JP2024014516A - Information processing device, information processing method and program - Google Patents

Abstract

According to the present invention, object detection accuracy can be improved. [Solution] An information processing apparatus according to the present invention that solves the above problems includes a specifying means for specifying an important position in an image, an extracting means for extracting a region of interest including the important position from the image, and an information processing apparatus for all parts of the image. The method further comprises: a setting means for setting a plurality of partial images to encompass a region; a detection means for detecting an object from a partial image of the image corresponding to the region of interest; and the plurality of partial images. Features. [Selection diagram] Figure 2

Description

The present invention relates to a technique for detecting an object from an image.

In Patent Document 1, a process is performed in which a target image is divided into a plurality of regions, cut out, and converted into an image with a predetermined number of pixels so that it can be treated as an input image to a learning device.

JP 2021-144589 Publication

In the technique disclosed in Patent Document 1, it becomes difficult to detect the object from the image depending on the relationship between the position where the object to be detected is frequently displayed and the divided regions.

The present invention aims to improve object detection accuracy.

An information processing apparatus according to the present invention that solves the above problems includes a specifying means for specifying an important position in an image, an extraction means for extracting a region of interest including the important position from the image, and an entire area of the image. The present invention is characterized by comprising: a setting means for setting a plurality of partial images, a partial image of the image corresponding to the region of interest, and a detection means for detecting an object from the plurality of partial images.

According to the present invention, object detection accuracy can be improved.

1 is a diagram illustrating a configuration example of an information processing device. 1 is a diagram illustrating an example of a functional configuration of an information processing device. FIG. 3 is a diagram illustrating an example of setting a region of interest. 3 is a flowchart illustrating processing executed by the information processing device. FIG. 3 is a diagram illustrating an example of an operation screen for setting a region of interest. 3 is a flowchart illustrating processing executed by the information processing device. FIG. 3 is a diagram illustrating an example of an operation screen for setting a region of interest. FIG. 3 is a diagram illustrating an example of setting a region of interest.

EMBODIMENT OF THE INVENTION Hereinafter, the form for implementing this invention is demonstrated using drawings.

[First embodiment]
In recent years, image analysis that uses images captured by imaging devices such as surveillance cameras to detect and track objects, estimate attributes, etc., and estimation of the number of objects using the results of such image analysis has become popular in various scenes. It is being carried out in Here, object detection is information processing that outputs, for example, the position and size of an object to be detected in an image, the attributes of the object, the reliability of the object, and the like. In machine learning models used for object detection, the size of data input to the model is determined in advance, and preprocessing such as reducing or dividing images is performed to match the input size. However, when dividing an image into images to be input into a machine learning model, there is a possibility that objects may become difficult to detect depending on the scene. Therefore, in this embodiment, an information processing apparatus (information processing method) that can improve object detection accuracy even when performing object detection processing by dividing an image into a plurality of partial images will be described.

FIG. 1 is a block diagram showing a configuration example of an information processing apparatus 100 according to this embodiment. The information processing device 100 in this embodiment has an object detection function that detects an object to be detected from an image captured by an imaging device such as a surveillance camera. In the following, a case will be described in which a human face is detected as an example of an object. The object is not limited to this, and can include vehicles, animals, and various objects, and can be applied to any system that analyzes images and detects a predetermined object.

The information processing device 100 according to the present embodiment includes a CPU 101, a memory 102, a communication interface (I/F) section 103, a display section 104, an operation section 105, and a storage section 106, which communicate with each other via a system bus 107. Possibly connected. Note that the information processing apparatus 100 according to the present embodiment may further include configurations other than this.

A CPU (Central Processing Unit) 101 controls the entire information processing device 100 . The CPU 101 controls the operation of each functional unit connected via a system bus 107, for example. The memory 102 stores data, programs, etc. used by the CPU 101 for processing. Further, the memory 102 functions as a main memory, a work area, etc. for the CPU 101. When the CPU 101 executes processing based on the program stored in the memory 102, the functional configuration of the information processing apparatus 100 shown in FIG. 2, which will be described later, and the processing shown in the flowchart shown in FIG. 4, which will be described later, are realized.

The communication I/F unit 103 is an interface that connects the information processing device 100 to a network. The display unit 104 includes a display member such as a liquid crystal display, and displays the results of processing by the CPU 101 and the like. The operation unit 105 includes operation members such as a mouse, a touch panel, and buttons, and inputs user operations to the information processing apparatus 100. The storage unit 106 stores, for example, various data necessary when the CPU 101 performs processing related to a program. Further, the storage unit 106 stores, for example, various data obtained by the CPU 101 performing processing related to a program. Note that data, programs, etc. used by the CPU 101 for processing may be stored in the storage unit 106.

FIG. 2 is a block diagram showing an example of the functional configuration of the information processing device 100. The information processing device 100 includes an image acquisition section 201, an object detection section 202, an image extraction section 203, a modification section 204, an output section 205, and a storage section 206.

The image acquisition unit 201 acquires an image to be subjected to object detection. In this embodiment, an image to be subjected to object detection is acquired from the outside (for example, an imaging device) through the communication I/F unit 103. Hereinafter, the data of the image that is the object of object detection, which is acquired by the image acquisition unit 201, will also be simply referred to as an "input image." In the following description, the input image is, for example, a 720×480 pixel RGB image with a width of 720 pixels in the horizontal direction (horizontal direction) and a height of 480 pixels in the vertical direction (vertical direction). Note that the input image is not limited to an RGB image of 720 x 480 pixels, and any image can be used as the input image, and for example, the width in the horizontal direction and the height in the vertical direction may be different. .

The object detection unit 202 performs object detection processing on each of the partial images obtained by dividing the input image using predetermined analysis processing. In this embodiment, the object detection unit 202 detects information indicating the position of a person's face as a specific object from a plurality of partial images obtained by dividing the input image acquired by the image acquisition unit 201. Further, the object detection unit 202 outputs a detection result using a machine learning model that has been trained to be able to detect the face of a person included in an image. That is, the predetermined analysis process is a process of detecting a specific object from an image of a predetermined size and outputting information indicating the position and size of the object on the image. For example, this can be realized by applying the technology described in Non-Patent Document 1 below.

(Non-patent Document 1) J. Redmon, A. Farhadi, “YOLO9000: Better Faster Stronger”, Computer Vision and Pattern Recognition (CVPR) 2016.
Note that the predetermined analysis process in the object detection unit 202 is not limited to the technique disclosed in Non-Patent Document 1, and various techniques can be applied as long as the technique can detect the object to be detected.

In this embodiment, the object detection unit 202 detects an RGB image (partial image) of 240 x 240 pixels, which has a width of 240 pixels in the horizontal direction (horizontal direction) and a height in the vertical direction (vertical direction), as an example. shall be carried out. If an image of another size is input, resizing or deformation processing may be performed using any method such as the generally known bicubic method.

The image extraction unit 203 extracts a partial image to be input to the object detection unit 202 from the input image acquired by the image acquisition unit 201. The image extraction unit 203 extracts partial images using roughly two types of methods. The first method is to extract a region of interest that includes important positions in the image from the image. The region of interest is a closed region indicated by the coordinate system of the input image, which is set in order to improve object detection accuracy in an input image when there is a region in which an object is particularly desired to be detected. The second method is to set a plurality of partial images equally or based on a predetermined rule for the entire area of the input image. The plurality of partial images are images suitable for performing the above-described predetermined analysis processing or for resizing an input image to an image of a predetermined size. Although the shape of the region of interest or partial image extracted here may be rectangular, it is assumed to be square in order to simplify the following explanation.

The predetermined rule is set based on position and size information for extracting a partial image. If position and size information for extracting a partial image is not particularly given, for example, as shown in Fig. 3(a), a 240 x 240 pixel square area obtained by equally dividing a 720 x 480 pixel input image into six Set six. Dotted lines 301 indicate the boundaries of the six partial images extracted at this time. On the other hand, if 300 pixels on each side is specified as the size information of the partial image, square areas of 300 x 300 pixels are uniformly arranged to cover the entire area of the image, as shown in Figure 3(b). Six partial images are extracted for this purpose. At this time, the extraction regions may overlap each other. Dotted lines 302 indicate the boundaries of each partial image extracted at this time.

For example, if information is given that the center coordinates of the region of interest are (500, 200) and the size is 240 pixels on one side, one partial image as shown in FIG. 3(c) is extracted. A square indicated by a dotted line 303 is a partial image corresponding to a region of interest of 240×240 pixels whose center coordinates are at a position (500,200) when the upper left of the image is the origin (0,0).

The image extraction unit 203 also acquires information on important positions on the image and information on the size of the region of interest, which will be described later, and performs processing to calculate the center coordinates of the region of interest and the size of the region of interest corresponding to the center coordinates. The image extraction unit 203 also performs processing to determine the position and size of each region of interest for automatically arranging partial images so as to encompass the entire image, and to extract each partial image.

The modification unit 204 modifies the detection results obtained by the object detection unit 202 for each partial image. For example, when a plurality of detection results are output for one object, or when one object is detected across a plurality of regions of interest, result correction processing is performed to integrate the detection results.

The output unit 205 outputs the region of interest to a display means. By displaying information indicating regions of interest and important positions on the display means, it becomes easier for the user to make desired settings, and detection accuracy can be improved. Further, the output unit 205 outputs the output from the correction unit 204, that is, the result of the analysis process. The output information includes coordinate information indicating the position of the detected object on the image, an image in which a detection frame represented by a circumscribed rectangle of the detected object is superimposed on the input image, and classification information of the detected object.

The storage unit 206 stores data used in processing in each of the functional units 201 to 205 of the information processing device 100, data obtained as a processing result, and the like.

Next, processing performed by the information processing apparatus 100 will be described with reference to FIGS. 4 and 5. FIG. 4 is a flowchart for explaining processing executed by the information processing device. The processing shown in the flowchart of FIG. 4 is executed by the CPU 101 of FIG. 1, which is a computer, according to a computer program stored in the storage unit 106. In the following description, each process (step) is indicated by adding S to the beginning thereof, thereby omitting the notation of the process (step). FIG. 5 is an example of an operation screen suitable for this embodiment.

In S401, the image acquisition unit 201 acquires an input image (an image to be subjected to object detection). In this embodiment, the input image is assumed to be a 720×480 pixel image as described above.

In S402, the image extraction unit 203 identifies important positions in the input image. Specifically, the image extraction unit 203 acquires information regarding important positions indicating areas where the object to be detected is frequently seen on the 720×480 pixel image. Further, as an example of a method of defining an important position, a predetermined position such as the center of the input image can be set. As another method, important positions are identified based on the position of the object detected by executing a predetermined analysis process on images acquired at a predetermined time. For example, the CPU 101 executes generally known moving body detection processing or person detection processing on an input image acquired at a predetermined timing, and defines the center of gravity of the circumscribed rectangle of the person detected at that time as an important position. . As yet another method, the image extraction unit 203 identifies a position on the input image specified by the user as an important position. For example, the image extraction unit 203 acquires information regarding important positions specified by the user using the operation screen shown in FIG. 5 displayed on the display unit 104 and the operation unit 105.

The attention area setting screen 500 in FIG. 5A includes an image display section 510, an OK button 501, a cancel button 502, a reset button 503, an automatic placement button 504, and an add attention area button 505. The above-mentioned input image is displayed on the image display section 510, and in FIG. 5(a), the vicinity of the entrance and exit of the elevator is displayed. In FIGS. 5(b) to 5(g), reference numerals are omitted for screen elements common to those in FIG. 5(a). When the user presses the add attention area button 505, a square attention area frame 520 of 240×240 pixels is displayed (FIG. 5(b)). In FIG. 5B, the attention area frame 520 is initially displayed at the upper left of the image display section 510, but the display position of the attention area frame immediately after addition is not particularly limited. Additionally, the size of the added attention area in Figure 5(b) is 240 x 240 pixels, which is suitable for performing object detection processing on this image, but the size of the added attention area may be changed depending on the conditions. It's also good to do. Alternatively, the user may be able to freely change the size. This attention area frame 520 can be moved on the screen by a user's operation. The user moves the attention area frame 520 to a position where object detection is to be performed or a position where the object to be detected is frequently seen, that is, an important position. FIG. 5C shows the attention area frame 520 moved so as to overlap the elevator entrance. The image extraction unit 203 acquires positional information of the attention area frame 520, the coordinates of the upper left apex and the lower right apex of the attention area 520, which is a square here, or information on the center coordinates and the length of one side of the square.

FIGS. 5(d) to 5(f) are other examples of operation screens for specifying important positions on the screen. In FIG. 5(c), the attention area frame is placed directly on the screen, whereas in FIG. 5(d), the operation screen is used to specify important positions with points using the pointer 530. The image extraction unit 203 acquires the coordinates specified with the mouse pointer or tapped on the touch panel. Further, in FIG. 5E, a predetermined figure resembling the object to be detected, here a humanoid figure 540, is placed at an important position. The image extraction unit 203 acquires the position information of this humanoid figure. Furthermore, FIG. 5(f) is an operation screen for specifying the vicinity of important positions by placing a predetermined figure or drawing it freehand. In this case as well, the image extraction unit 203 acquires the positional information of the drawn closed region 550 (such as the position of the center of gravity of the closed region).

In S403, the image extraction unit 203 determines the center coordinates of the region of interest based on the important position acquired in S402. If a square region of interest is defined as shown in FIG. 5(c), the coordinates of the center point of the square are acquired. In the example of FIG. 5(c), the attention area frame 520 has a size of 240×240 pixels and is displayed at a position with center coordinates (510, 210). That is, the center coordinates of the region of interest determined in S403 are (510, 210). In the case of FIG. 5D, the coordinates of the point specified by the pointer 530 are the center coordinates of the region of interest determined in S403. In the case of FIGS. 5(e) and (f), the representative point of the humanoid figure 540 or the irregularly shaped figure 550 is defined as the center of gravity or the center point of the circumscribed rectangle of the figure, and its coordinates are defined as the center point of the circumscribed rectangle of the figure. Set to center coordinates.

In S404, the image extraction unit 203 determines the size of a square indicating the region of interest based on the important position acquired in S402. In the example of FIG. 5C, the size of the region of interest frame 520 is set to 240×240 pixels, so in step S403, the image extraction unit 203 determines the size of the region of interest to be 240×240 pixels.

Furthermore, the image extraction unit 203 may extract regions of interest of different sizes depending on important positions in the input image. In the operation methods shown in FIGS. 5(d) to 5(f), the size is undetermined, so the size is determined according to the center coordinates (or important position) of the region of interest acquired in S403. At this time, if the input image is a shot of a scene with depth, the size of the object to be detected changes depending on the position on the image corresponding to the depth direction, so the size of the region of interest should also change accordingly. I can do it. Specifically, the closer the important position is to the front, the larger the rectangle size is, and the further the important position is to the back of the screen, the smaller the rectangle size is. This is because the detection accuracy of the analysis means is improved if the ratio between the size of the rectangle and the size of the object to be detected is constant. Further, if the size difference depending on the position on the image is not large, a constant attention area size may be applied regardless of the position. In this embodiment, the size of the region of interest determined in S404 is 240×240 pixels.

In S405, the image extraction unit 203 extracts from the input image a partial image corresponding to the region of interest of the size determined in S404 and centered on the coordinates determined in S403. Since there is a high possibility that an object will be comprehensively captured in the partial image corresponding to the region of interest, the accuracy of object detection processing can be improved.

In S406, the output unit 205 outputs the partial image corresponding to the region of interest to the output unit (display device). That is, the output unit 205 displays the region of interest determined in S405 in a manner that is visible and superimposed on the image display unit 510. Specifically, the area of interest is displayed on the display means in a specific manner (color, shape, transmittance, line, etc.) that allows it to be identified. The superimposed state here is the same as or similar to the attention area 520 shown in FIG. 5(c). By indicating where in the input image the region of interest is set, convenience for the user is improved.

In S407, the operation unit 105 determines whether the user has instructed a predetermined process. That is, the operation unit 105 determines whether the automatic placement button 504 has been pressed by the user. The automatic placement button 504 serves as a trigger for performing processing for setting a plurality of partial images for the input image. If the automatic placement button 504 is pressed, the process advances to S408. Unless the automatic placement button 504 is pressed, the processes of S408 and S409 are not executed.

In S408, the image extraction unit 203 sets a plurality of partial images so as to encompass the entire area of the image. Specifically, the input image is divided as shown in FIGS. 3(a) and 3(b). By inputting these partial images to a predetermined analysis process, the position of the specific object in the input image is detected.

In S409, the output unit 205 outputs the plurality of partial areas set for the input image. Specifically, the output unit 205 superimposes and displays the one or more partial images set in S408 on the image display unit 510 in a visually recognizable format. At this time, the plurality of partial images set for the input image and the region of interest drawn in S406 may be displayed in different manners. FIG. 5G shows how six 240×240 pixel partial images are displayed in S409 on the image display section 510 in a format in which dotted lines 560 indicate the boundaries of the regions. When the OK button 501 is pressed by the user, the image extraction unit 203 extracts each of the partial images determined through the processing up to this point and inputs them to the object detection unit 202.

In S410, the object detection unit 202 detects information indicating the position of a person's face as a specific object from a plurality of partial images obtained by dividing the input image acquired by the image acquisition unit 201.

As explained above, according to the first embodiment, when extracting a target partial image in object detection processing for an input image, since an important position is included in at least one partial image, an object at an important position Detection accuracy is improved.

[Second embodiment]
In the first embodiment, the process of determining a region of interest corresponding to an important position on an image and the process of setting one or more partial images to encompass the entire area of the image have been described. In the second embodiment, a process of setting other partial images based on one of the partial images determined from an important position on an image will be described. In the following description, the same reference numerals will be used for the same components as in the first embodiment, and the description will be omitted. As for the hardware configuration, the configuration shown in FIG. 1 is used as in the first embodiment.

FIG. 6 is a flowchart illustrating processing performed by the information processing apparatus 100 in this embodiment. The processing shown in the flowchart of FIG. 6 is executed by the CPU 101 of FIG. 1, which is a computer, according to a computer program stored in the storage device 106. Components that are common to the flowchart shown in FIG. 4 are designated by the same reference numerals as those in FIG. 4, and a description thereof will be omitted. FIG. 7 is an example of an operation screen for explaining the partial image setting method according to this embodiment.

In S601, the CPU 101 determines whether the user has performed an operation to add an important position, that is, whether the important position addition button 701 in FIG. 7A has been pressed. If it is determined that no important position has been specified, the process advances to step S408, and the image extraction unit 203 sets one or more partial images so as to cover the entire area of the image.

The image extraction unit 203 sets the arrangement of partial images based on one of the following functions. (1) A function specified in advance regarding the position and size of the object in the input image. (2) A function calculated based on the coordinates and size of one or more objects specified on the input image by the user. (3) A function calculated based on the coordinates and size of the object detected by executing a predetermined analysis process on images acquired within a predetermined period of time. Each function will be explained below.

An example of the operation screen is shown in FIG. 7(b). FIG. 7B shows an example of a method for arranging partial images suitable for an input image having a depth in which the size of the subject differs in the foreground (at the bottom of the image) and in the back (at the top of the image). The boundaries of the partial images are indicated by dotted lines 710. The size of the partial image that changes in the depth direction is (1) specified in advance using a function based on information indicating how the size of the subject changes depending on the coordinates within the image. For example, (2) a function is generated from size information obtained by the user specifying two or more sizes of the object to be analyzed, one in the foreground and the other in the background, on an operation screen (not shown). Alternatively, (3) it is also possible to perform simple analysis processing on the input image at a predetermined timing and generate a function from information about the detected object. This function is expressed as follows, assuming that the center coordinates of the square analysis processing area on the image are (x, y) and the length of one side is w pixels.
w=a×x+b×y+c...(Formula 1)
is defined as In the example of FIG. 7, there is almost no change in the size of the subject in the horizontal direction (x-axis direction), so a in Equation 1 is 0. Since the size of the subject changes in the vertical direction (y-axis direction), the length of one side is w pixels depending on the y-coordinate of the center of the analysis processing area.
w=0.48×y+68.6...(Formula 2)
It is assumed that a function is defined.

If it is determined in S601 that an operation to add an important position has been performed, the output unit 205 executes the processes in S402 to S406 as described in the first embodiment, and adds the key position corresponding to the specified important position. Output the region of interest to a display device.

In S602, the CPU 101 determines whether the user performs an operation to add or change important positions. If there is an operation to add or change an important position here, the processes of S402 to S406 are repeated. FIG. 7C shows an example of an operation screen when two important positions are designated and corresponding regions of interest 711 and 712 are drawn. If the determination in S602 is negative, the process proceeds to S603 after determining whether or not automatic placement is specified in S407.

In S603, the CPU 101 determines whether a reference region of interest has been specified by the user. The reference region of interest is designated by the user, for example, if one of the two regions of interest shown in FIG. 7C is in a selected state. If there is no reference region of interest specified, the process advances to step S408, where the image extraction unit 203 arranges one or more regions of interest so as to encompass the entire region of the image, and extracts an image corresponding to each region of interest. . FIG. 7(d) shows that a dotted line 710 indicating the boundary of the region of interest arranged by the process of S408 has been added to FIG. 7(c). On the other hand, FIG. 7E shows an example of an operation screen in which the attention area 714 is in a selected state. The boundary of the selected region of interest 714 is displayed as a double line, which is a different display format from the solid lines of the other region of interest frames 711, so that it can be easily identified. If the user presses the automatic placement button 504 while one of the attention areas is selected in this way, the CPU 101 determines in S603 that the reference attention area has been specified, and the process advances to S604.

In S604, the image extraction unit 203 sets one or more partial images to encompass the entire area of the image, using the region of interest as a reference. Here, the image extraction unit 203 uses the above-described function to determine the positions and sizes of other partial images based on the selected region of interest 714. FIG. 7F shows an example in which the attention area set in S604 is drawn in S409 and displayed as a dotted line 720 on the operation screen. Here, the region of interest 714 has center coordinates (480, 230) and a square side length of 179 pixels. In a region having the same y coordinate as the region of interest 714, regions of interest of the same size are arranged without overlapping (721).

Next, the position and size of a region of interest 722 one step above another region of interest 721 arranged at the same y coordinate as the region of interest 714 are calculated. First, the y-coordinate of the upper side of the attention area 714 is calculated as 141 from the position and size of the attention area 714. This y-coordinate becomes the lower side of the region of interest 722 one level higher. When the y-coordinate of the center point of the region of interest one step above is y1, and the length of one side of the square representing the size is w1, from the above equation 2, w1=0.48×y1+68.6. Furthermore, since the y coordinate of the lower side is calculated as 141, y1+w1/2=141. From these two equations, y1=86 and w1=110 are calculated. A region of interest 715 is arranged at the same x coordinate as the y coordinate and size of the region of interest 714 serving as a reference. Next, attention areas of the same size are set on the left and right without overlapping (722).

The other rows are determined in the same way, the common y coordinate of the attention area 723 placed at the top of the image is -3, the size is 67, the common y coordinate of the attention area 724 placed at the bottom step is 292, The size is 465.

In this example of the operation screen, the boundary of the reference area 714 is displayed as a double line, which is in a different format so that it can be distinguished from other areas of interest. In addition, the attention area placed by the user by the important position addition operation (solid line 711, double line 714) and the attention area placed by automatic placement (dotted line 720) are displayed in different display formats, so they can be easily distinguished. . The display format of the attention area is not limited to this, and the entire area may be displayed in the same format. Furthermore, it is also possible to change the display format of boundaries only for areas that need to be identified during editing, such as changing, deleting, or adding an attention area, or to color the area including the attention area.

As explained above, according to the second embodiment, in object detection processing for an input image, it is possible to extract a plurality of partial regions based on important positions on the image specified by the user. As the detection accuracy improves, the extraction efficiency of other partial regions also improves.

[Third embodiment]
In this embodiment, among the processing steps in the first and second embodiments, a process (S408) of setting one or more partial images to encompass the entire area of the input image will be described.

In order to input images extracted from the input image in partial image units to the object detection unit 202, one or more square partial images are set as the input image. As shown in the second embodiment, if a reference area of interest is specified among the areas of interest corresponding to important positions on the image, other partial images can be arranged using that area as a starting point. good. However, when a reference region of interest is not designated, there are many possible ways to arrange partial images. Therefore, in this embodiment, how to determine a reference region of interest when there is no information regarding important positions on an image will be described using FIG. 8.

Regions of interest 811, 821, 831, and 841 indicated by double lines in FIGS. 8(a) to 8(d) each indicate a partial image initially placed as a reference. The positions and sizes of other partial images are calculated based on the position of this partial image. Boundaries of other partial images are indicated by dotted lines 810, 820, 830, and 840. FIG. 8A shows an example in which partial images have the same size regardless of the position on the image, and the reference region of interest position is set at the center of the image. FIG. 8B is an example in which the partial images have the same size regardless of the position on the image, and the reference region of interest position is set at the bottom center of the image. FIG. 8C is an example in which partial images of different sizes are set depending on the position on the image. Specifically, this is an example in which the partial image size changes depending on the y-coordinate of the input image, and the reference region of interest position is set at the center of the image. FIG. 8D is also an example in which partial images of different sizes are set depending on the position on the image. Specifically, this is an example in which the partial image size changes depending on the y-coordinate of the input image, and the reference region of interest position is set at the bottom center of the image. It would be good if these four types of arrangement methods could be used depending on the characteristics of how the subject appears. Therefore, in order to easily estimate the characteristics of how a subject appears, the coefficients of Equation 1 used in the second embodiment are used as camera setting information and information regarding depth.

FIG. 8E is an example of an input image when a person 850 is photographed from a high position looking down almost directly below. For such images, it is assumed that the object to be detected is often seen at the center of the image, so the reference region of interest is placed at the center of the image. Furthermore, since the depth is small, the arrangement pattern shown in FIG. 8A is selected assuming that there is no change in size depending on the position on the image.

FIG. 8(f) is an example of an input image when a person is photographed from almost directly sideways in a place with a small depth. There is no significant difference in the size of the person 860 who is slightly in the foreground and the person 861 who is further behind him in the image. For such an image, the reference region of interest is placed at the bottom center, which can reduce the number of regions of interest that protrude outside the image area, rather than placing it at the center of the input image. In other words, the arrangement pattern shown in FIG. 8(b) is selected.

FIG. 8(g) is an example of an input image taken from a high position looking down diagonally. There is depth, and a person 871 appearing in the back (at the top of the image) is smaller in size than a person 870 appearing near the center. In addition, although a person is shown in the foreground (at the bottom of the image), if the person approaches more than a certain level, only a portion of the person, such as person 872, will appear larger at the bottom of the image, and the person will be removed from the object of interest. Therefore, for such an image, the reference region of interest is placed at the center of the input image. Since there is depth, it is necessary to change the size of the region of interest depending on the position on the image, so the arrangement pattern shown in FIG. 8(c) is selected.

FIG. 8(h) is an example of an input image taken from a lower position than FIG. 8(g). There is depth, and the person 860 in the foreground is captured from the front and appears at the bottom of the image. For such an image, the arrangement pattern shown in FIG. 8(d) in which the reference region of interest is arranged at the bottom center of the input image is selected.

Camera setting information is used as the first element for simply estimating how the subject shown in FIGS. 8(e) to 8(h) appears. For example, here, the angle of depression, which is the angle between the horizontal plane and the optical axis, is used to indicate the setting state of the camera. The depression angle information of the camera can be obtained by the user inputting it into the information processing apparatus 100 through the operation unit 105 at the time of installation, and also by inputting information generated by the camera's sensor function etc. to the information processing apparatus 100 through the communication I/F 103. In some cases. Furthermore, the information may include height information from the floor or the ground at the position where the camera is attached. The height information can be obtained by the user inputting it into the information processing apparatus 100 through the operation unit 105 at the time of installation.

Information regarding depth is used as the second element for easily estimating how the subject appears. Here, the coefficient b of Equation 1 (w=axx+b×y+c, length of one side of the analysis processing area at the center coordinates (x, y), w pixels) described in the second embodiment is used. The larger b is, the larger the change in the subject size on the image in the y-axis direction (depth direction) is. Table 1 shows an example of determining an array pattern that is a combination of the position of the reference analysis processing area and the presence or absence of a change in the size of the analysis processing area, depending on two conditions: the value of b and the camera depression angle. In Table 1, the value of b is, for example, 0≦b<0.5 for “small” and 0.5≦b≦1 for “large”. If this value cannot be obtained, it is marked as "unknown". Further, for the camera depression angle, for example, "small" is less than 20 degrees, "large" is 20 degrees or more, and if depression angle information cannot be obtained, it is treated as "unknown".

By predetermining the arrangement pattern of the analysis processing area based on the conditions in this way, it becomes possible to set the analysis processing area more suitable for the input image.

[Other embodiments]
The present invention provides a system or device with a program that implements one or more functions of the embodiments described above via a network or a storage medium, and one or more processors in a computer of the system or device reads and executes the program. This can also be achieved by processing. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

The embodiments described above are merely examples of implementation of the present invention, and the technical scope of the present invention should not be construed as limited by these embodiments. That is, the present invention can be implemented in various forms without departing from its technical idea or main features.

100 Information Processing Device 201 Image Acquisition Unit 202 Object Detection Unit 203 Image Extraction Unit 204 Correction Unit 205 Output Unit 206 Storage Unit

Claims (15)

identification means for identifying important positions in the image;
Extracting means for extracting a region of interest including the important position from the image;
Setting means for setting a plurality of partial images to encompass the entire area of the image;
An information processing device comprising: a partial image of the image corresponding to the region of interest; and a detection means for detecting an object from the plurality of partial images. The information processing apparatus according to claim 1, wherein the setting means sets the plurality of partial images based on the region of interest. The information according to claim 1, wherein the identifying means identifies the important position based on the position of the object detected by executing a predetermined analysis process on an image acquired at a predetermined time. Processing equipment. further comprising operation means for accepting a position in the image specified by a user;
The information processing apparatus according to claim 1, wherein the specifying unit specifies a position on the image designated by the user as the important position. 5. The information processing apparatus according to claim 4, wherein the operating means receives the position by arranging a predetermined figure on the image displayed on the display means. 2. The information processing apparatus according to claim 1, wherein the extraction means extracts the attention area of a different size depending on the important position in the image. The information processing apparatus according to claim 1, further comprising output means for outputting the region of interest to a display means. 8. The information processing apparatus according to claim 7, wherein the output means displays the plurality of set partial images and the region of interest in different manners. 8. The information processing apparatus according to claim 7, wherein the output means displays the region of interest and a region including an object detected from the plurality of partial images in different manners. The information processing apparatus according to claim 1, wherein the detection means detects the object by performing a predetermined analysis process on each of the partial images of a predetermined size. The identifying means identifies a predetermined position on the image as the important position,
3. The information processing apparatus according to claim 2, wherein the setting means sets the partial images so that one of the plurality of partial images is a partial image including the important position. The information processing apparatus according to claim 1, wherein the setting means sets the partial images to have different sizes depending on their positions in the image. The setting means is based on (1) a function specified in advance regarding the position and size of the object in the image, and (2) the coordinates and size of one or more of the objects specified on the image by the user. (3) a function calculated based on the coordinates and size of the object detected by executing a predetermined analysis process on images acquired within a predetermined time; 13. The information processing apparatus according to claim 12, wherein the arrangement of the partial images is set based on one of the following. an identification step of identifying important positions in the image;
an extraction step of extracting a region of interest including the important position from the image;
a setting step of setting a plurality of partial images to encompass the entire area of the image;
An information processing method comprising: a partial image of the image corresponding to the region of interest; and a detection step of detecting an object from the plurality of partial images. computer,
identification means for identifying important positions in the image;
Extracting means for extracting a region of interest including the important position from the image;
Setting means for setting a plurality of partial images to encompass the entire area of the image;
A program for functioning as an information processing apparatus, comprising a partial image of the image corresponding to the region of interest, and a detection means for detecting an object from the plurality of partial images.

Images