Mobile body detection device, mobile body detection system, mobile body detection method, and program storage medium

Abstract

In order to increase detection accuracy for detecting, from a captured image, a mobile body to be detected that appears small in the captured image, this mobile body detection device comprises an acquisition unit, a classification unit, an adjustment unit, and an output unit. The acquisition unit acquires a captured image of a detection area. The classification unit classifies, by using a pixel value and a threshold for classification determination, each of the pixels constituting the acquired captured image as either a foreground pixel that reflects the mobile body or a background pixel that is any pixel other than the foreground pixel. If a foreground / background ratio of the captured image is higher or lower than a predetermined allowable range, the adjustment unit adjusts the threshold for classification determination so that the foreground / background ratio falls within the allowable range. The foreground / background ratio is a ratio of the number of pixels classified as the foreground pixels to the number of pixels classified as the background pixels. When the foreground / background ratio is within the allowable range, the output unit outputs detection information indicating a position in the captured image of the pixels classified as the foreground pixels.

Description

MOBILE BODY DETECTION DEVICE, MOBILE BODY DETECTION SYSTEM, MOBILE BODY DETECTION METHOD, AND PROGRAM STORAGE MEDIUM

[0001] The present disclosure relates to a moving body detection device, a moving body detection system, a moving body detection method, and a program storage medium that detect a moving body using a captured image.

[0002] One of the technologies used by a surveillance system to monitor a surveillance area is a technology for automatically detecting a target moving object from images of the surveillance area captured by a camera. Patent Document 1 (JP 2009-110240 A) discloses a technology for determining whether each pixel of each frame image constituting a moving image is classified as foreground or background.

[0003] JP 2009-110240 A

[0004] Because the monitoring area of ​​a monitoring system is wide, it may be necessary to detect a moving object moving far away from the camera, for example, one kilometer away, from the image captured by the camera. In such cases, if the moving object moving far away from the camera is a large moving object such as an airplane and its shape is captured in the captured image, it can be detected from the captured image using object recognition technology or the like. However, there are cases where the moving object is a small moving object, such as an unmanned aerial vehicle (small unmanned aircraft), also known as a drone or UAV (Unmanned Aerial Vehicle). In such cases, the moving object in the far away area captured in the captured image is small and its shape is unclear, which makes it difficult to accurately detect the moving object from the captured image.

[0005] The present disclosure has been devised to solve the above-mentioned problems. That is, a main objective of the present disclosure is to provide a technology for improving the detection accuracy of detecting a moving object that appears small in a captured image from the captured image.

[0006] In order to achieve the above object, one aspect of the moving body detection device according to the present disclosure includes: an acquisition unit that acquires a captured image of a detection area; a classification unit that executes a classification process to classify each pixel constituting the captured captured image into either a foreground pixel that shows a moving body or a background pixel other than the moving body using a pixel value and a threshold for classification determination; an adjustment unit that, when a foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels, is higher or lower than a predetermined allowable range, adjusts the threshold for classification determination so that the foreground-background ratio falls within the allowable range; and an output unit that, when the foreground-background ratio is within the allowable range, outputs detection information that indicates the position in the captured image of the pixel classified as the foreground pixel.

[0007] In addition, one aspect of the moving object detection system according to the present disclosure includes: an imaging device that captures an image of a detection area; and the moving object detection device according to the present disclosure as described above that uses an image captured by the imaging device.

[0008] Furthermore, one aspect of the moving body detection method according to the present disclosure includes: acquiring a photographed image of a detection area by a computer; performing a classification process for classifying each pixel constituting the acquired photographed image into either a foreground pixel showing a moving body or a background pixel using pixel values ​​and a classification threshold; if a foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels, is higher or lower than a predetermined allowable range, adjusting the classification threshold so that the foreground-background ratio falls within the allowable range; and outputting detection information indicating the position in the photographed image of the pixel classified as the foreground pixel if the foreground-background ratio is within the allowable range.

[0009] Furthermore, in one aspect, the program storage medium according to the present disclosure stores a computer program that causes a computer to execute the following processes: a process of acquiring a captured image of a detection area; a process of performing a classification process to classify each pixel constituting the acquired captured image into either a foreground pixel that reflects a moving object or other background pixels using pixel values ​​and a threshold for classification determination; a process of adjusting the threshold for classification determination if a foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels, is higher or lower than a predetermined allowable range so that the foreground-background ratio falls within the allowable range; and a process of outputting detection information indicating the position in the captured image of the pixel classified as a foreground pixel if the foreground-background ratio is within the allowable range.

[0010] According to the present disclosure, it is possible to improve the detection accuracy of detecting a moving object that is a detection target and appears small in a captured image from the captured image.

[0011] 1 is a diagram illustrating an example of a moving body detection system according to the present disclosure. FIG. 2 is a diagram illustrating an example of a configuration of a moving body detection device according to the present disclosure. FIG. 3 is a diagram schematically illustrating an example of the imaging range of an imaging device viewed from the horizontal direction. FIG. 4 is a diagram schematically illustrating an example of the imaging range of an imaging device viewed from the zenith direction. FIG. 5 is a diagram illustrating an example of the form of a captured image. FIG. 6 is a diagram illustrating an example of display of detection information. FIG. 7 is a diagram illustrating an example of the execution timing of a process for updating a background model. FIG. 8 is a diagram illustrating another example of the execution timing of a process for updating a background model. FIG. 9 is a flowchart illustrating an example of the operation of a moving body detection device according to the present disclosure. FIG. 10 is a diagram illustrating an example of a segmented region in a captured image. FIG. 11 is a diagram illustrating another example of a segmented region in a captured image. FIG. 12 is a diagram illustrating another example of the configuration of a moving body detection device according to the present disclosure. FIG. 13 is a diagram illustrating yet another example of the configuration of a moving body detection device according to the present disclosure. FIG. 14 is a diagram illustrating yet another example of the configuration of a moving body detection system according to the present disclosure. FIG. 15 is a flowchart illustrating another example of the operation of a moving body detection device according to the present disclosure.

[0012] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0013] First Embodiment A mobile object detection system according to a first embodiment of the present disclosure is a system that detects a mobile object in a detection area using an image captured by a camera. The system includes a camera 3 and a mobile object detection device (hereinafter also referred to as a detection device) 5, as shown in FIG. 1 . In the first embodiment, the mobile object detection system 1 is applied to a surveillance system. An example of a surveillance system to which the mobile object detection system 1 is applied is a system that monitors the airspace above and surrounding areas of important facilities such as airports and nuclear power plants. The surveillance area includes no-fly zones where unmanned aerial vehicles are prohibited by law. Because the mobile object detection system 1 is applied to such a surveillance system, the detection area of ​​the mobile object detection system 1 is the surveillance area of ​​the applicable surveillance system. Furthermore, the target mobile object to be detected by the surveillance system is not limited to a predetermined area determined by the operator of the surveillance system, taking into account, for example, the type of facility being monitored and its surrounding environment. However, in the following description, it is assumed that an aerial vehicle is set as the target mobile object to be detected. Examples of flying objects, which are moving bodies that are set as detection targets, include unmanned aerial vehicles (unmanned aircraft) that can be flown by remote control or automatic piloting, such as unmanned airplanes, unmanned rotorcraft, and unmanned airships, as well as those flown by people using specific aviation equipment, such as balloons, hang gliders, and paragliders.

[0014] The camera 3 is a device that captures video of the detection area. The installation location of the camera 3 may be any location that allows it to capture video of the detection area, and is determined appropriately by the operator of the surveillance system, taking into consideration the presence or absence of obstacles that may interfere with capture and the ease of installation. The number of camera devices 3 installed is determined appropriately taking into consideration the width of the field of view of the camera 3 and the size of the detection area, and may be one or more, with no limit on the number. However, when multiple camera devices 3 are installed, these camera devices 3 are installed side by side. Furthermore, the camera directions of the multiple camera devices 3 are different from each other, and the respective camera directions are set so that by combining the images captured by the multiple camera devices 3, an image of the entire detection area that can be captured from the installation location can be obtained.

[0015] The detection device 5 is a computer device and is communicatively connected to the imaging device 3. As shown in FIG. 2 , the detection device 5 includes an arithmetic unit 50 and a storage device 40. The storage device 40 includes a storage medium for storing data and a computer program (hereinafter also referred to as a program) 41. There are multiple types of storage devices, such as magnetic disk drives and semiconductor memory devices. Furthermore, there are multiple types of semiconductor memory devices, such as RAM (Random Access Memory) and ROM (Read Only Memory). A computer device may be equipped with multiple types of storage devices depending on their intended use, but these storage devices will be collectively referred to as the storage device 40 without distinction. Furthermore, the type and number of storage devices 40 included in the detection device 5 are not limited, and a description thereof will be omitted. The detection device 5 may also be connected to a database 6, which is a storage device. In this case, the detection device 5 may write information to the database 6 or read information from the database 6. However, to avoid complication of explanation, such a case will not be described here.

[0016] The arithmetic device 50 is configured with a processor such as a CPU (Central Processing Unit) or a GPU (Graphics Processing Unit). The arithmetic device 50 can have functions based on a program 41 by reading and executing the program 41 stored in the storage device 40. Here, the arithmetic device 50 has a functional unit that detects a moving object from an image captured by the image capturing device 3.

[0017] The image captured by the camera 3 captures a range of distances from close to the camera 3 to far away. Here, it is assumed that a moving object to be detected, located in a range from close to medium distances from the camera 3, is visible (i.e., its shape is recognizable) in the image captured by the camera 3. Various object detection technologies for detecting such visible moving objects from captured images, such as YOLO (You Only Look Once) and pattern recognition technology, have been proposed. The detection device 5 has a function for detecting such visible moving objects from captured images. However, the detection technology is not limited and may be appropriately set by, for example, the designer of the detection device 5, and therefore, a description thereof will be omitted. The following mainly describes a method for detecting moving objects that are small and difficult to see in a captured image (moving objects whose shape is not clearly visible) from the camera. Here, the term "far away" is defined taking into account the distance from the installation location of the camera to the farthest area of ​​the detection area, and refers to an area in the captured image where the moving object is difficult to see. Furthermore, a medium distance from the camera refers to a region that includes the middle between the distant location and the location where the camera is installed, as described above, and is a region in which a moving object can be seen in the captured image. For example, a full high-definition image (full HD (High Definition) image) has an image size (1920 x 1080) of 1920 horizontal and 1080 vertical pixels. In this full HD image, a moving object located at a medium distance from the camera and captured at an image size of approximately 100 x 100 pixels is visible. A moving object captured at this image size can be detected from the captured image using object detection technology such as YOLO, as described above. In contrast, the same moving object is difficult to see in a captured image if it is captured at a distance and only a few pixels x a few pixels. Such moving objects are difficult to detect from the captured image using object detection technology such as YOLO.

[0018] The detection device 5 includes an acquisition unit 51, a classification unit 52, an adjustment unit 53, an update unit 54, and an output unit 55 as shown in Fig. 2 as functional units for detecting a moving object that is a detection target and difficult to see from a captured image. As mentioned above, there may be a plurality of image capture devices 3, but the processing executed on the captured images of the image capture devices 3 is the same regardless of the number of installed image capture devices 3. For this reason, in the following explanation, even in the case where the detection device 5 is connected to a plurality of image capture devices 3 and processes the captured images of each of the plurality of image capture devices 3, redundant explanation of the processing executed on the captured images will be omitted.

[0019] The acquisition unit 51 acquires captured images (video) from the image capture device 3. The acquisition unit 51 also stores (registers) the acquired captured images in the storage device 40. Note that the method for acquiring captured images from the image capture device 3 and the method for storing the acquired captured images in the storage device 40 are not limited, and therefore a description of these methods will be omitted.

[0020] The classification unit 52 performs a pixel classification process on the captured image acquired by the acquisition unit 51. Here, since the captured image is a moving image, the captured image is composed of multiple frame images (still images) arranged in chronological order. Of the multiple frame images that make up the captured image, the classification process is performed as follows on the pixels of the frame images to be processed. The classification unit 52 first extracts frame images to be processed from the captured image. The frame images to be processed are appropriately set by the device designer or the like and are not limited to these, and may be, for example, all frame images that make up the captured image. However, here, taking into consideration the processing capacity and efficiency of the detection device 5, a predetermined number of frame images in chronological order are extracted as frame images to be processed from the multiple frame images arranged in chronological order.

[0021] The classification process performed by the classification unit 52 classifies each pixel constituting the frame image to be processed into either a foreground pixel or a background pixel. Foreground pixels are pixels that show moving objects, and background pixels are all other pixels. Here, a method using pixel values, a background model, and a classification threshold is adopted as a method for classifying pixels of a captured image into either foreground pixels or background pixels. Pixel values ​​are quantified values ​​of the brightness and color (R (red), G (green), B (blue)) of a pixel. A background model is a model (quantified) of the background of a captured image using the pixel values ​​of each pixel in the captured image. Examples of background models include a background model using a Gaussian Mixture Model (GMM) and a background model using the average pixel value.

[0022] The background model using a Gaussian mixture model is a model that includes information that expresses the distribution state of pixel values ​​(colors (R, G, B)) of each pixel in the background of a captured image using a Gaussian mixture model. In other words, the background model using a Gaussian mixture model includes information that associates position information of each pixel in the captured image with a Gaussian mixture model of the pixel values ​​of the background corresponding to that pixel. Furthermore, the background model using the average pixel values ​​is a model that includes information on the average pixel value of each pixel in the background of a captured image, calculated as follows: The average pixel value here is a value obtained by averaging the pixel values ​​of pixels at the same position in a plurality of frame images in a time series that constitute the captured image. The background model using the average pixel values ​​includes information that associates position information of each pixel in the captured image with the above-mentioned average pixel value of the background corresponding to that pixel. Information on such a background model is stored in the storage device 40. Note that the background model employed here may be any model that digitizes the background in the captured image using the pixel values ​​of each pixel in the captured image, and the type of background model employed is not limited. In the following description, the Gaussian mixture model and average value information of pixel values ​​associated with position information of pixels included in the background model will also be simply referred to as information of background pixel values ​​in the background model.

[0023] The classification unit 52 classifies each pixel in the frame image to be processed as either a foreground pixel or a background pixel using the background model described above, a classification threshold, and the pixel values ​​of the pixels in the frame image to be processed, as follows: In the description of the first embodiment, the classification threshold is also referred to as the "classification threshold Th."

[0024] For example, the classification unit 52 compares the pixel value of a target pixel selected from among multiple pixels in the frame image to be processed with information about the pixel value of the background associated with the pixel at the same position in the background model, and calculates the degree of deviation. Furthermore, the classification unit 52 compares the calculated degree of deviation with a classification threshold Th. If the degree of deviation is equal to or less than the classification threshold Th, the pixel value of the target pixel is close to the pixel value information of the background, and therefore the target pixel reflects the background, and classifies the target pixel as a background pixel. On the other hand, if the degree of deviation is greater than the classification threshold Th, the pixel value of the target pixel deviates from the pixel value information of the background, and therefore the target pixel is determined not to be background, and classifies the target pixel as a foreground pixel. The classification unit 52 sequentially selects pixels as target pixels from the multiple pixels in the frame image to be processed, for example, in raster order, and performs the classification process described above on all pixels in the frame image to be processed. Information resulting from the classification process by the classification unit 52 is stored in the storage device 40, for example, as a binary image. This binary image is associated with frame identification information (for example, a frame number) that identifies the corresponding frame image.

[0025] The detection device 5 detects small, difficult-to-see moving objects as detection targets from the captured image through the pixel classification process described above. However, depending on the background of the captured image, the target moving object may not be detected accurately unless the classification threshold Th is adjusted. For example, if a forest or a beach is captured in the captured image, the classification process by the classification unit 52 may classify pixels reflecting trees and waves as foreground pixels (pixels reflecting moving objects), even though trees swaying in the wind and waves are not detection targets. In such cases, the foreground pixels reflecting trees and waves that are not detection targets become noise in the detection of the target moving object, hindering accurate detection of the target moving object.

[0026] Therefore, the present inventors have devised a method for adjusting the classification determination threshold Th from the following perspective. That is, when a moving object to be detected is detected from a frame image (captured image) using pixel classification processing by the classification unit 52, the moving object to be detected is small enough to be difficult to visually recognize. For this reason, most of the pixels in the frame image (i.e., pixels other than those that show the moving object to be detected) may be classified as background pixels. Using this, the classification determination threshold Th is adjusted as follows. Note that several methods have been proposed for setting the initial value of the classification determination threshold Th, and the setting methods are not limited here, so a description thereof will be omitted.

[0027] That is, the adjustment unit 53 adjusts the classification threshold Th when it determines that such adjustment is necessary. The classification result of the classification unit 52 for the frame image to be processed is used to determine whether adjustment of the classification threshold Th is necessary. That is, the following foreground-to-background ratio is used to adjust the classification threshold Th. The foreground-to-background ratio is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels in the frame image (captured image) to be processed. If such a foreground-to-background ratio falls outside a predetermined allowable range (i.e., is higher or lower than the allowable range), it is determined that adjustment of the classification threshold Th is necessary. The allowable range is set using the number of pixels expected to capture the target moving object to be detected in the frame image (i.e., the number of pixels corresponding to the image size of the target moving object captured in the frame image). One method for setting the allowable range is the following setting method, which uses information related to the shooting range (field of view) of the image capture device 3.

[0028] Fig. 3 shows a schematic diagram of an example of the imaging range of the imaging device 3 (i.e., the area shown in the captured image) as viewed from a direction along the ground surface. Fig. 4 shows a schematic diagram of an example of the imaging range of the imaging device 3 as viewed from the zenith. Fig. 5 shows a simplified example of an image captured by the imaging device 3. In the following description, the direction along the ground surface will also be referred to as the horizontal direction, and the direction perpendicular to the ground surface will also be referred to as the vertical direction.

[0029] In the method for setting the tolerance described below, the number of pixels in the horizontal direction of the rectangular frame image (captured image) shown in Fig. 5 is represented by Iw, and the number of pixels in the vertical direction is represented by Ih. As shown in Fig. 3, the vertical angle of view of the image capture device 3 is 2θ, and the distance from the image capture device 3 to the moving object 9 to be detected is approximately r (m (meters)). Furthermore, as shown in Fig. 4, the horizontal angle of view of the image capture device 3 is 2φ. Furthermore, the size of the moving object 9 is assumed to be such that the vertical length (height) is h (m (meters)) and the horizontal length (width) is w (m (meters)).

[0030] In such a case, the vertical length (height) H of the imaging range captured by the imaging device 3 is "2×r×tanθ" (m (meters)), so the number of vertical pixels hp of the moving object 9 captured in the frame image can be calculated by the formula hp = (Ih × h) ÷ (2×r × tanθ). Furthermore, the horizontal length (width) W of the imaging range captured by the imaging device 3 is "2×r × tanφ" (m (meters)), so the number of horizontal pixels wp of the moving object 9 captured in the frame image can be calculated by the formula wp = (Iw × w) ÷ (2×r × tanφ). Thus, the number of pixels corresponding to the image size of the moving object 9 to be detected captured in the frame image is "wp × hp." Furthermore, the ratio R of the number of pixels of the moving object 9 to the total number of pixels in the frame image (hereinafter also referred to as the foreground ratio of the detected object) can be calculated by the formula R = (wp × hp) ÷ (Iw × Ih). That is, the foreground ratio R of the detection target can be calculated using the pixel counts wp and hp calculated as described above and the pixel counts Iw and Ih of the frame image. The lower limit of the allowable range is set to α times the foreground ratio R of the detection target (0 < α < 1), and the upper limit of the allowable range is set to β times the foreground ratio R of the detection target (1 ≦ β). Note that the constants α and β are set appropriately depending on, for example, what appears as the background in the image captured by the image capture device 3, and are not limited to these values.

[0031] In this way, it is possible to set an acceptable range related to the adjustment of the classification threshold value Th. Another method for setting the acceptable range is, for example, to set a range of a predetermined width with the foreground ratio of the detection target as the middle. In this case, the background of the captured image is also taken into consideration.

[0032] If the foreground-background ratio in the frame image to be processed based on the classification result of the classification unit 52 is outside the above-mentioned allowable range (higher or lower than the allowable range), the adjustment unit 53 adjusts the threshold value Th for classification judgment so that the foreground-background ratio falls within the allowable range.

[0033] In other words, when the foreground-to-background ratio is higher than the allowable range, pixels that reflect objects other than the detection target are assumed to be classified as foreground pixels, which increases the likelihood of false detection of moving objects. Adjusting the classification threshold Th in this case can be considered an adjustment aimed at reducing false detection of moving objects. In other words, the classification threshold Th is changed so as to reduce the number of pixels classified as foreground pixels.

[0034] On the other hand, when the foreground-to-background ratio is lower than the allowable range, it is assumed that there are pixels that reflect the moving object to be detected but are not classified as foreground pixels, and therefore the number of moving object non-detections tends to increase. In this case, adjusting the classification determination threshold Th can be said to be an adjustment in the direction of suppressing the non-detection of moving objects. In other words, the classification determination threshold Th is changed in the direction of increasing the number of pixels classified as foreground pixels.

[0035] When adjusting the classification determination threshold value Th, the adjustment unit 53 calculates an updated classification determination threshold value Th by, for example, adding or subtracting a predetermined fixed value ΔTh to the classification determination threshold value Th so that the foreground-background ratio falls within an acceptable range.

[0036] In this way, when the classification determination threshold Th is updated, the classification unit 52 redoes the classification process on the frame image to be processed using the updated classification determination threshold Th. In other words, since the classification process using the pre-updated classification determination threshold Th may result in erroneous detection or non-detection of the moving object to be detected, the classification unit 52 redoes the classification process on the pixels in the frame image to be processed using the updated classification determination threshold Th. Then, using the number of background pixels and the number of foreground pixels classified by the redo classification process, the adjustment unit 53 again determines whether or not adjustment of the classification determination threshold Th is necessary, and if it is determined that adjustment of the classification determination threshold Th is necessary, the adjustment unit 53 updates the classification determination threshold Th in the same manner as described above.

[0037] As an example of adjusting the classification determination threshold Th, the adjustment unit 53 repeatedly updates the classification determination threshold Th by changing the classification determination threshold Th by a fixed value ΔTh, and the classification unit 52 performs classification processing. In this manner, the adjustment unit 53 adjusts the classification determination threshold Th so that the foreground-to-background ratio falls within an acceptable range. When the foreground-to-background ratio falls within the acceptable range through repeated processing by the adjustment unit 53 and the classification unit 52, the adjustment unit 53 determines that adjustment of the classification determination threshold Th is no longer necessary, and the adjustment of the classification determination threshold Th is completed. Furthermore, upon completion of this adjustment, the classification of the pixels of the frame image to be processed by the classification unit 52 is finalized. Information about the updated classification determination threshold Th adjusted by the adjustment unit 53 as described above is stored in the storage device 40 and is used in the pixel classification processing by the classification unit 52 for the next frame image to be processed. Furthermore, when the classification determination threshold Th is adjusted as described above, the classification unit 52 performs pixel classification processing on the same frame image multiple times. In such a case, for example, the information on the results of the classification process for the same frame image is overwritten and stored in the storage device 40, for example.

[0038] The pixel classification process by the classification unit 52 and the adjustment process of the threshold value Th for classification determination by the adjustment unit 53 as described above are processes related to the detection of a moving object to be detected from a frame image, and therefore, hereinafter, these processes will be collectively referred to as the detection process.

[0039] The update unit 54 is an example of an update means, and performs an update process of the background model used in the classification process of the classification unit 52. This background model update process takes into account changes in pixel values ​​of the background of the captured image over time. For example, when the image capture device 3 is capturing an outdoor scene, the brightness and color of the background change depending on the time of day due to changes in the sunlight. The update unit 54 updates the background model in accordance with such changes in the background over time. By updating the background model in this way, the reliability of the results of the classification process by the classification unit 52 can be increased.

[0040] The timing at which the update unit 54 performs the update process may be set appropriately, taking into account, for example, changes in the background within the shooting range of the image capture device 3. For example, the timing may correspond to the timing at which the classification unit 52 performs the classification process. In such a case, the update unit 54 extracts, from the captured images, a plurality of frame images in a time series that includes the same frame image as the frame image to be classified by the classification unit 52, and performs the background model update process using the extracted plurality of frame images. This update process may be performed, for example, after the detection process, including the classification process and adjustment process described above, as shown in FIG. 7 , or before the detection process. Alternatively, the update process may be performed in parallel with the detection process, as shown in FIG. 8 .

[0041] Regarding the background model updating process, methods have been proposed according to the type of background model, and the background model updating method is not limited here, and a description of the updating method will be omitted. However, a simple example is an update process in which a background model is generated using a plurality of frame images in a time series as described above, and the generated background model is used as the updated background model.

[0042] The output unit 55 outputs detection information regarding the detection of a moving object to be detected. That is, the detection information is information representing the position in the frame image (captured image) to be processed of pixels classified as foreground pixels by the above-described detection process (processing including the classification process by the classification unit 52 and the adjustment process by the adjustment unit 53). The pixels classified as foreground pixels are pixels classified as foreground pixels when the foreground-background ratio is within an acceptable range. That is, when the foreground-background ratio is within an acceptable range, noise suppression, which suppresses pixels representing moving objects that are not the detection target, such as trees or waves, from being classified as foreground pixels, and non-detection prevention, which prevents pixels representing the detection target moving object from being classified as background pixels, are achieved. Thus, the detection information can be said to be information representing the position of the detection target moving object in the frame image to be processed.

[0043] An example of an output destination of such detection information is the display device 7. The display device 7 is, for example, a display device provided in a monitoring system to which the mobile object detection system 1 including the detection device 5 is applied, and is a device used by a monitoring personnel in the monitoring system for monitoring. When the detection device 5 is directly connected to the display device 7, the output unit 55 also has a display control function for controlling the display operation of the display device 7, and the detection information is displayed on the display device 7 through the control operation of the output unit 55. An example of a display mode of the detection information is a mode in which the detection information is superimposed on the image captured by the imaging device 3. In such a case, for example, information that highlights the positions of pixels classified as foreground pixels in the captured image is associated with the detection information. The information that highlights the positions of foreground pixels is, for example, information such as a mark, letter, or symbol indicating the position of the foreground pixel, or information that makes the foreground pixel more noticeable in brightness, color, or blinking than the surrounding pixels, as shown in FIG. 6 .

[0044] Furthermore, when the detection device 5 and the display device 7 of the monitoring system are not directly connected, the output destination of the detection information from the output unit 55 may be, for example, a computer device 8 as indicated by the dotted line in Fig. 1. This computer device 8 is a computer device provided in the monitoring system to which the mobile object detection system 1 is applied, and controls the display of the display device 7. The detection information output from the output unit 55 to the computer device 8 is displayed on the display device 7 in the same manner as described above under the display control of the computer device 8.

[0045] By viewing the detection information displayed on such a display device 7, for example, a person in charge of monitoring the monitoring system can ascertain that a moving object to be detected is present in the distance in the monitoring area. This allows the person in charge of monitoring to perform operations such as monitoring the movement of the moving object to be detected in the distance by the detection device 5, or taking a photograph of the moving object to be detected in the distance with a telephoto camera, and to determine, for example, the danger posed by the moving object to be detected in the distance.

[0046] The detection device 5 has the configuration described above. Next, an example of the operation of the detection device 5 will be described with reference to FIG. 9 . The example of the operation of the detection device 5 described here is an example of the operation related to the detection of a distant moving object that is difficult to see from a captured image, and an explanation of the operation related to the process of detecting a moving object that is visible from a captured image as described above and the process of updating a background model will be omitted. Note that FIG. 9 is a flowchart showing an example of the operation of the detection device 5 to detect a moving object that is a detection object. FIG. 9 can also be said to be a diagram showing an example of a moving object detection method.

[0047] For example, first, the acquisition unit 51 of the detection device 5 starts acquiring captured images using the image capture device 3 (step 101 in FIG. 9 ). Thereafter, the classification unit 52 extracts a predetermined frame image to be processed from the captured image (step 102). Then, the classification unit 52 executes a classification process for each pixel in the frame image to be processed, using the pixel value, the background model described above, and the classification threshold Th to classify the pixel as either a foreground pixel or a background pixel (step 103).

[0048] After this classification process has completed classification of all pixels in the frame image to be processed, the adjustment unit 53 uses the classification results to calculate a foreground-to-background ratio for the frame image to be processed. Furthermore, the adjustment unit 53 determines whether the calculated foreground-to-background ratio is within a predetermined acceptable range (step 104). If the foreground-to-background ratio is within the acceptable range, the classification threshold Th is appropriate for the frame image to be processed. In other words, adjustment of the classification threshold Th is not necessary, and the classification of pixels by the classification unit 52 for the frame image to be processed is confirmed (step 106).

[0049] On the other hand, if the foreground-to-background ratio is higher or lower than the allowable range (outside the allowable range) in step 104, the classification determination threshold Th needs to be adjusted, and the adjustment unit 53 adjusts the classification determination threshold Th (step 105). For example, as described above, the adjustment unit 53 adds or subtracts a predetermined fixed value ΔTh to or from the classification determination threshold Th so that the foreground-to-background ratio falls within the allowable range, thereby calculating an updated classification determination threshold Th. Then, using the updated classification determination threshold Th, the classification unit 52 redoes the classification process for the pixels of the frame image being processed. In other words, the operation of step 103 is performed again by the classification unit 52. If the foreground-to-background ratio is still not within the allowable range even after redoing the classification process using the updated classification determination threshold Th, the adjustment process by the adjustment unit 53 and the classification process by the classification unit 52 are performed again, similar to those described above. By repeating these adjustment and classification processes, the foreground-to-background ratio falls within the allowable range, and the adjustment of the classification determination threshold Th by the adjustment unit 53 is completed. Then, when the foreground-background ratio is within the allowable range, the classification of the pixels in the frame image to be processed is determined.

[0050] Thereafter, the output unit 55 outputs detection information relating to the frame image to be processed (step 107). Thereafter, the arithmetic unit 50 determines whether a predetermined termination condition is met (step 108). The termination condition is a condition for terminating the process relating to the detection of the target moving object in the captured image, and an example of such a condition is that an end instruction is input to the detection device 5 by the person in charge of monitoring.

[0051] If the termination condition is not satisfied, the operations from step 102 onwards are repeatedly executed, and the same processing as described above is performed for the next frame image to be processed. If the termination condition is satisfied, the calculation device 50 terminates the processing related to the detection of the moving object to be detected in the captured image.

[0052] The detection device 5 constituting the moving object detection system 1 in the first embodiment can achieve the following effects by having the above-described configuration. Specifically, the detection device 5 is configured to classify each pixel in a captured image (frame image) as either a foreground pixel or a background pixel using a classification threshold Th, thereby enabling detection of a target moving object that appears in the captured image so small that it is difficult to see as a foreground pixel. Furthermore, the detection device 5 is configured to adjust the classification threshold Th in consideration of the image size of the target moving object to be detected in the captured image. By adjusting the classification threshold Th in this manner, the detection device 5 can suppress noise caused by moving objects that are not the detection target, such as trees or waves (i.e., noise in detecting the target moving object). This allows the detection device 5 to increase the likelihood (detection accuracy) of detection of a target moving object that appears in the captured image so small that it is difficult to see.

[0053] The detection device 5 also includes an update unit 54, which is configured to update the background model used in the pixel classification process. This allows the background model to be updated in accordance with changes in the background in the captured image, preventing the background model from deviating from the background conditions in the captured image. This configuration also contributes to improving the detection accuracy (detection precision) of the detection device 5's detection of the target moving object. By improving the detection accuracy of the target moving object in this way, the detection device 5 can increase the reliability of its detection of the target moving object.

[0054] Second Embodiment A second embodiment according to the present disclosure will now be described. In the description of the second embodiment, components having the same names as those in the description of the first embodiment will be denoted by the same reference numerals, and a duplicate description of the common parts will be omitted.

[0055] In the second embodiment, the moving object detection system 1, like the first embodiment, includes an imaging device 3 and a moving object detection device (detection device) 5. The detection device 5 also includes an acquisition unit 51, a classification unit 52, an adjustment unit 53, an update unit 54, and an output unit 55, which are functional units that detect moving objects that are difficult to see due to their small image size from captured images, similar to those in the first embodiment. However, the second embodiment differs from the first embodiment in the following points.

[0056] That is, in the first embodiment, a common classification threshold Th is set for the entire region of the frame image, whereas in the second embodiment, a classification threshold is set for each predetermined segment region of the frame image.

[0057] One example of the segmented regions of a frame image is partial regions obtained by dividing the frame image into a grid pattern, as shown in Fig. 10. Alternatively, the segmented regions of a frame image may be partial regions obtained by dividing the frame image by semantic segmentation, as shown in Fig. 11. That is, in the example of Fig. 11, the segmented regions of the frame image are regions divided into an image region showing mountains, an image region showing the sea, an image region showing the sky, and an image region showing vegetation. As such, there are multiple methods for setting segmented regions in a frame image, and an appropriate setting method may be adopted by the device designer or the like.

[0058] In the second embodiment, a threshold value for classification determination is set for each such divided area, and the classification unit 52 performs pixel classification processing for each divided area using the threshold value for classification determination set for that divided area.

[0059] Furthermore, an acceptable range for the foreground-background ratio is defined for each partitioned region. In the second embodiment, the adjustment unit 53 calculates the foreground-background ratio for each partitioned region using the pixel classification results obtained by the classification unit 52. Furthermore, when the foreground-background ratio for each partitioned region is outside the acceptable range for that partitioned region (when the foreground-background ratio is higher or lower than the acceptable range), the adjustment unit 53 determines that adjustment of the classification threshold is necessary. The adjustment unit 53 then adjusts the classification threshold determined to require adjustment. An example of a method for adjusting the classification threshold for each partitioned region is the method of repeatedly changing (updating) the classification threshold and classifying pixels, as described in the first embodiment. When using such a method, the region in which pixels are reclassified using the updated classification threshold is not the entire frame image, but the partitioned region in which the classification threshold is adjusted.

[0060] In the second embodiment, the classification of the pixels of the frame image to be processed by the classification unit 52 is determined by completing adjustment of the classification determination threshold for all partitioned regions in the frame image.

[0061] The configurations of the detection device 5 and the moving body detection system 1 including the detection device 5 in the second embodiment other than those described above are the same as those in the first embodiment.

[0062] The detection device 5 and the moving body detection system 1 including the detection device 5 in the second embodiment are configured to adjust the classification judgment threshold in consideration of the image size of the detection target moving body in the captured image, as in the first embodiment. As a result, in the second embodiment, as in the first embodiment, the detection device 5 and the moving body detection system 1 including the detection device 5 can achieve the effect of increasing the accuracy of detecting from the captured image a detection target moving body that appears in the captured image so small that it is difficult to see.

[0063] In the second embodiment, a classification threshold is set for each divided area in the captured image (frame image), and pixel classification and classification threshold are adjusted for each divided area. As a result, for example, in cases where the background differs depending on the image position within the captured image, it is expected that the accuracy of detecting a moving object to be detected from the captured image will be improved more than when a classification threshold common to all areas of the frame image is used.

[0064] Third Embodiment A third embodiment according to the present disclosure will now be described. In the description of the third embodiment, components having the same names as those in the descriptions of the first and second embodiments will be denoted by the same reference numerals, and a duplicate description of the common parts will be omitted.

[0065] In the third embodiment, in addition to the configuration of the first or second embodiment, the detection device 5 further includes a filtering unit 56 as a functional unit, as shown in FIG. 12 . The filtering unit 56 is an example of a filtering means. If the foreground region containing pixels classified as foreground pixels by the classification unit 52 is larger than a predetermined reference value, the filtering unit 56 changes the pixels contained in the foreground region larger than the reference value to background pixels. That is, as described above, the detection device 5 detects pixels classified as foreground pixels in the captured image as foreground pixels representing the target moving object. The image size of the target moving object thus detected is assumed to be very small, for example, having a few pixels in both the vertical and horizontal directions. Therefore, if the foreground region in the captured image (the region containing pixels classified as foreground pixels) is clearly larger than the image size of the target moving object, the foreground region is considered to represent something other than the target moving object. In other words, the pixels in such a foreground region are not foreground pixels representing the target moving object, and are therefore reclassified as background pixels.

[0066] Here, the reference value used by the filtering unit 56 is expressed by the number of pixels. The reference value is appropriately set by the device designer or the like, and is set to, for example, the number of pixels that represents the expected image size of the moving object to be detected with some leeway.

[0067] The configuration of the detection device 5 in the third embodiment other than the above-described configuration is the same as the configuration of the detection device 5 in the first or second embodiment.

[0068] The detection device 5 of the third embodiment has the configuration of the first or second embodiment, and can therefore achieve the same effects as the first or second embodiment. Furthermore, the detection device 5 of the third embodiment includes a filtering unit 56. By including this filtering unit 56, the detection device 5 can change pixels that are classified by the classification unit 52 as foreground pixels (i.e., pixels that are thought to reflect a moving object) and that are thought to reflect an object other than the detection target to background pixels. In other words, by changing foreground pixels that cause noise in the detection of the target moving object to background pixels, the noise in the detection of the target moving object can be suppressed, and therefore the detection device 5 can improve the detection accuracy of the target moving object even with this configuration.

[0069] In the third embodiment, an example is shown in which the filtering unit 56 changes pixels included in a foreground area larger than a reference value to background pixels. Alternatively, the filtering unit 56 may delete images that appear in a foreground area larger than a reference value from the captured image. There are multiple methods for deleting a portion of an image from a captured image, and the present invention is not limited to these, so a description thereof will be omitted.

[0070] Other Embodiments The present disclosure is not limited to the first to third embodiments and may be embodied in various ways. For example, the detection device 5 may include a preprocessing unit 57 as shown in FIG. 13 in addition to the configuration of any of the first to third embodiments. The preprocessing unit 57 performs image processing on the captured image acquired by the acquisition unit 51 before processing by the classification unit 52 or the update unit 54, making the captured image easier to process by the classification unit 52 or the update unit 54. Examples of image processing performed by the preprocessing unit 57 on the captured image include gamma correction (brightness adjustment), contrast enhancement (histogram equalization), image data normalization, fog and mist removal, and superimposing an infrared image of the detection area (monitoring area) captured by an infrared camera. One or more of these image processing methods are appropriately selected by the device designer or the like and executed by the preprocessing unit 57. The selection of the image processing method takes into account factors such as the environment of the area captured by the imaging device.

[0071] Furthermore, the detection device 5 in the first to third embodiments includes an update unit 54. Alternatively, for example, if the change in the background of the image captured by the image capture device 3 is very small and it is considered possible to prevent deterioration in the detection accuracy of the moving object to be detected without updating the background model, the update unit 54 may be omitted.

[0072] Furthermore, the first to third embodiments illustrate an example in which the mobile object detection system 1 including the detection device 5 is applied to a surveillance system. The application of the mobile object detection system 1 including the detection device 5 illustrated in the first to third embodiments is not limited to surveillance systems. For example, in a traffic control system for unmanned aircraft such as drones in the vicinity of a logistics hub, the mobile object detection system of the present disclosure may be applied to detect managed unmanned aircraft as a detection target mobile object. Furthermore, for example, when optical space communication is performed between the ground and a satellite or between satellites, the mobile object detection system of the present disclosure may be applied to identify the position of the communicating satellite. Furthermore, the detection device 5 may be incorporated into the imaging device 3.

[0073] Furthermore, the mobile object detection device may also have a configuration as shown in Fig. 14. The mobile object detection device 80 shown in Fig. 14 is, for example, a computer device, and includes, as functional units realized by executing a computer program, an acquisition unit 81, a classification unit 82, an adjustment unit 83, and an output unit 84. The acquisition unit 81, the classification unit 82, the adjustment unit 83, and the output unit 84 correspond to acquisition means, classification means, adjustment means, and output means, respectively.

[0074] The acquisition unit 81 acquires a captured image of the detection area. The classification unit 82 executes a classification process to classify each pixel constituting the acquired captured image into either a foreground pixel or a background pixel using a pixel value and a classification threshold. Foreground pixels are those that reflect a moving object. Background pixels are pixels other than foreground pixels.

[0075] If the foreground-background ratio in the captured image is higher or lower than a predetermined allowable range, the adjustment unit 83 adjusts the classification determination threshold so that the foreground-background ratio falls within the allowable range. The foreground-background ratio is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels.

[0076] The output unit 84 outputs the detection information. The detection information is information that indicates the position in the captured image of a pixel that is classified as a foreground pixel when the foreground-background ratio is within the allowable range. Note that the acquisition unit 51, classification unit 52, adjustment unit 53, and output unit 55 of the detection device 5 in the first to third embodiments described above are examples of the acquisition unit 81, classification unit 82, adjustment unit 83, and output unit 84, respectively.

[0077] The moving body detection device 80 has the above-described configuration. The moving body detection device 80 constitutes a moving body detection system 100 together with an imaging device 90 as shown in FIG.

[0078] Next, an example of the operation of the mobile object detection device 80 will be described with reference to Fig. 16. Fig. 16 is a flowchart illustrating an example of the operation of the mobile object detection device 80. Fig. 16 can also be considered to be a diagram illustrating an example of a mobile object detection method.

[0079] For example, when the acquisition unit 81 acquires a photographed image (step 201), the classification unit 82 executes a classification process (step 202). As described above, the classification process here is a process of classifying each pixel constituting the acquired photographed image into either a foreground pixel or a background pixel using a pixel value and a classification threshold value.

[0080] Thereafter, if the foreground-to-background ratio in the photographed image is higher or lower than the allowable range, the adjustment unit 83 adjusts the classification threshold so that the foreground-to-background ratio falls within the allowable range (step 203). Then, if the foreground-to-background ratio resulting from the classification of the pixels in the photographed image is within the allowable range, the output unit 84 outputs detection information indicating the positions in the photographed image of the pixels classified as foreground pixels (step 204).

[0081] The above-described moving object detection device 80 and the moving object detection system 100 including the moving object detection device 80 can detect foreground pixels depicting moving objects from a captured image through classification processing by the classification unit 82. As a result, the moving object detection device 80 and the moving object detection system 100 including the moving object detection device 80 can detect moving objects that appear so small in a captured image that they are difficult to see. Furthermore, by adjusting the classification threshold so that the foreground-to-background ratio falls within an acceptable range, the moving object detection device 80 can adjust the classification threshold so that pixels that are thought to depict moving objects that are not the detection target are less likely to be classified as foreground pixels. This allows the moving object detection device 80 to suppress foreground pixels that become noise in the detection of the target moving object, thereby improving the detection accuracy of detecting the target moving object that appears small in the captured image.

[0082] Some or all of the above-described embodiments may be described as, but are not limited to, the following supplementary notes: (Supplementary Note 1) A moving body detection device comprising: an acquisition unit that acquires a captured image of a detection area; a classification unit that performs a classification process to classify each pixel constituting the acquired captured image into either a foreground pixel that depicts a moving body or a background pixel other than a moving body using a pixel value and a threshold for classification determination; an adjustment unit that adjusts the threshold for classification determination when a foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels in the captured image, is higher or lower than a predetermined allowable range so that the foreground-background ratio falls within the allowable range; and an output unit that outputs detection information indicating the position in the captured image of the pixel classified as the foreground pixel when the foreground-background ratio is within the allowable range. (Supplementary Note 2) The moving object detection device according to Supplementary Note 1, wherein the classification unit performs classification processing for each partitioned area using a classification threshold for each predetermined partitioned area in the captured image, rather than using a classification threshold common to all areas of the captured image, and the adjustment unit adjusts the classification threshold for the partitioned area when a foreground-to-background ratio in the partitioned area is higher or lower than the allowable range. (Supplementary Note 3) The moving object detection device according to Supplementary Note 1, further comprising a filtering unit that, when a foreground area in which pixels classified as foreground pixels are concentrated is larger than a predetermined reference value, changes pixels included in the foreground area larger than the reference value to the background pixel, or deletes images of the foreground area larger than the reference value. (Supplementary Note 4) The moving object detection device according to Supplementary Note 1, wherein the detection information output by the output unit is associated with information that highlights the positions in the captured image of the pixels classified as foreground pixels. (Supplementary Note 5) The moving object detection device according to Supplementary Note 1, wherein the classification threshold used by the classification unit is a value for determining whether to classify a pixel of the captured image as the foreground pixel or the background pixel, using a degree of deviation of a pixel value from a background model representing a background in the captured image of the detection area. (Supplementary Note 6) The moving object detection device according to Supplementary Note 5, further comprising an update unit that updates the background model using the captured image acquired by the acquisition unit.(Supplementary Note 7) The moving body detection device according to Supplementary Note 1, wherein the tolerance range is set using an expected image size of a moving body to be detected that appears in a captured image. (Supplementary Note 8) A moving body detection system comprising: an imaging device that images a detection area; and the moving body detection device according to Supplementary Note 1, which uses images captured by the imaging device. (Supplementary Note 9) A moving body detection method, comprising: acquiring, by a computer, the captured image of the detection area; performing a classification process to classify each pixel constituting the acquired captured image into either a foreground pixel that shows a moving body or a background pixel other than the moving body using pixel values ​​and a threshold for classification determination; adjusting the threshold for classification determination if a foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels in the captured image, is higher or lower than a predetermined tolerance range; and outputting detection information indicating the position in the captured image of the pixel classified as foreground pixel if the foreground-background ratio is within the tolerance range. (Supplementary Note 10) A program storage medium that stores a computer program that causes a computer to execute the following processes: a process of acquiring a photographed image of a detection area; a process of executing a classification process that classifies each pixel that constitutes the acquired photographed image into either a foreground pixel that shows a moving object or other background pixel using pixel values ​​and a threshold for classification determination; a process of adjusting the threshold for classification determination so that the foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels in the photographed image, falls within a predetermined allowable range when the foreground-background ratio is within the allowable range; and a process of outputting detection information that indicates the position in the photographed image of the pixel classified as a foreground pixel when the foreground-background ratio is within the allowable range.

[0083] Note that part or all of the configurations described in Supplementary Notes 2 to 7 that are dependent on Supplementary Note 1 above may also be dependent on Supplementary Notes 8, 9, and 10 in the same dependent relationship as Supplementary Notes 2 to 7. Furthermore, not limited to Supplementary Notes 1, 8, 9, and 10, part or all of the configurations described as Supplements may be made dependent on various hardware, software, various recording means for recording software, or systems, within the scope of each of the above-mentioned embodiments.

[0084] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Furthermore, each embodiment can be combined with other embodiments as appropriate.

[0085] This application claims priority based on Japanese Patent Application No. 2024-090427, filed on June 4, 2024, the disclosure of which is incorporated herein in its entirety by reference.

[0086] 1,100 Mobile object detection system 5, 80 Mobile object detection device 51, 81 Acquisition unit 52, 82 Classification unit 53, 83 Adjustment unit 54 Update unit 55, 84 Output unit 56 Filtering unit

Claims

1. A moving object detection device comprising: an acquisition means for acquiring a photographed image of a detection area; a classification means for executing a classification process for classifying each pixel constituting the acquired photographed image into either a foreground pixel showing a moving object or a background pixel other than a moving object using a pixel value and a threshold for classification judgment; an adjustment means for adjusting the threshold for classification judgment when a foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels in the photographed image, is higher or lower than a predetermined allowable range so that the foreground-background ratio falls within the allowable range; and an output means for outputting detection information indicating the position in the photographed image of the pixel classified as a foreground pixel when the foreground-background ratio is within the allowable range.

2. The moving body detection device of claim 1, wherein the classification means performs classification processing for each predetermined divided area in the captured image using a classification threshold for each predetermined divided area in the captured image, rather than using a classification threshold common to all areas of the captured image, and the adjustment means adjusts the classification threshold for each divided area when the foreground-to-background ratio in the divided area is higher or lower than the allowable range.

3. A moving body detection device according to claim 1 or claim 2, further comprising filtering means for, when a foreground region in which pixels classified as foreground pixels are concentrated is larger than a predetermined reference value, changing pixels included in the foreground region larger than the reference value to background pixels, or deleting images of the foreground region larger than the reference value.

4. A moving body detection device as described in any one of claims 1 to 3, wherein the detection information output by the output means is associated with information that highlights the positions in the captured image of pixels classified as foreground pixels.

5. A moving body detection device as claimed in any one of claims 1 to 4, wherein the classification threshold used by the classification means is a value for determining whether a pixel in the captured image should be classified as a foreground pixel or a background pixel, using the degree of deviation of the pixel value from a background model representing the background in the captured image of the detection area.

6. The moving body detection device according to claim 5, further comprising an update means for updating a background model using the photographed image acquired by said acquisition means.

7. A moving body detection device according to any one of claims 1 to 6, wherein the allowable range is set using an estimated image size of the moving body to be detected that appears in the captured image.

8. A moving body detection system comprising: an imaging device that captures an image of a detection area; and a moving body detection device according to any one of claims 1 to 7 that uses an image captured by the imaging device.

9. A moving object detection method comprising: acquiring a photographed image of a detection area using a computer; performing a classification process to classify each pixel constituting the acquired photographed image into either a foreground pixel showing a moving object or a background pixel using pixel values ​​and a classification threshold; if a foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels in the photographed image, is higher or lower than a predetermined allowable range, adjusting the classification threshold so that the foreground-background ratio falls within the allowable range; and outputting detection information indicating the position in the photographed image of the pixel classified as a foreground pixel when the foreground-background ratio is within the allowable range.

10. The moving object detection method according to claim 9, wherein the computer performs classification processing for each predetermined divided area in the captured image using a classification threshold for each predetermined divided area in the captured image, rather than using a classification threshold common to all areas of the captured image, and adjusts the classification threshold for the divided area when the foreground-to-background ratio in the divided area is higher or lower than the allowable range.

11. The moving object detection method according to claim 9 or claim 10, wherein the computer further: when a foreground region in which the pixels classified as foreground pixels are concentrated is larger than a predetermined reference value, changes the pixels included in the foreground region larger than the reference value to background pixels, or deletes the image of the foreground region larger than the reference value.

12. A moving object detection method according to any one of claims 9 to 11, wherein the detection information output by the computer is associated with information that highlights the positions in the captured image of pixels classified as foreground pixels.

13. A moving object detection method as claimed in any one of claims 9 to 12, wherein the classification threshold used by the computer is a value for determining whether a pixel in a photographed image should be classified as a foreground pixel or a background pixel, using the degree of deviation of the pixel value from a background model representing the background in the photographed image of the detection area.

14. The moving object detection method according to claim 13, wherein the computer further updates the background model using the acquired photographed image.

15. A moving body detection method according to any one of claims 9 to 14, wherein the allowable range is set using an estimated image size of the moving body to be detected that appears in the captured image.

16. A program storage medium storing a computer program that causes a computer to execute the following processes: a process of acquiring a photographed image of a detection area; a process of performing a classification process to classify each pixel that constitutes the acquired photographed image into either a foreground pixel that shows a moving object or a background pixel that does not show a moving object, using pixel values ​​and a threshold for classification determination; a process of adjusting the threshold for classification determination so that the foreground-background ratio, which is the ratio of the number of pixels classified as foreground pixels to the number of pixels classified as background pixels in the photographed image, falls within a predetermined allowable range, if the foreground-background ratio is within the allowable range; and a process of outputting detection information that indicates the position in the photographed image of the pixel classified as a foreground pixel, if the foreground-background ratio is within the allowable range.

17. A program storage medium as claimed in claim 16, which stores a computer program that causes a computer to execute the following steps: in the classification process, rather than using a classification threshold common to all areas of the photographed image, a process that uses a classification threshold for each predetermined divided area in the photographed image and executes the process for each divided area; and in the process of adjusting the classification threshold, a process that adjusts the classification threshold corresponding to the divided area when the foreground-to-background ratio in the divided area is higher or lower than the allowable range.

18. A program storage medium according to claim 16 or claim 17, which stores a computer program that causes a computer to further execute a process of changing pixels included in a foreground area larger than the reference value to background pixels, or deleting images of the foreground area larger than the reference value, when the foreground area in which the pixels classified as foreground pixels are gathered is larger than a predetermined reference value.

19. A program storage medium according to any one of claims 16 to 18, which stores a computer program that causes a computer to execute a process of associating the detection information with information that highlights the positions in the captured image of pixels classified as foreground pixels and outputting the information.

20. A program storage medium according to any one of claims 16 to 19, wherein the threshold value for classification is a value for determining whether a pixel in a captured image is classified as a foreground pixel or a background pixel, using the degree of deviation of the pixel value from a background model representing the background in the captured image of the detection area.

21. The program storage medium according to claim 20, further storing a computer program that causes a computer to execute a process of updating a background model using the acquired captured image.

22. A program storage medium according to any one of claims 16 to 21, wherein the allowable range is set using an estimated image size of the moving object to be detected that appears in the captured image.

Images