object recognition and tracking method using the same.

As the recognition technology for objects such as people and objects is improved, and the necessity of tracking objects occurs in various situations, many technologies have recently been developed for technologies for recognizing objects or tracking objects based on the objects. .

For example, Korean Patent Laid-Open Publication No. 10-2008-0057500 discloses a technique for extracting and tracking a motion region by comparing a pixel of a current frame with a previous frame that is input as related to a people motion tracking system and method thereof. In addition, the Republic of Korea Patent Publication No. 10-2012-0072009 relates to a multi-user interaction recognition apparatus and method through asynchronous image processing asynchronously recognizes the face of the multi-user through the multi-calculation and gesture point of the user through the tracking It discloses a technique for recognizing this.

As described above, a number of techniques for recognizing an object from an image captured to date, tracking the object, and identifying the characteristics of the movement of the object have been developed.

However, in the prior art, tracking of the movement of a plurality of objects often fails to track the corresponding object, and when newly tracking the same object, it is not possible to determine whether the same object is the new object. There is a problem to be recognized as.

This occurs even when an object is tracked on one screen through one camera, but especially when tracking an object through multiple cameras or when the object tracking is complicated due to various changes in the surrounding environment. This can lead to problems that cannot be tracked continuously.

Republic of Korea Patent Publication No. 10-2008-0057500 Republic of Korea Patent Publication No. 10-2012-0072009

Accordingly, the technical problem of the present invention has been conceived in this respect, the object of the present invention can easily determine whether the same object for the newly appeared object, in particular, newly appeared on all screens of the multi-camera or to various surrounding environments The present invention relates to a segmented appearance information-based object recognition and tracking system capable of object recognition and tracking even in difficult situation of object recognition.

Another object of the present invention relates to an object recognition and tracking method using the object recognition and tracking system.

Object recognition and tracking system according to an embodiment for realizing the above object of the present invention includes a plurality of tracking units, a plurality of storage units, determination unit and integration unit. The tracking units track a plurality of objects in each tracking area, and extract location and appearance information of each object. The storage units store location and appearance information of the objects tracked by the tracking units, respectively. When the object is destroyed in any one of the tracking units and the new object is tracked in any one of the tracking units, the determination unit determines the similarity between the destroyed object and the new object. The integration unit allocates the new object as the destroyed object when it is determined that the new object is similar to the destroyed object, and integrates the tracked objects by allocating the new object as a new object when it is determined that the new object is similar to the destroyed object.

In one embodiment, each of the storage unit, when the objects are destroyed from the tracking unit, may store the location and appearance information of the object immediately before the destruction.

In one embodiment, each of the objects is a person, the appearance information can be extracted and stored by dividing the person into a head, upper body and lower body.

In one embodiment, the determination unit may include a comparison determination unit to determine the degree of similarity by comparing the extinct object and the position information of the new object, and the appearance information divided into the head, upper body and lower body, respectively.

In one embodiment, the comparison determiner, in determining the similarity between the extinct object and the new object, the first weight to the similarity of the upper body, the second weight to the similarity of the lower body, and to the similarity of the head The similarity is determined by giving a third weight, and the first to third weights may be different from each other.

In example embodiments, the second weight may be greater than the third weight, and the first weight may be greater than the second weight.

In one embodiment, the determination unit may include a deletion unit for deleting the location and appearance information about the expired object that is not determined to be similar to the new object even after a predetermined time elapses.

In one embodiment, each of the tracking areas tracked by each of the tracking units may be fixed.

Object recognition and tracking method according to an embodiment for realizing another object of the present invention described above to extract the position and appearance information of each object by tracking a plurality of objects in each tracking area. The location and appearance information of the tracked objects are respectively stored. When an object is destroyed in any one of the tracking units and a new object is tracked in any one of the tracking units, the degree of similarity between the destroyed object and the new object is determined. If it is determined that the new object is similar to the destroyed object, the new object is allocated to the destroyed object, and if not determined to be similar, the new object is allocated to the new object to integrate the tracked objects.

In an embodiment, in storing the location and appearance information of the objects, respectively, when the objects are destroyed, the location and appearance information of the object immediately before the destruction may be stored.

In an embodiment, in the determining of the similarity, when the extinct object and the new object are humans, the similarity may be determined by dividing the appearance information of the person into head, upper body, and lower body.

In an embodiment, in the determining of the similarity, the similarity of the upper body is assigned a first weight, the second weight is assigned to the similarity of the lower body, and the similarity of the head is assigned to the similarity. The second weight may be greater than the third weight, and the first weight may be greater than the second weight.

In one embodiment, if it is determined that the destroyed object is similar to the new object even after a certain time elapses, the location and appearance information of the destroyed object may be deleted.

According to embodiments of the present invention, whether the new object is an expired object is determined by determining the similarity with the new object newly introduced in the tracking unit with respect to the objects destroyed by any one of the plurality of tracking units. It can be determined on the fly, further improving the object tracking ability.

In particular, even if there are a large number of objects that are destroyed due to poor recognition ability or poor tracking performance because the resolution of the tracking unit is not high, continuous tracking of one object is possible by repeatedly determining similarity of new objects. Furthermore, even if the object tracking is repeatedly disconnected in various environments such as the characteristics of the surrounding environment or the fast moving speed of the object, continuous object tracking may be performed through similarity determination.

In this case, if the destroyed object and the new object are determined to be the same, the new object is assigned to the same ID as that of the destroyed object and managed, thereby allocating IDs to a plurality of objects to be tracked. Ease of use is improved.

Furthermore, since the plurality of tracking units share information extracted in conjunction with each other, when the tracking unit is a fixed camera that observes the tracking area fixedly, effective continuous tracking is possible even for an object deviating from the tracking area.

In addition, when the object is a human, the external body information is divided into head, upper body, and lower body, and the weight of the similarity determination is given to the upper body, and the upper body having the smallest change in the appearance of the person actually moving is centered. Since the similarity of the appearance is determined, the similarity discrimination power of the appearance can be improved.

In addition, the storage space of the storage unit may be used more efficiently by deleting the location and appearance information of the destroyed object that is not determined to be similar to the new object even after a certain time elapses.

1 is a block diagram illustrating an object recognition and tracking system according to an embodiment of the present invention.
FIG. 2 is a schematic diagram illustrating a situation in which a plurality of objects are tracked by each of the tracking units in the object recognition and tracking system of FIG. 1.
3 is a schematic diagram illustrating a situation in which appearance information about a plurality of objects tracked by the first tracking unit of FIG. 2 is extracted.
FIG. 4 is a schematic diagram illustrating an example of object integration in the discriminating unit and the integrating unit in the object recognition and tracking system of FIG. 1.
5 is a flowchart illustrating an object recognition and tracking method using the object recognition and tracking system of FIG. 1.

As the inventive concept allows for various changes and numerous modifications, the embodiments will be described in detail in the text. However, this is not intended to limit the present invention to a specific disclosed form, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. In describing the drawings, similar reference numerals are used for similar elements. Terms such as first and second may be used to describe various components, but the components should not be limited by the terms.

The terms are used only for the purpose of distinguishing one component from another. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. Singular expressions include plural expressions unless the context clearly indicates otherwise.

In this application, the terms "comprise" or "consist of" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in the commonly used dictionaries should be construed as having meanings consistent with the meanings in the context of the related art and shall not be construed in ideal or excessively formal meanings unless expressly defined in this application. Do not.

Hereinafter, with reference to the accompanying drawings, it will be described in detail a preferred embodiment of the present invention.

1 is a block diagram illustrating an object recognition and tracking system according to an embodiment of the present invention. FIG. 2 is a schematic diagram illustrating a situation in which a plurality of objects are tracked by each of the tracking units in the object recognition and tracking system of FIG. 1. 3 is a schematic diagram illustrating a situation in which appearance information about a plurality of objects tracked by the first tracking unit of FIG. 2 is extracted. FIG. 4 is a schematic diagram illustrating an example of object integration in the discriminating unit and the integrating unit in the object recognition and tracking system of FIG. 1. 5 is a flowchart illustrating an object recognition and tracking method using the object recognition and tracking system of FIG. 1.

Hereinafter, the object recognition and tracking system 10 and the object recognition and tracking method 20 using the same will be described at the same time with reference to FIGS. 1 to 5.

First, referring to FIG. 1, the object recognition and tracking system 10 according to the present embodiment includes a tracking unit 100, a storage unit 200, a determination unit 300, and an integrated unit 400.

More specifically, referring to FIGS. 1, 2 and 5, the tracking unit 100 includes a plurality of first to nth tracking units 101, 102,..., 103, and the tracking units Tracks a plurality of objects with respect to the first to n th tracking areas 111, 112,..., 113.

In this case, the first to n-th tracking areas 111, 112,..., 113 may overlap each other as shown in FIG. 2, and thus, objects (eg, located in the overlapping area) The 530 and 540 may be simultaneously tracked by the plurality of tracking units 101 and 102.

In addition, each of the tracking units 101, 102,..., 103 may be an imaging device such as a two-dimensional or three-dimensional camera, and may be a device that can simultaneously acquire location information on a corresponding object in addition to an image. .

Accordingly, in the present exemplary embodiment, each of the tracking units 101, 102,..., 103 is a plurality of objects in the first to n-th tracking regions 111, 112,..., 113. 510, 520, 530, and 540 are tracked, respectively, and location information and appearance information are extracted for each of the plurality of objects 510, 520, 530, and 540 (step S10).

In this case, when one object is included in the plurality of tracking areas, the location information and the appearance information are extracted by being tracked by the plurality of tracking units, and the overlapping information is one by the integration unit 400 described later. It is recognized as an object of and assigned an ID.

On the other hand, the object may be, for example, a person, if the object is a human, each of the tracking units (101, 102, ..., 103), as shown in Figure 3, the location of the object and Extract appearance information.

That is, when the first to third objects 510, 520, and 530 are tracked in the first tracker 101, the first tracker 101 may be configured to include the first to third objects 510 and 520. And 530, each of which is divided into a head 511, 521, and 531, an upper body 512, 522, and 532, and a lower body 513, 523, and 533, to extract appearance information.

In this case, each appearance information 511, 512, and 513 of the divided object 510 is divided into one pixel form as shown and extracted as appearance information of the corresponding pixel.

However, the head, the upper body and the lower body constitute one object. When the first tracking unit 101 extracts the location information of each object, the whole object is one, that is, one pixel 514 and 524. 534, it is preferable to extract the position information of the pixel (514, 524, 534).

On the other hand, when extracting the respective appearance information for the divided head, upper body and lower body, a feature descriptor is generated in each of the divided head, upper body and lower body, and the generated feature descriptors are extracted as appearance information. In addition, various methods of extracting appearance information of a specific pixel may be applied.

As such, with respect to the first object 510 tracked by the first tracking unit 101, the entire first object 510 is treated as one pixel 514 to extract position information of the pixel 514. The first object 510 is divided into a head 511, an upper body 512, and a lower body 513 to extract appearance information of the pixels 511, 512, and 513.

In this case, as described above, if the first object 510 is also tracked by the second tracking unit 102, the location information and the appearance information are similarly extracted by the second tracking unit 102.

Thus, location information and appearance information may be extracted for all objects tracked by all tracking units, and some objects may be overlapped by a plurality of tracking units, and location information and appearance information may be extracted.

Meanwhile, referring to FIGS. 1 and 5, when location information and appearance information of all objects are extracted by the tracking units, the storage unit 200 stores the extracted location information and appearance information (step S20).

That is, the storage unit 200 also includes first to nth storage units 201, 202,..., 203, and thus, location information and appearance information extracted from the first tracking unit 101. Is stored in the first storage unit 201, the location information and appearance information extracted from the second tracking unit 102 are stored in the second storage unit 202, the n-th tracking unit 103 The location information and the appearance information extracted from may be stored in the n-th storage unit 203, respectively.

Alternatively, the storage unit 200 may be stored in a form of one storage unit, distinguishing the location information and the appearance information extracted from the first to n-th tracking units from each other at once.

Meanwhile, in the present embodiment, it is difficult to repeatedly or continuously store all the location information and the appearance information of all the objects extracted by the tracking units in the storage unit 200 at regular time intervals, because of the problem of storage capacity. At the start of the object recognition and tracking method according to the present embodiment, the location information and the appearance information stored in the storage unit 200 are primarily provided to the integration unit 400, the integration unit 400 Based on the provided location information and appearance information, if location information and appearance information of the same object are extracted by the plurality of tracking units, the corresponding objects may be regarded as one same object and assigned to the same ID.

On the contrary, for objects newly tracked by the respective tracking units, a temporary ID is assigned to distinguish the tracking unit, and the corresponding object is divided into corresponding IDs (COTO, COT1, COT2, C1T0, C1T1, and C1T2 of FIG. 4). Etc.), the integration unit 400 may further determine whether the same object is the same and assign the same object to the same ID.

Accordingly, when the same ID is assigned to the same object, the plurality of tracking units track the same object with the same ID, and the storage unit 200 stores all the location information and the appearance information extracted from the object. no need.

Thus, in the present embodiment, the storage unit 200 is only the position information and the appearance information of the object immediately before the object is destroyed, if the extinction object occurs among the objects tracked by the tracking unit Save (step S20).

That is, the tracking unit 101, 102, ..., 103 of any one of the tracking unit 101, 102, ..., 103 while extracting the location information and the appearance information repeatedly or continuously at a predetermined time interval for all the objects assigned to a specific ID Only when the object is destroyed in the occurrence of the, the storage unit 200 stores the location information and appearance information immediately before the destruction for the destroyed object.

In addition, when a new object is newly introduced from any one of the tracking units 101, 102,..., 103, and tracking starts, the location information and the appearance information of the new object are also stored in the storage unit 200. )

In this case, the tracking unit in which the occurrence of the destructive object is recognized may be the same as or different from the tracking unit in which the appearance of the new object is recognized, and thus, the identity of the tracking unit may be different.

1, 4, and 5, when the new object appears and is tracked, the determination unit 300 compares the extinct object with the positional information and appearance information of the new object ( In step S30, the degree of similarity between the destroyed object and the new object is determined (step S40).

More specifically, the determination unit 300 includes an extinction information input unit 310, a new information input unit 320, a tracking information input unit 330, and a comparison determination unit 340.

That is, when the extinct object is generated, the extinction information input unit 310 receives the position information and the appearance information of the object immediately before the extinction of the object from the storage unit 200.

In addition, when the new object appears in the new information input unit 320, location information and appearance information of the new object are input from the storage unit 200.

Meanwhile, the tracking information input unit 330 may additionally input location information, appearance information, or an assigned ID of the tracked objects.

The comparison determination unit 340 is the position information and appearance information immediately before the extinction of the extinct object input to the extinction information input unit 310, the position information and appearance information of the new object input to the new information input unit 320 Compare

In this case, the location information of the destroyed object is compared with the location information of the new object, and the appearance information of the destroyed object is compared with the appearance information of the new object.

As described above, the position information of the object is regarded as one pixel (514, 524, 534) as the entire object is stored as the position of the pixel, that is, coordinate information, and thus the pixels of the destroyed object and the pixel of the new object Compare the coordinates with each other.

In this case, the coordinate information of the destroyed object and the new object is compared, and if the coordinate is within a preset range, it is determined that the new object is the same as the destroyed object, and if it is outside the preset range, it is determined to be different. .

In addition, the preset range may be set in various ways in consideration of an error in coordinate recognition, and for example, may be set within a radius of 1m.

In addition, the appearance information of the object is divided into the head, the upper body and the lower body to regard each as a pixel to store the characteristics of the appearance of the pixel, accordingly, the head, upper body and lower body of the destroyed object and new objects Compare the appearance characteristics with each other.

In this case, when determining the similarity between the new object and the extinct object, it may be determined by applying different weights to the appearance characteristics of the head, upper body and lower body.

That is, the first weight may be assigned to the similarity of the appearance characteristics of the upper body, the second weight may be assigned to the similarity of the appearance characteristics of the lower body, and the third weight may be assigned to the similarity of the appearance characteristics of the head. To third weights may be set differently.

In general, when the object is a person, the hand can move freely while the person moves, especially when the hand is raised to the head side including the face. In addition, the face may rotate while looking forward, side and rear, and may cause various changes such as wearing or taking off a hat, wearing or taking off glasses, or changing a facial expression. Of course, some changes may occur in the case of the lower body walking.

Considering the change in the appearance of the person as a whole during the movement of the person, the upper body (particularly, the body except for the arms) can be seen that the relatively small change in the appearance of the appearance, the less the change in the appearance of the lower body, The characteristic change of the appearance of the head can be said to be the largest of the three parts.

Therefore, it is more accurate to determine the similarity between two objects by comparing the external characteristics of the lower body than by comparing the external characteristics of the head, and to compare the external characteristics of the lower body. It is more accurate to judge the similarity by comparing the appearance characteristics.

Accordingly, in the present embodiment, with respect to the first to third weights, the second weight is greater than the third weight, and the first weight is set to be greater than the second weight, so that the upper body similarity, the lower body The similarity of, and the similarity of the head can increase the influence on the similarity judgment.

That is, the comparison determination unit 340, in determining whether the extinction object and the new object has a high similarity in appearance, respectively, summing the similarity of the weighted upper body, the lower body similarity, and the head similarity, respectively. Based on the result, the similarity can be judged.

Of course, in this case, the criteria for determining the similarity, that is, whether to determine similarity may be set in advance, and may be variously set in consideration of the type of the estimated object and the surrounding environment.

As described above, the comparison determining unit 340 determines whether the extinguished object and the new object have high similarity based on whether the extinction object and the location information and appearance information of the new objects have high similarity. If it is determined that the same object and the similarity is not high, the new object is determined as a new object.

Furthermore, the integration unit 400 considers the new object as the same object as the destroyed object when the destroyed object and the new object are determined to be the same object based on the determination result of the comparison determining unit 340. Then, the same ID as the ID assigned to the destroyed object is allocated to the new object, and the new object is treated as a continuous object of the destroyed object so as to be tracked by the tracking unit 100 (step S60).

On the other hand, when it is determined that the destroyed object and the new object are not the same object, the integrator 400 determines the new object as a new object, allocates a new ID, and tracks it by the tracking unit 100. (Step S50).

On the other hand, when a new object is newly appeared, since it is necessary to determine whether the same object as described above with respect to all existing objects that have been destroyed, a lot of time may be required. Accordingly, in the present embodiment, the determination unit 300 further includes a deletion unit 350, and even after a predetermined time has elapsed, the information on the expired object that is not determined to be the same object as the new object is deleted. As described above, the new object having a high degree of similarity with the destroyed object is not considered to be present, and an additional comparison determination step is omitted (step S70).

In this case, the time for deleting the information about the expired object may be set in advance, which may be variously set in consideration of the types of tracked objects or the environment in which the objects are moved.

Meanwhile, referring to FIG. 4, an example of position comparison in the determination unit 300 and the integration unit 400 will be described below.

That is, as shown in FIG. 4, after the first to third objects 510, 520, and 530 are destroyed in the first tracking unit 101, the second to third tracking units 102 may include eleventh to thirteenth. A case where the objects 610, 620, and 630 newly appear will be described as an example.

In this case, the first to third objects 510, 520, and 530 are regarded as the pixels 514, 524, and 534, and location information of each pixel is extracted by the first tracking unit 101. And is stored in the first storage unit 201. Similarly, location information of each pixel is extracted by the second tracking unit 102 by considering the eleventh to thirteenth objects 610, 620, and 630 as the pixels 614, 624, and 634. It is stored in the second storage unit 202.

The stored information is provided to the extinction information input unit 310 and the new information input unit 320, and then compared by the comparison determining unit 340 to determine whether each object is the same object based on similarity. .

In this case, based on the similarity of the position, for example, the comparison determiner 340 determines that the second object 520, 524 and the eleventh object 610, 614 are the same object. When the third object 530, 534 and the twelfth object 620, 624 are determined to be the same object, the integrator 400 integrates the second and eleventh objects into the same object. The third and twelfth objects are integrated into the same object.

Meanwhile, before the first to third objects 510, 520, and 530 are integrated by the integrator 400 into the objects tracked by the first tracker 101, C0T0 and C0T1 are integrated. , ID can be assigned to C0T2. Similarly, the eleventh through thirteenth objects 610, 620, and 630 are IDs of C1T0, C1T1, and C1T2 before being integrated by the integrator 400 into objects tracked by the second tracking unit 102. Can be allocated.

Thereafter, after comparing the first to third objects 510, 520, and 530 that are extinguished with the newly introduced eleventh to thirteenth objects 610, 620, and 630, the integrator 400 is determined. Assigns M1 having the same ID to the second and eleventh objects 520 and 610 which are the same object, and M2 having the same ID to the third and twelfth objects 530 and 620 which are the same object. do.

Meanwhile, objects determined to be different objects may also be allocated with a new ID through the integrator 400. Accordingly, the first object 510 is assigned an ID of M1 and the thirteenth object. The object 630 may be assigned an ID of M3.

According to the embodiments of the present invention as described above, it is determined whether the new object is an expired object by determining the similarity with the new object newly introduced in the tracking unit with respect to the objects destroyed by any one of the plurality of tracking units. It is possible to immediately determine whether or not to improve the object tracking ability.

In particular, even if there are a large number of objects that are destroyed due to poor recognition ability or poor tracking performance because the resolution of the tracking unit is not high, continuous tracking of one object is possible by repeatedly determining similarity of new objects. Furthermore, even if the object tracking is repeatedly disconnected in various environments such as the characteristics of the surrounding environment or the fast moving speed of the object, continuous object tracking may be performed through similarity determination.

In this case, if the destroyed object and the new object are determined to be the same, the new object is assigned to the same ID as that of the destroyed object and managed, thereby allocating IDs to a plurality of objects to be tracked. Ease of use is improved.

Furthermore, since the plurality of tracking units share information extracted in conjunction with each other, when the tracking unit is a fixed camera that observes the tracking area fixedly, effective continuous tracking is possible even for an object deviating from the tracking area.

In addition, when the object is a human, the human body information is divided into head, upper body, and lower body, and in particular, the weight of the similarity determination is given to the upper body with the highest weight. Since the similarity of the appearance is determined, the similarity discrimination power of the appearance can be improved.

In addition, the storage space of the storage unit may be used more efficiently by deleting the location and appearance information of the destroyed object that is not determined to be similar to the new object even after a certain time elapses.

While the foregoing has been described with reference to preferred embodiments of the present invention, those skilled in the art will be able to variously modify and change the present invention without departing from the spirit and scope of the invention as set forth in the claims below. It will be appreciated.

An object recognition and tracking system based on partitioned appearance information and an object recognition and tracking method using the same according to the present invention have industrial applicability that can be used in a device for recognizing and tracking a moving person or an object.

10: object recognition and tracking system
100: tracking unit 200: storage unit
300: discrimination unit 310: extinction information input unit
320: new information input unit 330: tracking information input unit
340: comparison unit 400: integrated unit

Claims (13)

A plurality of tracking units which track a plurality of objects in each tracking area and extract location and appearance information of each object;
A plurality of storage units respectively storing position and appearance information of the objects tracked by each of the tracking units;
A determination unit for determining similarity between the destroyed object and the new object when an object is destroyed in one of the tracking units and a new object is tracked in any one of the tracking units; And
And an integrating unit for integrating the tracked objects by allocating the new object to the destroyed object if it is determined that the new object is similar to the destroyed object, and assigning the new object to the new object if it is not determined to be similar. ,
The determination unit includes a comparison determination unit for comparing the extinction object and the position information of the new object, and the appearance information divided into the head, upper body and lower body, respectively, to determine the similarity,
In the similarity determination between the extinct object and the new object, the comparison determining unit assigns a first weight to the similarity of the upper body based on a change in the characteristic of the appearance of the object, and compares the similarity of the lower body to the similarity of the lower body. A small second weight is given, and the similarity of the head is assigned a third weight smaller than the second weight to determine similarity,
Each of the storage units, when the objects are destroyed from the tracking unit, stores the position and appearance information of the object just before the destruction,
Each of the objects is a person, the appearance information is extracted and stored by dividing the person into a head, upper body and lower body,
The determination unit may include a deletion unit that deletes the location and appearance information of the destroyed object that is not determined to be similar to the new object even after a certain time has elapsed.
And each tracking area tracked by each of the tracking units is fixed.
delete delete delete delete delete delete delete delete delete delete delete delete

Images