Image recognition device, image recognition method, and storage medium

Abstract

The present invention provides an image recognition device, a method for image recognition device, and a storage medium, which can detect a desired object more accurately without requiring a user to perform complex pre-input operations. The image recognition device includes: an image acquisition unit (11) for acquiring an image; a detection unit (12) for detecting a detection object from the image; an index generation unit (14) for generating an index for determining whether the detection object is a target object based on a history of the position and size of the detection object; and a determination unit (15) for determining whether the detection object is the target object using the index, wherein the index is information indicating a distribution tendency of the position and size of the detection object in the history.

Description

Technical Field

[0001] The present invention relates to an image recognition device, an image recognition method and a storage medium. Background Art

[0002] With the widespread use of network cameras, technologies for detecting people and objects from captured images are increasingly being utilized. When detecting people from captured images, objects with similar shapes or combinations of shapes within the captured image may be mistakenly detected as people. If these false detections occur frequently, proper image recognition cannot be performed.

[0003] Patent Document 1 discloses a technique for improving the detection accuracy of a detection target object by detecting an object of a specific size as a detection target object based on the distance from a reference position in a fisheye image.

[0004] Prior art literature

[0005] Patent Literature

[0006] Patent Document 1: Japanese Patent Application Publication No. 2019-159739 Summary of the Invention

[0007] Problems to be solved by the invention

[0008] However, the prior art described above requires pre-registering the relationship between the size of the detection object, which is to be determined as a person, and the distance of the detection object from the reference position. Consequently, the user must input information related to the camera's shooting environment, such as the camera's installation height, the height of the detection object, and the relationship between the distance from the center of the image (reference position) and the distance in real space. This input process places a significant burden on the user.

[0009] Therefore, an object of one embodiment of the present invention is to realize an image recognition device that detects a desired object with higher accuracy without requiring a user to perform complicated prior input operations.

[0010] Technical means to solve the problem

[0011] In order to solve the above-mentioned problem, an image recognition device according to an embodiment of the present invention includes: an image acquisition unit for acquiring a captured image; a detection unit for detecting an object of a specified type from the captured image as a detection object through image recognition; a recording control unit for controlling the recording of the history of the position and size of the detection object detected by the detection unit in the captured image; an indicator production unit for producing an indicator for determining whether the detection object is the target object of the detection object based on the history; and a determination unit for determining whether the detection object is the target object by comparing the position and size of the detection object in the captured image detected by the detection unit with the indicator.

[0012] The method of the image recognition device of another embodiment includes: an image acquisition step of acquiring a captured image; a detection step of detecting a specified type of object as a detection object from the captured image through image recognition; a recording control step of controlling the history of the position and size of the detection object detected by the detection step in the captured image; an indicator creation step of creating an indicator for determining whether the detection object is the target object of the detection object based on the history; and a determination step of determining whether the detection object is the target object by comparing the position and size of the detection object in the captured image detected by the detection step with the indicator.

[0013] Effects of the Invention

[0014] According to an embodiment of the present invention, an image recognition device can be realized, which can detect a desired object more accurately without requiring a user to perform complicated prior input operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a block diagram showing the configuration of the main parts of the image recognition system according to the first embodiment.

[0016] Figure 2 This is a model diagram showing a situation where a camera is installed above and takes pictures from a top-down direction.

[0017] Figure 3 This diagram shows the temporal changes in a captured image when the camera is installed upward and captures images in a top-down direction.

[0018] Figure 4 This is a model diagram showing the situation where the camera is installed downward and shooting is performed in an upward direction.

[0019] Figure 5 This diagram shows the temporal changes in a captured image when the camera is installed downward and captures images in an upward direction.

[0020] Figure 6 This is an example of a case where the wrong object was detected when the camera was mounted upward.

[0021] Figure 7 This is an example of a case where the wrong object was detected when the camera was installed downward.

[0022] Figure 8 This is a diagram showing the correlation between the position (Y coordinate) and size of the detection object.

[0023] Figure 9 This is a flowchart showing an example of the operation of the image recognition system according to the first embodiment.

[0024] [Explanation of Symbols]

[0025] 1: Image recognition device

[0026] 11: Image acquisition unit

[0027] 12: Testing Department

[0028] 13: Recording control unit

[0029] 14: Indicator Production Department

[0030] 15: Judgment Department

[0031] 20: Camera

[0032] 31: Object

[0033] 32: Detecting Objects

[0034] 33: Detection frame

[0035] 51: Multiple points

[0036] 52: Approximate curve (indicator)

[0037] 53: Point

[0038] 54: Distance

[0039] 100: Image recognition system DETAILED DESCRIPTION

[0040] [Implementation Method 1]

[0041] Hereinafter, an embodiment of one aspect of the present invention (hereinafter also referred to as "this embodiment") will be described with reference to the drawings. In the drawings, the same or corresponding parts are denoted by the same reference numerals and their description will not be repeated.

[0042] §1. Application Examples

[0043] Figure 1This is a block diagram showing the main components of an image recognition system 100 according to Embodiment 1. Image recognition system 100 includes a camera 20 and an image recognition device 1 that acquires images captured by camera 20 and performs image recognition. Once camera 20 is installed, its viewing angle and imaging direction remain constant and unchanged.

[0044] In the image recognition device 1, the detection unit 12 detects a predetermined type of object from a captured image acquired by the image acquisition unit 11. The determination unit 15 determines whether the detected object is a target object. The determination unit 15 determines whether the detected object is a target object based on an indicator acquired by the recording control unit 13.

[0045] The index predefines a correspondence between the position of the detection object in the captured image and the size of the target object at that position. Therefore, if the size of the detection object is close to the size in the index corresponding to the position of the detection object in the captured image, the detection object can be determined to be the target object.

[0046] §2.Structure Example

[0047] (Structure of Image Recognition Device 1)

[0048] The image recognition device 1 includes an image acquisition unit 11 , a detection unit 12 , a recording control unit 13 , a storage unit 131 , an index creation unit 14 , and a determination unit 15 .

[0049] The image acquisition unit 11 acquires an image captured by the camera 20 and outputs the captured image to the detection unit 12 .

[0050] The detection unit 12 detects a predetermined type of object in the input captured image through image recognition. While a person is typically used as the object, the object is not limited to a person and may be any object. The position and size of the detected object, as the detection result obtained by the detection unit 12, are output to the recording control unit 13 and the determination unit 15.

[0051] Here, the position of the detected object is the center of the frame surrounding the detected object (detection frame). The method for deriving the position of the detected object is not limited to the above method, and other derivation methods such as an arbitrary point on the detection frame or the center of gravity of the detected object can also be used.

[0052] Alternatively, the size of the detection object can be any of the detection frame's lengths. Examples of detection frame lengths include diagonal length, vertical length, horizontal length, any length in both the vertical and horizontal directions, or any length in both the vertical and horizontal directions. Alternatively, the size of the detection object can be the area of ​​the detection frame. Furthermore, when specifying the area occupied by the detection object itself, the size of the detection object can also be any of the lengths or areas of that area.

[0053] The recording control unit 13 controls the data recorded in the storage unit 131, which stores the program and various parameters of the image recognition device 1. Specifically, the recording control unit 13 records the correspondence between the position and size of the detected object as a detection history (history) in the storage unit 131. The recording control unit 13 outputs the detection history recorded in the storage unit 131 to the index creation unit 14.

[0054] The storage unit 131 is not limited to being a functional block of the image recognition device 1 but may be provided in an external device. That is, the recording control unit 13 may access the storage unit 131 provided in an external device via various communication means to record and acquire the detection history.

[0055] Based on the detection history, the index generator 14 generates an index representing the distribution trend of the position and size of the detected object. The index is, for example, information expressed in the form of a function and is used to determine whether the detected object is the target object. The index generator 14 outputs the generated index to the determination unit 15.

[0056] The determination unit 15 determines whether the object detected by the detection unit 12 is the target object based on the indicator generated by the indicator generation unit 14. The determination unit 15 outputs the determination result. Examples of output destinations for the determination result obtained by the determination unit 15 include a display control unit of a display device (not shown) or a communication control unit for transmitting the result to an arbitrary server or terminal device via a communication network.

[0057] (Photographed image accompanying installation of camera 20)

[0058] Figure 2 This is a model diagram showing a case where the camera 20 is installed upward and images are taken in a top-down direction. Figure 3 1 and 2 are diagrams showing temporal changes in captured images when the camera 20 is mounted upward and captures images in a top-down direction. Figure 4 This is a model diagram showing a case where the camera 20 is installed downward and images are taken in an upward direction. Figure 5 This is a diagram showing temporal changes in a captured image when the camera 20 is installed downward and captures images in an upward direction.

[0059] The image recognition device 1 detects an object assumed to be a person from a captured image as a detection object 32. Hereinafter, the captured image will be described with an origin at the lower left, an X-axis in the left-right direction, and a Y-axis in the up-down direction.

[0060] First, if Figure 2 As shown in FIG, consider the case where the camera 20 is installed above and photographs the object 31 in a downward direction. Figure 3 As shown in the image 41, the object far from the camera 20 is smaller in the upper part of the captured image. Figure 3 As shown in the image 42 , objects near the camera 20 appear larger in the lower portion of the captured image.

[0061] like Figure 3 As shown, the detection object 32 that moves from far away from the camera 20 to near the camera 20 gradually increases in size as the distance from the camera 20 decreases, and its position in the captured image moves from top (larger Y coordinate) to bottom (smaller Y coordinate).

[0062] In addition, if Figure 4 As shown in FIG, consider the case where the camera 20 is installed below and shoots the object 31 in an upward direction. Figure 5 As shown in the image 43, the objects far away from the camera 20 are smaller in the lower part of the captured image. Figure 5 As shown in the image 44, objects near the camera 20 appear larger in the upper portion of the captured image.

[0063] like Figure 5 As shown, the detection object 32 that moves from far away from the camera 20 to near the camera 20 gradually increases in size as the distance from the camera 20 decreases, and its position in the captured image moves from bottom (small Y coordinate) to top (large Y coordinate).

[0064] That is, regardless of the method of mounting the camera 20 , the position and size of the detection object change according to the movement of the detection object.

[0065] (False detection by camera 20)

[0066] Figure 6 This is an example of a case where an erroneous object is detected when the camera 20 is mounted upward. Figure 7 This is an example of a case where an erroneous object is detected when the camera 20 is installed downward.

[0067] like Figure 6As shown, when the camera 20 is mounted upward, the image recognition device 1 may mistakenly detect a non-human object at the position indicated by the detection frame 33. This may occur, for example, when the background image resembles the shape of a human body. When a non-human object is mistakenly detected, the size of the position (Y coordinate) of the display detection frame 33 is likely to be different from that when a human is detected.

[0068] Likewise, if Figure 7 As shown, when the camera 20 is installed downward, the image recognition device 1 may erroneously detect an object other than a person at the position indicated by the detection frame 33 .

[0069] That is, in Figure 6 In the case where the camera 20 is installed at the top, the detection frame 33 is displayed at the bottom of the image. Originally, since the detection frame 33 is located at the bottom of the image, the detection frame should be large, but the detected detection frame 33 is small. Figure 7 In the example above, when camera 20 is mounted downward, a detection frame 33 is displayed above the image. Since detection frame 33 is positioned above the image, it should be large, but the detected detection frame 33 is small. This suggests that the position and size of the detection frame during a false detection are likely different from those in the case of a detected person. Therefore, false detections can be determined based on indicators indicating the position and size of the detection frame.

[0070] (Correlation between the Position and Size of Detected Object 32)

[0071] Figure 8 The diagram shows the correlation between the position (Y coordinate) and the size of the detection object 32. The storage unit 131 records the detection history in which the position and size of the detection object detected in the past are associated with each other. The detection history is plotted as a detection history distribution on a two-axis scatter diagram. Figure 8 In the detection history distribution, the horizontal axis plots the size of the detection object, and the vertical axis plots the position of the detection object (the position in the vertical direction of the captured image (Y coordinate)). Multiple points 51 represent a certain degree of correlation and form a discrete distribution.

[0072] The index generator 14 acquires the detection history from the storage 131 and derives an approximate curve 52 based on the plurality of points 51. The least squares method can be used as an example of a method for deriving the approximate curve 52, but the method is not limited thereto and any other method of deriving the approximate curve may be used.

[0073] Here, during the detection history creation stage, if the detection history includes erroneously detected data, it is considered that there may be points that deviate significantly from the approximate curve. However, if these significantly deviated points are relatively few in number, their impact on approximate curve 52 is minimal and can be ignored. In other words, if the probability of erroneous detection is low, even if erroneous detection data is included during the detection history creation stage, it is still possible to create a sufficiently appropriate indicator. Furthermore, after the number of detection histories exceeds a certain value, only data that indicates that the detected object is the target object, as determined by the determination unit 15, is added to the detection history, thereby eliminating the adverse effects of erroneous detection data on indicator creation.

[0074] The index creation unit 14 creates an approximate curve 52 representing the distribution trend of the plurality of points 51, but this is not limited to this; any index may be used. For example, a function may be used. The approximate curve 52 representing the function is not limited to a continuous curve; it may also be a discontinuous curve or a straight line within a specified range of position and size. Furthermore, the function may be expressed in a table format, with values ​​determined within a specified range of position and size.

[0075] Here, the determination process of the determination unit 15 is described in the state of obtaining the above-mentioned index. If the detection object 32 is detected, the determination unit 15 compares the position and size of the detection object 32 with the function as the index. In the comparison process, the position and size of the detection object 32 are conceptually plotted on the Figure 8 At point 53 shown, distance 54 from approximate curve 52 is calculated and compared with a predetermined value. Here, distance 54 is the shortest distance between point 53 and approximate curve 52. Alternatively, distance 54 may be the difference in the Y coordinates between point 53 and approximate curve 52. The method for calculating distance 54 is not limited to this, and any calculation method may be used.

[0076] §3. Action Examples

[0077] Figure 9 This is a flowchart showing an example of the operation of the image recognition system 100 according to the first embodiment.

[0078] In S11, the image acquisition unit 11 acquires a captured image from the camera 20. The acquired captured image is output to the detection unit 12. The detection unit 12 detects the object 31 as the detected object 32 from the input captured image.

[0079] In S12, the detection unit 12 determines whether at least one detection object 32 has been detected. If at least one detection object 32 has not been detected (No in S12), the process proceeds to S13. On the other hand, if at least one detection object 32 has been detected (Yes in S12), the process proceeds to S14.

[0080] In S13 , the determination unit 15 displays that there is no detected object 32 .

[0081] In S14, the recording control unit 13 determines whether the detection history is equal to or greater than a predetermined number. If it is less than the predetermined number (No in S14), the process proceeds to S19. On the other hand, if it is equal to or greater than the predetermined number (Yes in S14), the process proceeds to S15.

[0082] In S15 , the index generating unit 14 reads the detection history from the storage unit 131 and derives an approximate curve 52 reflecting the distribution tendency of the detection history as an index.

[0083] In S16, the distance 54 between the approximate curve 52 and the point 53 corresponding to the position and size of the detection object 32 detected this time is calculated. If the calculated distance 54 is greater than a predetermined threshold (No in S16), the process proceeds to S17. On the other hand, if the calculated distance 54 is less than the predetermined threshold (Yes in S16), the process proceeds to S18.

[0084] In S17 , the determination unit 15 determines that the detected object 32 is not a person (target object) but is an erroneous detection.

[0085] In S18 , the determination unit 15 determines that the detected object 32 is a person.

[0086] In S19 , the recording control unit 13 stores the position and size of the detected object in the detection history.

[0087] In S20 , the determination unit 15 outputs the results of the detection unit 12 and the determination unit 15 , and displays them on the display device.

[0088] Here, the recording control unit 13 can be configured to exclude detection objects that are clearly misdetected or misjudged by the user from the results of the visual detection unit 12 and the judgment unit 15 from being used to generate the approximate curve in the detection history distribution by setting a flag indicating misdetection. This operation improves the accuracy of determining the target object among the detected objects.

[0089] §4. Function and Effect

[0090] Whether the camera 20 is mounted above and shooting from a downward angle or below and shooting from an upward angle, moving the detection object, which is an object in the image, causes the position and size of the detection object in the image to change. Because the changes in the position and size of the detection object are related, an index is derived based on the distribution trend derived from this relationship. By comparing the position and size of the detection object with this index, it is possible to determine whether the detection object is the target object for detection. This makes it possible to determine whether an object other than the target object has been mistakenly detected, thereby improving the accuracy of target object detection.

[0091] Therefore, the position and size of detected objects are automatically recorded. When the number of recorded position and size combinations exceeds a predetermined number, it is possible to determine whether the detected object is the target object (e.g., a person) or another object. Therefore, human detection and identification can be performed without prior input of information such as the size of a person, depending on the installation environment corresponding to the camera 20's installation position, thereby improving user convenience.

[0092] By collecting a predetermined number of histories of the positions and sizes of the detected objects, the distribution trend of the positions and sizes of the detected objects can be known, and an index with high accuracy can be created.

[0093] In particular, the indicator may be a function, and further an approximate curve in a two-dimensional space, and may only consider the up and down directions in the image to determine whether the detected object is a target object.

[0094] Furthermore, the approximate curve is considered in terms of the size of the detected object relative to the vertical direction, but this is not limiting. It can also be the size of the detected object relative to a one-dimensional position in any uniaxial direction. That is, the approximate curve can be the size of the detected object relative to the horizontal direction, or relative to a straight line with an arbitrary slope in the captured image. The approximate curve can be a function derived from a group of points representing the one-dimensional position and the size of the detected object. For example, if camera 20 is installed on the side wall of a corridor at approximately the same height as a human body, and the shooting direction is determined so that the far side of the corridor is the left side of the captured image and the near side of the corridor is the right side of the captured image, the size of the target object increases from the left side to the right side of the captured image. In this case, the approximate curve can simply be set to a curve corresponding to the size of the detected object relative to the horizontal direction of the captured image.

[0095] [Variant 1]

[0096] In Embodiment 1, only the vertical direction of the image is considered when detecting the position of the object. However, in Modification 1, an example is described in which both the vertical and horizontal directions of the image are considered. That is, the size of the detected object is associated with the X and Y coordinates on the image.

[0097] Therefore, the detection history distribution is obtained as a scatter plot of three axes: the X coordinate of the detected object, the Y coordinate of the detected object, and the size of the detected object. Furthermore, the indicator derived from the detection history distribution is an approximate surface. An approximate surface can also be represented by a function. Furthermore, the approximate surface representing a function is not limited to a continuous surface and can also be a discontinuous surface or plane within a specified range of position and size. Furthermore, a function can also be represented in the form of a table whose values ​​are determined within a specified range of position and size.

[0098] Therefore, in Variation 1, a new point representing the X- and Y-coordinates and size of the detected object in three dimensions can be used to calculate the distance relative to the approximate surface. This distance can be calculated using the distance to the point on the approximate surface closest to the new point. Alternatively, the distance to the point on the approximate surface corresponding to the size of the new point's X- and Y-coordinates can be used. If these distances are less than a specified value, the detected object is determined to be a target object (person).

[0099] In Embodiment 1, in Modification 1, since the detection history distribution is represented by a three-axis scatter plot, an approximate surface can be created that takes into account the relationship of three-dimensional depth. Therefore, even if the Y coordinate is the same, as long as the X coordinate is different, it can also be used to deal with different sizes.

[0100] For example, if the captured space contains steps or other height differences, the size of the target object may change differently depending on the location of the captured image. In this case, by setting an approximate curved surface corresponding to the height difference, it is possible to accurately determine whether the object is the target object.

[0101] Furthermore, if an area where objects, such as a wall, cannot penetrate is within the range of the captured image, the area where the object cannot exist can be clearly defined by presetting the size of the detected object to 0 or a maximum value. In this case, the user must input information related to the relationship between the position and size of the detected object in advance.

[0102] Furthermore, the approximate surface is considered based on the size of the detected object relative to the two-dimensional space of the X and Y coordinates, but this is not limiting. The size of the detected object relative to its two-dimensional position in predetermined biaxial directions may also be considered. In this case, the function can be an approximate surface in three dimensions based on the two-dimensional position of the detected object and its size.

[0103] [Variant 2]

[0104] In the first embodiment, the camera 20 is angled relative to the vertical direction. However, in the second modification, the optical axis is oriented in the vertical direction. For example, the camera 20 is mounted on the ceiling facing directly downward.

[0105] In this case, it is common to use a fisheye lens or a wide-angle lens, etc. With these lenses, even if they are equidistant from the camera 20, objects located at an angle to the optical axis of the camera 20 have the property of appearing smaller in the captured image than objects located on the optical axis.

[0106] Furthermore, the captured image has the property of being arranged concentrically around the optical axis. Therefore, the detection history differs from the tendency of the first embodiment, and the position of the detected object is expressed not by the Y coordinate but by the radius from the center position.

[0107] [Variant 3]

[0108] In the first embodiment, the relationship between the approximate curve and the position and size of the detected object is used to determine whether the detected object is the target object. In the third variant, the approximate curve is not used, and artificial intelligence (AI) is used to determine whether the detected object is the target object based on the detection history distribution and the position and size of the detected object. For example, machine learning can also be used to determine the deviation value.

[0109] In this case, even if a significant outlier value corresponding to an erroneous detection is mixed in the detection history distribution, the influence of the outlier value is minimized in the learning phase and does not become a problem.

[0110] 〔Summarize〕

[0111] In order to solve the above-mentioned problem, an image recognition device according to an embodiment of the present invention includes: an image acquisition unit for acquiring a captured image; a detection unit for detecting an object of a specified type from the captured image as a detection object through image recognition; a recording control unit for controlling the recording of the history of the position and size of the detection object detected by the detection unit in the captured image; an indicator production unit for producing an indicator for determining whether the detection object is the target object of the detection object based on the history; and a determination unit for determining whether the detection object is the target object by comparing the position and size of the detection object in the captured image detected by the detection unit with the indicator.

[0112] This structure allows an indicator to be created based on the history of the detected object's position and size, and determines whether the detected object is the target object based on its position and size relative to the indicator. This eliminates the need for the user to input information based on the environment in advance, improving target object detection accuracy.

[0113] The index creation unit may create a function representing a relationship between a position of the detection object and a size of the detection object as the index.

[0114] With this structure, a function can be derived as an index. By using the function, the size of the detection object at any position of the detection object can be uniquely derived.

[0115] The determination unit may determine that the detection object is the target object when a distance determined based on the function and the position and size of the detection object in the captured image is smaller than a predetermined threshold.

[0116] With the above configuration, the determination unit can calculate the distance between a point determined by the position and size of the detection object and the function, and determine whether the detection object is a target object when the distance is smaller than a predetermined threshold.

[0117] It may also be that the recording control unit performs control for recording the correspondence between the one-dimensional position of the detection object in a specified uniaxial direction in the captured image and the size of the detection object, the function is an approximate curve in two dimensions of the one-dimensional position of the detection object and the size of the detection object, and the distance is the distance between a point representing the one-dimensional position of the detection object and the size of the detection object and the approximate curve.

[0118] With the above configuration, it is possible to determine the size of the target object and to determine false detection of the detected object by considering only the up-down direction in the image.

[0119] It may also be that the recording control unit performs control for recording the correspondence between the two-dimensional position of the detection object in the specified biaxial direction in the captured image and the size of the detection object, the function is an approximate surface in three dimensions of the two-dimensional position of the detection object and the size of the detection object, and the distance is the distance between a point representing the two-dimensional position of the detection object and the size of the detection object and the approximate surface.

[0120] This structure allows for accurate detection of objects by taking into account the vertical, horizontal, and vertical directions of the image. By also including the horizontal and vertical directions in the determination, accurate detection is possible even when there are steps or other height differences in the captured space.

[0121] The method of the image recognition device of another embodiment includes: an image acquisition step of acquiring a captured image; a detection step of detecting a specified type of object as a detection object from the captured image through image recognition; a recording control step of controlling the history of the position and size of the detection object detected by the detection step in the captured image; an indicator creation step of creating an indicator for determining whether the detection object is the target object of the detection object based on the history; and a determination step of determining whether the detection object is the target object by comparing the position and size of the detection object in the captured image detected by the detection step with the indicator.

[0122] The image recognition device of each embodiment of the present invention can also be implemented by a computer. In the case, by making the computer act as the various parts (software components) included in the image recognition device, the image recognition program of the image recognition device implemented by the computer and the computer-readable recording medium recording the program also fall within the scope of the present invention.

[0123] [Example of implementation through software]

[0124] The functions of the image recognition device 1 (hereinafter referred to as the "device") can be realized by the following program, which is a program for enabling a computer to function as the device, and is used to enable the computer to function as each control block of the device (especially each part included in the image recognition device 1).

[0125] In this case, the apparatus includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The functions described in each embodiment are implemented by executing the program by the control device and the storage device.

[0126] The program may be recorded in one or more non-temporary computer-readable recording media. The recording medium may or may not include the device. In the latter case, the program may be supplied to the device via any transmission medium, whether wired or wireless.

[0127] Furthermore, some or all of the functions of each control block can be implemented using logic circuits. For example, integrated circuits incorporating logic circuits that function as each control block are also within the scope of the present invention. Furthermore, the functions of each control block can also be implemented using, for example, a quantum computer.

[0128] [Additional Notes]

[0129] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the technical claims. Embodiments obtained by appropriately combining technical means disclosed in different embodiments are also included in the technical scope of the present invention.

Claims

1. An image recognition device, comprising: An image acquisition unit, which acquires a captured image; a detection unit that detects a predetermined type of object as a detection object from the captured image through image recognition; a recording control unit that performs control for recording a history of the position and size of the detection object detected by the detection unit in the captured image; an index generating unit for generating an index for determining whether the detection object is a target object of detection based on the history; as well as The determination unit determines whether the detection object is the target object by comparing the position and size of the detection object in the captured image detected by the detection unit with the index.

2. The image recognition device according to claim 1, wherein: The index creation unit creates a function representing a relationship between a position of the detection object and a size of the detection object as the index.

3. The image recognition device according to claim 2, wherein: The determination unit determines that the detection object is the target object when the distance determined based on the function and the position and size of the detection object in the captured image is smaller than a predetermined threshold.

4. The image recognition device according to claim 3, wherein: The recording control unit performs control for recording a correspondence relationship between a one-dimensional position of the detection object in a predetermined uniaxial direction and a size of the detection object in the captured image. The function is an approximate curve in two dimensions of the one-dimensional position of the detection object and the size of the detection object. The distance is a distance between a point indicating the one-dimensional position of the detection object and the size of the detection object, and the approximate curve.

5. The image recognition device according to claim 3, wherein: The recording control unit performs control for recording a correspondence relationship between the two-dimensional position of the detection object in predetermined biaxial directions and the size of the detection object in the captured image. The function is an approximate surface in three dimensions of the two-dimensional position of the detection object and the size of the detection object. The distance is a distance between a point indicating the two-dimensional position of the detection object and the size of the detection object, and the approximate curved surface.

6. A method for an image recognition device, comprising: An image acquisition step, acquiring a captured image; a detection step of detecting objects of a specified type from the captured image as detection objects through image recognition; a recording control step for performing control for recording a history of the position and size of the detection object detected in the detection step in the captured image; an indicator making step of making an indicator for determining whether the detection object is a target object of detection based on the history; as well as The determination step determines whether the detection object is the target object by comparing the position and size of the detection object in the captured image detected by the detection step with the index.

7. A computer-readable storage medium storing an image recognition device program, which is an image recognition program for enabling a computer to function as the image recognition device as described in claim 1, wherein the image recognition device program is used to enable the computer to function as the image acquisition unit, the detection unit, the recording control unit, the indicator creation unit, and the judgment unit.

Images