JP2005128815A - Person detection device and person detection method - Google Patents

Abstract

A person detection apparatus capable of performing person determination in a short time and with high accuracy is provided.
In addition to a configuration for determining a person based on a feature amount, a configuration for determining a person by pattern recognition, which includes a shape feature extraction unit 101 and a pattern recognition unit 102, is provided. In addition, a tracking number setting unit 104 that sets a different number of times of tracking when the person is determined by pattern recognition is provided to shorten the person determination time. In addition, it is possible to individually set the number of person determinations that are determined to be a person based on the distance at which the moving object is detected based on pattern recognition and the case of determination based on physical feature amounts. To prevent it from being judged.
[Selection] Figure 1

Description

  The present invention relates to a person detection apparatus and a person detection method for detecting a person by processing an image taken by an imaging apparatus such as a video camera.

  Conventionally, a person detection method as described above has been proposed (see, for example, Patent Document 1). Patent Document 1 discloses a human image extraction device, which reduces the color of an input image, creates an index map corresponding to each color ID of the color map and the color map, and creates an image area based on the created index map. An area map is created by dividing and assigning an ID to each area, while creating an edge map from which edges are extracted. A person image is cut out based on the created area map and edge map.

  By the way, although the person image extraction device disclosed in Patent Document 1 can detect a person, it cannot track the person, and is not suitable for use to detect an intruder in a detection area. As a method for detecting an intruder into a detection area, the following method is considered.

  FIG. 7 is a block diagram illustrating a configuration of a person detection apparatus according to a method for detecting an intruder into a detection area. In this figure, a person detection apparatus 700 includes an input unit 702, an input image memory 703, a moving body extraction unit 704, a background image memory 705, a feature amount calculation unit 706, an object determination unit 707, and a display unit 708. It consists of and.

  The input unit 702 captures a video signal from an imaging device 701 such as a video camera, filters out high frequency components, generates A / D (Analog to Digital) conversion, and generates input image data. The input image data thus written is written into the input image memory 703. The background image memory 705 stores input image data several frames before. The moving body extraction unit 704 compares the input image data several frames before stored in the background image memory 705 with the current input image data stored in the input image memory 703 and extracts difference data. Then, the extracted difference data is compared with a predetermined threshold, and binarization processing is performed by setting a pixel exceeding the threshold to “1” and a pixel not exceeding “0” to extract a luminance change region in the input image. To do. After that, after removing noise components by dilation processing / shrinkage processing, labeling processing is performed, and if adjacent pixels are at the same level, a unique label is given to separate individual connected components to accurately move the moving object region. Extract. When the moving object region is determined, the center of gravity and area of the region are calculated.

  The moving object extraction unit 704 registers the extracted moving object region as a template by a difference method or optical flow estimation, and extracts a moving object region by searching for an area most similar to the template in the input image. Chase. The feature amount calculation unit 706 extracts the moving object region by estimating the background difference or the optical flow, and then calculates the feature amount (area ratio, aspect ratio, moving amount, moving speed, center of gravity, height, etc.) of the object region. .

The area ratio is calculated by calculating the number of vertical and horizontal pixels S ₀ of the frame circumscribing the object region extracted by the moving body extraction unit 704 and calculating the ratio with the reference pixel number S ₁ as a person. Area ratio = S ₀ / S ₁
Is calculated for each processing frame interval.

The amount of movement is defined as the amount of movement of the center of gravity from the start of tracking to the farthest position. The aspect ratio is obtained by dividing the vertical pixel number A of the frame circumscribing the object region by the horizontal pixel number B. Aspect ratio = A / B
Calculated by The aspect ratio is usually greater than “1”. Further, the moving speed is obtained by dividing the difference between the position of the center of gravity (X ₀ , Y ₀ ) at the time t and the position of the center of gravity (X ₁ , Y ₁ ) at the time t + Δt by the processing interval Δt. | (X ₁ , Y ₁ ) − (X ₀ , Y ₀ ) | / Δt
Calculated by

  The object determination unit 707 determines the feature amount (area ratio, aspect ratio, movement amount, movement speed, center of gravity, height, etc.) of the moving body region calculated by the feature amount calculation unit 706. FIG. 8 is a flowchart showing the moving object detection process. First, in the moving body extraction step in step S801, a moving object region is extracted by a difference method or optical flow estimation. Next, in the feature amount calculation step in step S802, feature amounts such as area ratio, aspect ratio, movement amount, and movement speed are calculated from the moving object region extracted in the previous step. Next, in the feature amount determination step in step S803, the feature amount calculated in the previous feature amount calculation step is compared with a predetermined specified value (a specified value possessed by the device when regarded as a person).

The area ratio must be within the specified range.Movement must be greater than the specified value.Aspect ratio must be within the specified range.Movement speed must be less than the specified value.

  If these feature values are not satisfied even with one item compared to a predetermined specified value (specified value as a person possessed by the device), or the ratio that the feature value was within the specified value as a person when tracked for the number of tracking times is If the specified value is not satisfied, the tracking is stopped. If the feature amount satisfies all the items, it is determined in the center-of-gravity position determination step in step S804 whether or not the center of gravity is within the detection area. If the center of gravity is outside the detection area, no tracking is performed. If the center of gravity is within the detection area, the search range in the current frame and the search range in the previous frame are defined as a search range that includes an area of several to several tens of pixels in the vertical and horizontal directions of the object area of the previously processed frame. Judgment is made if they overlap, and if they do not overlap, the tracking is stopped.

  On the other hand, if they overlap, tracking is established. When the tracking is established, the process proceeds to the tracking number determination step of step S805, where it is determined whether or not the number of times the tracking has been continued up to the present time is equal to or greater than the specified number. And the process returns to the moving body extraction step of step S801. When the number of tracking times exceeds the specified number, an alarm is issued and an alarm is notified. The object to be notified is an intruder into the detection area, and other objects are out of the target.

  Note that the input data processing interval is constant at intervals of several frames regardless of the distance at which the person is detected, and processing is performed at that interval. Therefore, it is necessary to set the number of tracking as large as possible in an environment where there are many moving objects other than a person so that a moving object other than a person is not detected as a person, so as to prevent misreporting caused by a moving object other than a person. . For this reason, it takes a time corresponding to the number of tracking times × processing time (interval) + hold time until the detected moving object is determined as a person and issued.

JP 2002-245441 A

  However, in the conventional person detection device capable of detecting intruders, the physical feature amounts such as the area ratio / height and the moving speed of the moving object are calculated, and those values are less than the prescribed values of the device, In addition, it is determined that the person is a person when the period of time specified in the number of tracking is continued, but there is a problem that a moving object other than a person is often determined as a person, and it takes time to determine a person. .

  The present invention has been made in view of this point, and an object of the present invention is to provide a person detection apparatus and a person detection method that can perform person determination in a short time and with high accuracy.

  In order to solve the above-described problems, a human detection device according to the present invention includes a moving object detection unit that detects a moving object that is a report target by performing image processing on an image captured by an imaging device, and the moving object detection unit. The tracking means for tracking the moving object by repeatedly capturing the output of the information, the person determining means for determining whether the moving object is a person from the result of tracking the detected moving object multiple times, and the person determining means A tracking number setting unit that sets a different number of times of tracking according to a difference in a method for determining the moving object as a person.

  According to this configuration, since the number of times of tracking is reduced by using pattern recognition, which is a technique capable of performing person determination with high accuracy, it is possible to shorten the time required to determine a person.

  Further, the person detection device of the present invention is the person detection device of the invention according to claim 1, wherein the person determination means performs person determination by pattern recognition or person determination by a physical feature amount, and the tracking number setting means Has a configuration in which the number of times of tracking is set to a smaller value when the person determination unit determines that the person is based on pattern recognition than when the person determination unit determines that the person is based on the physical feature amount.

  According to this configuration, by using pattern recognition that can perform person determination with high accuracy, the number of times of tracking can be reduced compared to the case of determining a person by physical feature amount, and thus the time required to determine a person. Can be shortened.

  The human detection device according to the present invention is the human detection device according to the first or second aspect, wherein the hold time for checking the notification for a predetermined time or more immediately after the tracking of the moving object is performed for the tracking number of times. And hold time setting means for changing the length of the hold time according to the difference in the method for the person determination means to determine the moving object as a person.

  According to this configuration, after the tracking is performed for the number of times of tracking, the notifications are forgotten only for the hold time, and only when the person other than the person is not determined within the hold time, the notification is issued after the hold time. Therefore, even when the number of tracking times is satisfied, it is possible to prevent a moving object other than a person from being erroneously determined as a person.

  The human detection device of the present invention is the human detection device according to any one of claims 1 to 3, further comprising a distance measuring unit that measures a distance from the imaging device to the detected moving object, The person determination unit performs person determination of the moving object mainly by pattern recognition when the position of the detected moving object is close to the imaging apparatus, and when the position of the detected moving object is far from the imaging apparatus, the detected movement Performing person determination by either physical feature amount or pattern recognition of the object, and further setting the number of person determinations for determining the moving object as a person according to the distance from the imaging device to the detected moving object, The tracking number setting unit is configured to set the tracking number according to a distance from the imaging device to the detected moving object.

  According to this configuration, the method for determining the person is changed according to the difference in the distance from the imaging device to the moving object, and the number of tracking times and the number of person determinations are set. It is possible to make a determination, and it is possible to reduce misinformation caused by determining a moving object other than a person as a person.

  According to another aspect of the present invention, there is provided a person detection device according to any one of the first to fourth aspects, wherein the person detection device generates a video image of moving an object in the shape of a person by image processing. And a person image synthesizing means for synthesizing the person image generated by the person image generating means on the image photographed by the imaging device.

  According to this configuration, a video in which an object in the shape of a person moves by image processing is created, and the video is synthesized with the video from the imaging device, and the video in which a person actually enters the screen The person detection function of the device can be confirmed by creating and recognizing the device.

  The human detection method of the present invention includes a step of performing image processing on a captured image to detect a moving object that is a report target, and whether or not the moving object is a person based on a result of tracking the moving object multiple times. A person detection method comprising a step of determining and a step of issuing a notification when it is determined that the person is a person, wherein in the person determination for the detected moving object, determination by pattern recognition and physical characteristics While the number of tracking is set to a different value depending on the amount of determination, the number of person determination can be set as a reference for determining a person based on the number of times determined by the pattern recognition. And

  According to this method, the number of times of tracking is reduced by using pattern recognition, which is a technique that can perform person determination with high accuracy, and the number of times of tracking can be reduced by setting the number of person determinations to be used as a reference for determining a person. Thus, the person can be determined, and the time required to determine the person can be shortened.

  The human detection method of the present invention includes a step of performing image processing on a captured image to detect a moving object that is a report target, and whether or not the moving object is a person based on a result of tracking the moving object multiple times. A person detection method comprising a step of determining and a step of issuing a notification when it is determined that the person is a person, wherein the predetermined hold is performed immediately after the tracking of the moving object by the number of tracking times. Even when the moving object is determined as a person even after a lapse of time, and the length of the hold time is determined using person recognition by pattern recognition and using a physical feature amount In the case of pattern recognition, the value is set shorter than in the case of a physical feature amount.

  According to this method, after the tracking is performed for the number of tracking times, the notifications are forgotten for the hold time, and the notification is issued after the hold time only when the person other than the person is not determined within the hold time. Therefore, even when the number of tracking times is satisfied, it is possible to prevent a moving object other than a person from being erroneously determined as a person.

  In addition, the person detection method of the present invention includes a step of performing image processing on an image captured by an imaging device to detect a moving object that is a report target, and a step of performing notification when the moving object is detected. When the position of the detected moving object is within a range set in advance from the imaging device, person determination is performed on the moving object by pattern recognition, and the position of the detected moving object is determined. If the range exceeds the range from the imaging device, the person is determined for the moving object by pattern recognition or physical feature quantity, and the tracking count and the person determination count are further determined according to the distance from the imaging device to the detected moving object. Set.

  According to this method, the person determination method is changed according to the difference in the distance from the imaging device to the moving object, and the tracking number and the person determination number are set. It is possible to make a determination, and it is possible to reduce misinformation caused by determining a moving object other than a person as a person.

  In addition, the person detection method of the present invention includes a step of performing image processing on an image captured by an imaging device to detect a moving object that is a report target, and a step of performing notification when the moving object is detected. In the person detection method, a person image is created by image processing, the created person image is synthesized on the image photographed by the imaging device, and further moved within the screen.

  According to this method, an image in which an object in the shape of a person moves is created by image processing, the image is synthesized with the image from the imaging device, and an image in which a person actually enters the screen The person detection function of the device can be confirmed by creating and recognizing the device.

  According to the present invention, it is possible to reduce erroneous detection of a moving object other than a person as a person even in an environment where a moving object other than a person is frequently detected, and to increase the time required to determine the moving object as a person. In addition to shortening, it is possible to perform highly accurate person detection. In addition, since the person image generated by the image processing is synthesized and moved in the captured image, the person intrusion test to the area where the normal person should not be present is actually put in the detection area. It becomes possible to do without.

  Hereinafter, the best mode for carrying out the present invention will be described in detail with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a configuration of a person detection apparatus according to Embodiment 1 of the present invention. In addition, in this figure, the same code | symbol is attached | subjected to what has the function which is common in each part in FIG. 7 mentioned above. 1, a human detection apparatus 100 according to the present embodiment includes an input unit 702, an input image memory 703, a moving body extraction unit 704, a background image memory 705, a feature amount calculation unit 706, and an object determination unit. 103, a shape feature extraction unit 101, a pattern recognition unit 102, a tracking number setting unit 104, and a display unit 105.

  The input unit 702 captures a video signal from an imaging device 701 such as a video camera, filters the captured video signal to remove a high frequency component, and then performs A / D conversion to generate input image data. Then, the generated input image data is stored in the input image memory 703. The moving body extraction unit 704 obtains difference data by comparing the image several frames before stored in the background image memory 705 with the current input image stored in the input image memory 703, and calculates the difference data as a predetermined value. Compared with the threshold value, a pixel that exceeds the threshold value is set to “1”, and a pixel that does not exceed the threshold value is set to “0”, thereby performing binarization processing and extracting a luminance change region in the input image. The extracted luminance change region is subjected to expansion / contraction processing to remove noise components, and then subjected to labeling processing. If adjacent pixels are at the same level, each connected component is given a unique label. To accurately extract the object region. When the object region is determined, the center of gravity and area of the region are calculated. That is, the moving body extraction unit 704 obtains a difference between a background image in which no moving object exists and the current image in the input image memory 703, extracts a changed area (that is, a moving body area), and also extracts an input image. And an input image several frames before, an optical flow is obtained, and a moving object region is extracted by applying a labeling process to a block in which the flow is generated.

  The feature amount calculation unit 706 calculates the physical feature amount of the moving object region extracted by the moving body extraction unit 704, that is, the area ratio, movement amount, movement speed, aspect ratio, center of gravity, height, etc. for the person. . The shape feature extraction unit 101 extracts edge information by performing filter processing such as a Gabor filter on the input image stored in the input image memory 703. The pattern recognition unit 102 recognizes a person from the edge information from the shape feature extraction unit 101. That is, the pattern recognizing unit 102 recognizes the recognition model and SVM (Support Vector Machine) obtained by learning in advance based on the physical feature amount of the moving body region calculated by the filtering process in the shape feature extracting unit 101. A person is recognized using etc. The object determination unit 103 compares the feature amount as a person calculated by the feature amount calculation unit 706 with a value that the apparatus has as a specified value for the person and is obtained from the recognition result of the pattern recognition unit 102. The personness threshold is compared with a prescribed value of personiness possessed by the apparatus to determine whether the moving object is a person.

  The tracking number setting unit 104 sets different tracking numbers depending on whether the object determination unit 103 determines the moving object as a person, that is, whether the moving object is determined as a person based on the pattern recognition result or the feature amount calculation result. That is, the number of times of tracking is changed between the pattern recognition result and the feature amount calculation result. In particular, since the pattern recognition result can be more accurate than the feature amount calculation result for the person itself, the number of tracking is reduced when the pattern recognition result is selected.

  Further, the tracking number setting unit 104 calculates the distance of the moving object from the imaging device 701, and sets the number of tracking times and the person determination number serving as a threshold for determining a person by pattern recognition according to the distance. When the apparatus detects a moving object and starts tracking, if the moving object is determined as a person continuously for the number of times that satisfies the tracking number set by the tracking number setting unit 104, the object determining unit 103 detects the moving object. It judges that it is a person and issues an alarm and notifies an alarm. The display unit 105 displays the detection area and displays a frame in the area determined to be a person. Further, a menu for the user to make various settings is displayed.

  Here, the processing of the tracking number setting unit 104 will be described in detail. Based on the height, area ratio, aspect ratio, movement amount, movement speed, etc. of the moving object calculated by the feature amount calculation unit 706, the specified value as a person possessed by the apparatus is satisfied, and the center of gravity of the detected moving object Is detected within the detection area set by the apparatus. The pattern recognizing unit 102 calculates how much the shape feature of the detected moving object looks like a person as compared to the shape feature of the device, and determines that the device is a person if the shape feature is equal to or greater than a specified value of the device. Based on these results, the number of tracking was set based on whether the moving object was determined to be a person based on the feature value determination result or the pattern recognition result, and the number of times that the moving object was determined to be a person by pattern recognition during tracking was calculated. If the value is equal to or greater than the number of person determinations that the apparatus has as a threshold for determining a person, it is determined as a person.

  Furthermore, a prescribed value for determining a person by pattern recognition (number of times of person determination indicating how many times a person is determined during tracking determines that the moving object is a person) calculates the distance between the moving object and the imaging device 701. By doing this, if the moving object is near the imaging device 701 (for example, within 200 m), the specified value is set to be large, and if the moving object is far from the imaging device 701 (for example, 200 m or more), the specified value is set to be small. Configure. The distance to the moving object can be calculated by instructing the distance from the outside at intervals of several tens of meters on the screen projected by the imaging device 701 in advance.

  Next, the operation of the person detection apparatus according to the present embodiment will be described with reference to the flowchart shown in FIG. In FIG. 2, first, in a moving body extraction step in step S <b> 201, a difference between a background image in which no moving object exists and the current image is obtained, and a changed area, that is, a moving body area is extracted. Also, an optical flow is obtained using the input image and a memory image several frames before, and a moving body region is extracted by performing a labeling process on the block in which the flow is generated. Next, in a feature amount calculation step 202 in step S202, physical feature amounts such as an area ratio, an aspect ratio, a movement amount, and a movement speed are calculated from the moving body region extracted in the previous moving body extraction step.

  Next, in the feature amount determination step in step S204, the feature amount calculated in the previous feature amount calculation step is compared with a specified value possessed by a device for regarding a person. In this case, if all the items of the feature amount are within the specified value, or if the ratio of the feature amount within the specified value as a person when tracking for the number of times of tracking is a specific value or more, it is a person Judge. Next, in the gravity center position determination step in step S206, it is determined whether or not the detected gravity center of the moving object is within the detection area. If the center of gravity of the moving object is within the detection area, the search range of the previous frame is set as a search range that includes several to several tens of pixels in the vertical and horizontal directions of the moving object area detected in the previous frame. It is determined whether or not the search ranges in the current frame overlap. If they overlap, it is considered that tracking has been established, and the process proceeds to the tracking number determination step in step S208.

  On the other hand, in the pattern recognition process in step S203, shape features are recognized from the moving object extracted in the moving body extraction process in step S201. Next, in the person determination step of step S205, the shape feature of the moving object extracted in the previous pattern recognition step is compared with the shape feature as a person possessed by the apparatus to calculate how much the person looks like, If the threshold of the person-likeness possessed by the device is exceeded, it is determined that the device is a person. Otherwise, it is determined that the device is not a person.

  Next, in the person determination factor determination step in step S207, it is determined whether the factor that determined that the moving object is a person is the result of calculation from the feature amount or the degree of personiness by pattern recognition. To do. Then, according to the determination, the tracking number n is set in the tracking number setting step in step S209, the distance to the moving object is calculated, and the threshold value (person) is determined by setting the number of tracking according to the distance and pattern recognition. Determine the number of judgments). Based on the result of the tracking number setting step in step S209, a predetermined value of the tracking number that the apparatus has is determined in the tracking number determination step in step S208. If the number of times of tracking is equal to or greater than a specified value, it is determined that the person is a person and is issued. If the tracking number is equal to or less than the specified value, “1” is added to the current tracking number (step S210), and the process returns to the moving body extraction step in step S201.

  As described above, according to the person detection device according to the present embodiment, the prescribed number of tracking times is determined based on the factor that determines the moving object as a person, and the tracking number and pattern recognition are further performed according to the distance at which the moving object is detected. The threshold value (number of person determinations) for determining a person can be set by the above, and if the person is determined by high-precision pattern recognition, if the number of person determinations is satisfied before the number of tracking times is reached, a notification is issued at that time Therefore, even in an environment where a moving object other than a person is frequently detected, it is possible to reduce false detection of detecting a moving object other than a person as a person, and to reduce the time until a moving object is determined as a person. It can be shortened further. That is, a person can be detected in a short time and with high accuracy.

(Embodiment 2)
FIG. 3 is a block diagram showing the configuration of the person detection apparatus according to Embodiment 2 of the present invention. In this figure, the same reference numerals are given to those having the same functions as those in FIG. 1 described above. The person detection apparatus according to the present embodiment is different from the person detection apparatus according to the first embodiment in that a hold time setting unit 301 is provided.

  Shape features from the edge data of the input image extracted by the shape feature extraction unit 101 to the person's head (circular or elliptical) and the shoulder that follows it, head width and wider shoulder width (number of pixels), etc. The pattern recognition unit 102 compares the shape feature with the shape feature as a person possessed by the device, calculates how much the person seems to be, compares the result with a specified value of the degree of personiness of the device, and the device If it is equal to or greater than the specified value, the person is determined. Also, the feature quantity such as the height / area ratio / aspect ratio / movement amount of the moving object calculated by the feature quantity calculation unit 706 is compared with the specified value as the person of the device, and all values are within the specified value. Is determined to be a person. The set values of the tracking number setting unit 104 and the hold time setting unit 301 are determined according to which of the processes of determining the moving object detected by the apparatus as a person is determined as a person.

  Details of the hold time setting unit 301 will be described here. In a human detection device that detects only a person from a video signal in an environment where there are many moving objects other than a normal person or objects similar to a person, the same moving object is continuously applied for the number of tracking times that the device has. Is detected as a person, the moving object is determined as a person. Therefore, when it is desired to quickly detect a moving object as a person, it is necessary to set the number of tracking times to be small. However, in this case, there is a high possibility that a moving object other than a person is erroneously determined as a person. Therefore, even when the hold time setting unit 301 is provided and the tracking number of the device is satisfied, the alarm and the alarm notification are further checked for the hold time, and if the person other than the person is not determined within the hold time, that is, the hold Only when it is determined that the person is within the time, the alarm is issued after the hold time.

  FIG. 4 is a flowchart showing the detection process of the person detection apparatus according to the present embodiment. Steps S201 to S210 are the same as those in the first embodiment described with reference to FIG. First, in the person determination factor determination step in step S207, the factor that the apparatus has determined the moving object as a person depends on the processing result of the feature amount calculation step in step S202, the feature amount determination step in step S204, or the pattern recognition in step S203. It is determined whether the process is based on the process result of the person determination process of step S205, and the setting value in the hold time determination process of step S402 is determined in the hold time setting process of step S401 based on the result.

  Normally, when a person is determined based on the results of the pattern recognition process in step S203 and the person determination process in step S205, the person is determined by the feature amount calculation process in step S202 and the feature amount determination process in step S204. Set the hold time short compared to the case. In addition, when the distance from the imaging device 701 of the detected moving object is calculated and the distance to the moving object is short, the hold time is set short, and when the distance to the moving object is long, the hold time is set long. It is also possible to do.

  As described above, according to the person detection device according to the present embodiment, after detecting a moving object and satisfying the prescribed value of the number of times of tracking, based on the factor that determined the moving object as a person and the distance at which the moving object was detected. In an environment where a moving object other than a person is frequently detected by configuring the apparatus to perform tracking for a long time set by the hold time setting unit 301 and determine whether or not the moving object is a person. It becomes possible to detect a person with high accuracy.

(Embodiment 3)
FIG. 5 is a block diagram showing a configuration of a person detection apparatus according to Embodiment 3 of the present invention. In this figure, the same reference numerals are given to those having the same functions as those in FIG. 1 described above. In FIG. 5, the human detection apparatus according to the present embodiment includes an intruder data creation unit 501 and an intruder data composition unit 502. The intruder data creation unit 501 generates an intruder image by image processing, The person data composition unit 502 synthesizes the intruder image generated by the intruder data creation unit 501 on the background image of the image photographed by the photographing device 701 and moves it within the screen to simulate the intrusion of a person. .

  The operations of the intruder data creation unit 501 and the intruder data composition unit 502 will be described in detail. Conventionally, it is dangerous to confirm that the device is functioning reliably when monitoring a person who is not located near 20m from a distance of 20m or an area where a normal person should not be detected. It was necessary to confirm the detection operation of the device by entering a person in the monitoring area regardless of the area or location. Therefore, an intruder data creation unit 501 and an intruder data composition unit 502 are provided in the previous stage of the input image memory 703 of the conventional human detection device. The intruder data creation unit 501 creates a video in which a human-shaped object moves by image processing using computer graphics, and the intruder data synthesis unit 502 combines the video with the video from the imaging device 701. The apparatus is configured so that the person detection function of the apparatus can be confirmed by creating an image in which a person actually enters the screen and causing the apparatus to recognize it. Even if it is set so that a person created by computer graphics is automatically moved arbitrarily, the movement can be selected by a menu on the display unit 105.

  FIG. 6 is a block diagram showing the configuration of the intruder data composition unit 502. The scan converter unit 601 converts the R, G, and B signals from the intruder data creation unit 501 into a composite signal, and the mixing unit 602 synthesizes it with the video signal from the imaging device 701. The synchronization signal generator 603 generates a clock (CLK), a horizontal synchronization signal (HSYNC), a vertical synchronization signal (VSYNC), and a field signal (FIELD) synchronized with the video signal from the imaging device 701, and is the same for all blocks. Supply synchronization signal. A switch unit 604 is a video switching switch for switching between a video obtained by synthesizing human data and a video from the imaging device 701. When the detection operation is confirmed, the switch unit 604 switches to the output of the mixing unit 602.

  As described above, according to the human power detection apparatus according to the present embodiment, when a person intrudes into an area where a normal person should not be present, and when it is detected and notified, the operation of the apparatus is confirmed. It is possible to confirm the operation of the apparatus without actually allowing a person to enter the detection area.

  The present invention can reduce erroneous detection of a moving object other than a person as a person even in an environment where a moving object other than a person is frequently detected, and further shortens the time until a moving object is determined to be a person. In addition, there is an effect that it is possible to detect a person with high accuracy, and the present invention can be applied to a purpose of detecting and notifying an entry into a specific place such as a place where entry is prohibited.

1 is a block diagram showing a configuration of a person detection device according to Embodiment 1 of the present invention. The flowchart for demonstrating operation | movement of the person detection apparatus which concerns on Embodiment 1 of this invention. The block diagram which shows the structure of the person detection apparatus which concerns on Embodiment 2 of this invention. The flowchart for demonstrating operation | movement of the person detection apparatus which concerns on Embodiment 2 of this invention. The block diagram which shows the structure of the person detection apparatus which concerns on Embodiment 3 of this invention. The block diagram which shows the structure of the intruder data synthetic | combination part of the person detection apparatus which concerns on Embodiment 3 of this invention. The block diagram which shows the structure of the conventional person detection apparatus Flowchart for explaining the operation of a conventional person detection device

Explanation of symbols

100, 300, 500 Human detection device 101 Shape feature extraction unit 102 Pattern recognition unit 103 Object determination unit 104 Tracking count setting unit 105 Display unit 301 Hold time setting unit 501 Intruder data creation unit 502 Intruder data synthesis unit 601 Scan converter unit 602 Mixing unit 603 Synchronization signal generating unit 604 Switch unit 701 Imaging device 702 Input unit 703 Input image memory 704 Moving object extraction unit 705 Background image memory 706 Feature quantity calculation unit

Claims (9)

A moving object detection means for detecting a moving object that is a subject of notification by performing image processing on an image captured by an imaging device;
Tracking means for tracking the moving object by repeatedly capturing the output of the moving object detection means;
Person determination means for determining whether or not the moving object is a person from a result of tracking the detected moving object multiple times;
Tracking number setting means for setting a different number of tracking times according to a difference in a method for the person determination means to determine the moving object as a person;
A human detection device comprising:   The person determination unit performs person determination based on pattern recognition or person determination based on a physical feature amount. When the person determination unit determines that the person is determined by pattern recognition, the physical feature amount is determined. The person detection device according to claim 1, wherein the tracking number is set to a value smaller than that in the case where the person is determined to be a person.   A hold time for setting off the notification for a predetermined time or more immediately after the tracking of the moving object for the number of tracking times is set, and the person determination means determines the length of the hold time as the person. The person detection device according to claim 1, further comprising a hold time setting unit that changes according to a difference in a technique for the purpose.   A distance measuring unit that measures a distance from the imaging device to the detected moving object, and the person determination unit mainly performs pattern recognition when the position of the detected moving object is close to the imaging device. A person is determined as a moving object, and if the person is far from the imaging device, the person is determined based on either the physical feature amount or pattern recognition of the detected moving object, and the detected moving object is detected from the imaging device. The number of person determinations for determining the moving object as a person is set according to the distance to the tracking object, and the tracking number setting unit sets the number of tracking according to the distance from the imaging device to the detected moving object. The person detection device according to claim 3.   Human image generation means for generating an image in which an object in the shape of a person moves by image processing, and human image composition for combining the human image generated by the human image generation means on an image taken by the imaging device The person detection apparatus according to claim 1, further comprising: means. A process of detecting a moving object to be reported by performing image processing on the captured image, a step of determining whether or not the moving object is a person from a result of tracking the moving object a plurality of times, and a determination of being a person A person detection method comprising a step of issuing a notification when
In the person determination for the detected moving object, the tracking number is set to a different value between the case of performing the determination by pattern recognition and the case of the determination by the physical feature amount, while the result of tracking the set number of tracking times, A person detection method capable of setting the number of person determinations serving as a reference for determining a person based on the number of times determined to be a person by pattern recognition. A process of detecting a moving object to be reported by performing image processing on the captured image, a step of determining whether or not the moving object is a person from a result of tracking the moving object a plurality of times, and a determination of being a person A person detection method comprising a step of issuing a notification when
Not only immediately after tracking the moving object for the number of tracking times, but only when the moving object is determined as a person even after a predetermined hold time has passed, and the length of the hold time Is set to a different value in the case of performing person determination by pattern recognition and in the case of using physical feature quantity, and in the case of pattern recognition, the person detection method is set shorter than in the case of physical feature quantity. . A person detection method comprising: a step of detecting a moving object as a notification target by performing image processing on an image captured by an imaging device; and a step of performing notification when the moving object is detected,
When the position of the detected moving object is within a range set in advance from the imaging device, person determination is performed on the moving object by pattern recognition, and the position of the detected moving object exceeds the range from the imaging device. In this case, a person detection method for performing person determination on the moving object based on pattern recognition or a physical feature amount, and further setting a tracking number and a person determination number according to a distance from the imaging device to the detected moving object. In a person detection method comprising: a step of detecting a moving object that is a subject of notification by performing image processing on an image captured by an imaging device; and a step of notifying when the moving object is detected,
A person detection method in which a person image is created by image processing, the created person image is synthesized on an image photographed by the imaging device, and further moved in a screen.

Images