WO2015136828A1 - 人物検出装置および人物検出方法 - Google Patents
人物検出装置および人物検出方法 Download PDFInfo
- Publication number
- WO2015136828A1 WO2015136828A1 PCT/JP2015/000356 JP2015000356W WO2015136828A1 WO 2015136828 A1 WO2015136828 A1 WO 2015136828A1 JP 2015000356 W JP2015000356 W JP 2015000356W WO 2015136828 A1 WO2015136828 A1 WO 2015136828A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- moving object
- image
- person
- still
- area
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B15/00—Systems controlled by a computer
- G05B15/02—Systems controlled by a computer electric
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
- G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G08—SIGNALLING
- G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
- G08B13/00—Burglar, theft or intruder alarms
- G08B13/18—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength
- G08B13/189—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems
- G08B13/194—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems
- G08B13/196—Actuation by interference with heat, light, or radiation of shorter wavelength; Actuation by intruding sources of heat, light, or radiation of shorter wavelength using passive radiation detection systems using image scanning and comparing systems using television cameras
- G08B13/19602—Image analysis to detect motion of the intruder, e.g. by frame subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
Definitions
- the present invention relates to an apparatus for detecting a person using an image.
- Typical methods for detecting specific objects include the AdaBoost algorithm based on Haar-Like features and HOG (Histograms of Oriented Gradients) features, which are mainly used to detect faces and people.
- AdaBoost AdaBoost algorithm based on Haar-Like features and HOG (Histograms of Oriented Gradients) features, which are mainly used to detect faces and people.
- HOG Histograms of Oriented Gradients
- the background subtraction method of (2) if an image in which no person is shown is used as a background image, the position of the person can be accurately extracted by taking the difference from the image in which the person is shown.
- this method is known to be vulnerable to changes in ambient illuminance.
- a method of appropriately updating the background image according to the external environment is also considered, but it is optimal in an environment where people frequently enter and exit such as offices and factories, or in an environment where lighting is turned on or off. It is difficult to select a background image.
- the inter-frame difference method (3) since the difference between the previous image and the current image is taken in time series, it is possible to detect a moving object and is relatively resistant to changes in ambient illuminance. . However, since a non-moving object cannot be detected, a stationary person cannot be detected. As a conventional method for detecting a stationary object by the inter-frame difference method, there is a tracking method for tracking a moving object in an image and recognizing that the object is stationary at a position where the movement stops.
- the human detection device includes an image acquisition unit that acquires a captured image, a position of the first moving object region extracted from the image acquired by the image acquisition unit, and an image of the first moving object region. Based on the distance from the position of the second moving body region extracted from the previous image, a stationary region determination unit that determines that the moving body of the second moving body region is in a stationary state; A stationary person determination unit that determines the presence / absence of a person using a change amount of the feature amount.
- the present invention determines a still region based on the distance between the position of the first moving object region extracted from the image and the position of the second moving object region extracted from an image before the image of the first moving object region. Since the presence / absence of a person is determined using the amount of change in the feature amount of the image in the still region, the device is controlled with the person information including the still person without collecting and registering a huge amount of image feature amount in advance. can do.
- FIG. 13 is a diagram showing a feature change amount by which the still person determination unit 10 determines a still person in the embodiment of the present invention.
- FIG. 1 is a configuration diagram of a person detection apparatus 1 according to an embodiment of the present invention.
- a person detection device 1 acquires image data from a camera 2 that captures a specific area, detects the number and positions of people in the specific area, and outputs it as person information.
- the camera 2 is described as being installed on the ceiling, but the camera may be installed anywhere.
- the person detection device 1 and the camera 2 may be configured separately or integrated.
- the camera 2 is installed on the ceiling, the person detection device 1 is installed on a floor or a control room of a building, and the camera 2 and the person detection device 1 are connected by a LAN (Local Area Network) cable or a coaxial cable.
- the person detection device 1 may acquire image data from a plurality of cameras 2 and perform person detection.
- the person detection apparatus 1 includes an image acquisition unit 3, a moving object extraction unit 4, a moving object tracking unit 5, a moving object history storage unit 6, a still region image storage unit 7, a still region image storage unit 8, a feature change amount extraction unit 9, a still person It comprises a determination unit 10 and a person information output unit 11.
- the moving object extraction unit 4 includes a moving object region extraction unit 12, and the moving object tracking unit 5 includes a stationary region determination unit 13.
- the human detection device 1 is equipped with a calculation device such as a CPU (Central Processing Unit), and each processing unit (3 to 13) is started as a program on the calculation device. Further, the person detection device 1 is equipped with a recording device such as a ROM ((Read Only Memory), a RAM (Random Access Memory), a flash memory, or a hard disk, and each storage unit (6 and 8) stores these recording devices. May be used.
- a calculation device such as a CPU (Central Processing Unit)
- each processing unit (3 to 13) is started as a program on the calculation device.
- a recording device such as a ROM (Read Only Memory), a RAM (Random Access Memory), a flash memory, or a hard disk, and each storage unit (6 and 8) stores these recording devices. May be used.
- the image acquisition unit 3 acquires image data captured by the camera 2 from the camera 2. Then, the latest one-frame image is extracted from the acquired image data. Then, the latest extracted one-frame image data is stored in a primary storage device (not shown).
- the moving object region extraction unit 13 of the moving object extraction unit 4 receives the latest image data input to the primary storage device and the previous frame stored in the primary storage device. The difference from the image data is calculated. Then, image blocks whose differences are equal to or greater than the set threshold are grouped and extracted as moving object regions. Then, the extracted moving object data (for example, the position of the moving object region and the coordinates of the region) is output to the moving object tracking unit 5.
- the moving object tracking unit 5 refers to the history of moving object data recorded in the moving object history recording unit 6, and all the previous frames are recorded.
- the position of the moving object area is acquired as the previous position.
- the stationary region determination unit 13 of the moving body tracking unit 5 is based on the distance between the positions of all the moving body regions one frame before and the position of the moving body region input from the moving body region extraction unit 12 of the moving body extraction unit 4.
- the moving body area is determined to be a stationary area. Then, the position of the moving object region (for example, the center position of the moving object region), the coordinates of the moving object region, the state of the moving object region (moving state or stationary state), and the moving object ID (indentification) that identifies the moving object. Is output to the moving body history storage unit 6 and stored as the latest moving body data.
- moving body data of the moving body region is input from the moving body tracking unit 5 and stored.
- the moving body history information of the moving body data stored in the moving body history storage unit 6 is referred to by the moving body tracking unit 5, the still region image storage unit 7, the feature change amount 8, and the person information output unit 11. If the stationary person determination unit 10 determines that there is no person in the stationary region, the data of the stationary region in the moving object history storage unit 6 is deleted.
- the still area image storage unit 7 acquires information on all the still areas stored in the moving object history storage unit 6. Then, the latest image data stored in the primary storage device is acquired, a still area is cut out from the acquired image data, and is output to and stored in the still area image storage unit 8.
- the image data of the still area of the latest image data is input from the still area image storage unit 7 and stored. Then, the information on the still area stored in the still area image storage unit 8 is referred to from the feature change amount extraction unit 9. If the still person determination unit 10 determines that there is no person in the still area, the data of the still area in the still area image storage unit 8 is deleted.
- the amount of change (feature change amount) with the feature amount of the image of the still area one frame before stored in (2) is extracted. Then, the feature change amount of the extracted image is output to the still person determination unit 10.
- the still person determination unit 10 stops when the feature change amount of the input image is within the set range after the set time. It is determined that there are people in the area. If it is outside the set range, it is determined that there is no person in the still area. Then, still person data indicating that there is a person who is stationary with respect to the determined still area is output to and stored in the moving body history storage unit 6. In addition, the still area determined to have no person is deleted from the moving body history storage unit 6.
- the still person determination unit 10 may include the processing of the still area image storage unit 7 or the feature change amount extraction unit 9.
- the person information output unit 11 is a communication unit that outputs information including the number and position detected as a person to an external device.
- the person information output unit 11 refers to the data of the moving object history recording unit 6 at predetermined time intervals or according to an instruction from an external device, for example, and determines the number and positions of persons from the moving object area and the still area detected as a person.
- the included person information is calculated, and the person information is output to the external device.
- the external device is a device or system that controls energy or raises an alarm according to the number and position of people, such as a lighting control device, an air conditioning control device, and an on-site alarm device. The above is the description of each configuration of the person detection device 1.
- 2 to 6 are flowcharts showing processing of the person detecting apparatus 1 in the embodiment of the present invention.
- processing of the image acquisition unit 3 and the moving object extraction unit 4 will be described with reference to FIG.
- the image acquisition unit 3 acquires the image data captured by the camera 2 from the camera 2, and acquires the latest one frame image from the acquired image data (S1).
- the acquired image data is an analog signal, it is converted from a digital signal into a one-frame digital image. If the acquired image data is compression-encoded image data such as Motion JPEG (Joint Photographic Experts Group) or MPEG (Moving Picture Experts Group), it is combined and converted into a digital image of one frame.
- the digital image may be a color image in RGB format or YUV format, or may be a luminance image (grayscale image).
- FIG. 7 is an image diagram illustrating the processing of the moving object extraction unit 4 in the embodiment of the present invention.
- the image 701 and the image 702 in FIG. 7 are images stored in the primary storage device, and the image 701 is an image one frame before the latest image 702 acquired by the image acquisition unit 3.
- An image 701 shows an object 703 and an image 702 shows an object 704 being captured.
- the moving body extraction unit 4 performs a frame interval between the latest image 702 and the previous frame 701 stored in the primary storage device.
- the difference is calculated (S3).
- an image block 705 having a difference between frames equal to or greater than a preset threshold is extracted as a motion block 706 (S4).
- the image block 705 is an area where the image is divided equally, and the number of divisions is a number determined in advance according to the image size.
- the threshold value may be a predetermined value determined in advance or may be changed dynamically.
- the motion blocks 706 are grouped by labeling the extracted motion blocks 706 (S5).
- a rectangle including the grouped motion blocks is defined as a moving object region 707.
- the position of the moving object region 707 for example, the center position or the center of gravity position of the moving object region is calculated (S6), and the information on the calculated position and area coordinates is output to the moving object tracking unit 5 as moving object data of the moving object region 707 and stored. To do.
- the moving object tracking unit 5 tracks the latest moving object region input from the moving object extracting unit 4 using the moving object data of the past moving object region stored in the moving object history storage unit 6.
- the moving object tracking unit 5 has a threshold M as the shortest distance Dmin between the position of the moving object area one frame before and the position of the latest moving object area, and ⁇ 1 as the identification number Min of the past moving object area as the shortest distance Dmin.
- the identification number i of the latest moving body region input from the extraction unit 4 is set to 1 and the identification number j of the past moving body region stored in the moving body history storage unit 6 is set to 1 and initialized (S7).
- the moving object tracking unit 5 acquires the history data of the moving object data from the moving object history storage unit 6 and performs the following processing.
- the position PB (j) of the moving body region B (j) one frame before acquired from the moving body history storage unit 6 is acquired as the previous position (S8).
- the distance D (j) between the position PA and PB (j) of the moving object in the moving object region A, which is the latest moving object data input from the moving object extracting unit 4, is calculated (S9).
- the calculated distance D (j) is smaller than a preset Dmin (S10). If the calculated distance D (j) is smaller than the preset Dmin, the calculated distance D (j) is set to Dmin, and the moving object region identification number j for which the distance is calculated is set to Min (S11). If the calculated distance D (j) is greater than or equal to the preset Dmin, the process of S12 is performed. It is determined whether j + 1 is larger than the number of all past moving object regions acquired from the moving object history storage unit 6 (S12). If j + 1 is larger than the number of all moving object regions in the past, the process of S13 is performed.
- j + 1 is less than or equal to the number of all past moving object areas, j is set to j + 1 and the process returns to S8 to calculate the distance from the latest moving object area to the latest moving object area. To do. If j + 1 is larger than the number of all moving object regions in the past, it is determined whether Dmin is smaller than the threshold value M (S13).
- the moving object tracking unit 5 acquires information on all moving object regions one frame before the moving object region extracted by the moving object extracting unit 4 from the moving object history storage unit, and the moving object extracting unit 4 extracts the information.
- the distance between the position of the moving object area and the position of the moving object area one frame before is calculated, and the moving object of the moving object area one frame before that distance is smaller than the threshold and the shortest distance is extracted by the moving object extraction unit 4 Tracking is performed assuming that the moving objects in the selected moving object region are the same moving object.
- the stationary region determination unit 13 of the moving body tracking unit 5 is a stationary region if the position of the moving body region one frame before is not updated to the position of the moving body region extracted by the moving body region extraction unit 12 of the moving body extraction unit 4. Is determined.
- the threshold value M has been described as being set in advance, it may be a specified value or may be changed dynamically.
- the distance from the position of the moving object region one frame before is determined for each moving object region, and whether or not they are the same moving object. Determine.
- the stationary region determination unit 13 has been described as a process after determining whether or not the moving object in the moving object region extracted by the moving object extraction unit 4 is the same as the moving object in the moving object region one frame before.
- Process 5 (a process for determining whether or not the moving object is the same as the moving object in the previous frame) may be performed.
- FIG. 8 is an image diagram for explaining the operation of the moving object tracking unit 5 in the embodiment of the present invention.
- An image 801 in FIG. 8 shows moving body data one frame before acquired by the moving body tracking unit 5 from the moving body history storage unit 6 in the image one frame before the latest image stored in the primary storage device.
- An image 802 shows moving body data input from the moving body extraction unit 4 in the latest image stored in the primary storage device.
- the image 801 a moving body 803 and a moving body 804 are photographed. Further, the moving body 803 and the moving body 804 are captured in the image 802, and the moving body area 805 of the moving body 803 and the moving body area 806 of the moving body 804 extracted by the moving body extraction unit 4 are illustrated. Further, the image 801 shows the history 807, 808, 809, and 810 of the position of the moving object 804 and the position 811 of the moving object 803 from the moving object data stored in the moving object history storage unit 6.
- the image 802 includes a position 812 of the moving object 803 input from the moving object extraction unit 4, a position 813 of the moving object 804, and moving object position histories 807, 808, 809 of the moving object data acquired from the moving object history storage unit 6.
- the positions 810 and 811 of the moving body one frame before are shown.
- the moving object tracking unit 5 moves the moving object data 804 of the moving object 804 and the position 811 of the moving object 803, which are moving object data one frame before, from the moving object history storage unit 6. Get as position. Then, the distance between the position 813 of the moving object 804 in the moving object region 806 and the position 810 of the moving object region acquired from the moving object history storage unit 6, which is the latest moving object data input from the moving object extraction unit 4, and the input from the moving object extraction unit 4. The distance between the position 813 of the moving body 804 in the moving body area 806 and the position 811 of the moving body area acquired from the moving body history storage unit 6 is calculated as the latest moving body data.
- the distance between the position 813 of the moving object region 806 input from the moving object extraction unit 4 and the position 811 of the moving object region acquired from the moving object history storage unit 6 is smaller than the threshold. Further, it is assumed that the distance between the position 813 of the moving object region 806 input from the moving object extraction unit 4 and the position 810 of the moving object region acquired from the moving object history storage unit 6 is smaller than the threshold. At this time, the position of the moving object region 806 input from the moving object extraction unit 4 is greater than the distance between the position 813 of the moving object region 806 input from the moving object extraction unit 4 and the position 811 of the moving object region acquired from the moving object history storage unit 6.
- the moving object data of the moving object region position 810 and the moving object of the moving object region 806 input from the moving object extraction unit 4 are shorter, so the moving object data of the moving object region position 810 and the moving object of the moving object region 806 input from the moving object extraction unit 4 are the same moving object. It is determined that Then, the position of the moving object region 806 input from the moving object extraction unit 4 is stored as the position of the same moving object stored in the moving object history storage unit 6.
- FIG. 9 is a diagram of an image for explaining the operation of the still region determination unit 13 of the moving object tracking unit 5 in the embodiment of the present invention.
- An image 901 in FIG. 9 shows the tracking position of the moving object data acquired by the moving object tracking unit 5 from the moving object history storage unit 6 in the image one frame before the latest image stored in the primary storage device.
- An image 902 shows a moving object region input from the moving object extracting unit 4 in the latest image stored in the primary storage device.
- An image 901 shows the moving object 903a captured in the image one frame before the latest image
- an image 902 shows the moving object 903b and moving object 904 captured in the latest image
- An image 901 shows a position (905, 906, 907, 908a) where the moving object 903a acquired from the moving object history storage unit 6 is tracked and a moving object region 909 of the moving object 903a.
- An image 902 shows a position (905, 906, 907, 908b) where the moving object 903b is tracked, and a position 911 of the moving object region 910 input from the moving object extraction unit 4.
- the moving object tracking unit 5 receives the position of the moving object region extracted by the moving object extraction unit 4 and the position of the moving object region one frame before acquired from the moving object history storage unit 6. The moving object is determined from the distance. If it is determined that the moving body is the same as the moving body one frame before, the position of the moving body is updated to the position of the moving body area extracted by the moving body extraction unit 4.
- the stationary area determination unit 13 of the moving object tracking unit 5 Is determined to be stationary.
- the moving object 903b of the moving object history storage unit 6 is set as the moving object region 909 of the moving object 903a extracted by the moving object region extracting unit 12 of the moving object extracting unit 4 in the image one frame before. Set the status information of the to static state.
- FIG. 10 is an example of moving object data stored in the moving object history storage unit 6 in the embodiment of the present invention.
- the moving object history storage unit 6 includes, for example, a moving object ID for identifying a moving object, a position of the moving object region (center position or center of gravity position), coordinates of the moving object region, and state information (moving state or (Still state) is stored.
- the moving object data is input from the moving object tracking unit 5 and stored in the moving object history storage unit 6.
- the moving object tracking unit 5 receives the distance between the extracted position of the moving object region and the position of the moving object region stored in the moving object history storage unit 6. Based on the above, it is determined whether the moving body in the extracted moving body area is in a moving state, a stationary state, or a new moving body. Then, the determined result is output to the moving object history storage unit 6 and stored.
- the still area image storage unit 7 uses the latest image stored in the primary storage device.
- the image of the still region B (k) is cut out and output to the still region image storage unit 8 to store the latest image of the still region B (k) (S20).
- FIG. 11 is an example of still area image data stored in the still area image storage unit 8 according to the embodiment of the present invention.
- the still area image storage unit 8 stores, for example, a moving object ID 1101, an image 1102, and an update count 1103.
- the moving object ID 1101 stores the same ID as the moving object ID stored in the moving object history storage unit 6.
- the image 1102 is the latest image obtained by cutting out only a still area. Since the images for the same moving object ID may be updated any number of times in the still region image storage unit 8, the update count 1103 indicates the number of times the image for the same moving object ID has been updated.
- the feature change amount extraction unit 9 performs the following processing.
- the feature change amount extraction unit 9 stores the still region information in the moving body history storage unit 6 (when the stationary state information is stored in the moving body data), the feature change amount extraction unit 9 stores the information in the still region image storage unit 8.
- a difference FD (k) between the image feature amounts of the still region B (k) image and the still image of the latest image stored in the primary storage device is calculated (S24), and the image feature change amount is calculated.
- the still region identification number k + 1 is larger than the number of still regions (S26), and feature amount change amounts are extracted until feature change amounts are extracted for all still regions. If it is less than or equal to the number of still areas, the identifier of the next still area is set and the process returns to S24. If the number is larger than the number of still areas, the calculated feature change amount is output to the still person determination unit 10 and the process of the still person determination unit 10 after S27 is performed.
- FIG. 12 is an image diagram for explaining the operation of the feature change amount extraction unit 9 in the embodiment of the present invention.
- FIG. 12 shows still area image data 1201 stored in the still area image storage unit 8 and the latest image 1202 acquired by the image acquisition unit 3 and stored in the primary storage device.
- the still area image data 1201 indicates that a still area image 1203 corresponding to the moving object ID3 and a still area image 1204 corresponding to the moving object ID4 are stored.
- the feature change amount extraction unit 9 acquires information (position and region coordinates) of the still area corresponding to the moving object ID 3 from the moving object history storage unit 6.
- a latest area 1205 of the latest image 1202 corresponding to the acquired still area is shown in the latest image 1202.
- information on the still area corresponding to the moving object ID 4 is acquired from the moving object history storage unit 6.
- the latest image 1204 shows the still region 1207 of the latest image 1202 corresponding to the acquired still region.
- An image of the moving object 1208 is present in the static area 1207.
- the feature change amount extraction unit 9 then stores the image feature amount of the still region 1203 stored in the still region image storage unit 8 and the image feature amount of the still region 1205 cut out from the latest image acquired from the primary storage device.
- the difference is calculated.
- the calculation of the difference between the feature amounts uses a difference between templates based on template matching, a difference between HOG feature amounts, or a difference between luminance histograms. This difference may be used as a change amount of the image feature amount, or the above process may be repeated for a specified time or a specified number of times, and the sum of the feature amount differences may be used as the change amount of the image feature amount.
- the timing of updating the still region image of the still region image recording unit 8 to the still region of the latest image may be performed every time or when a certain condition is satisfied. May be. For example, when the feature amount difference is equal to or greater than a specified value, the latest image is updated.
- the still person determination unit 10 determines the presence or absence of a person in the still region and reflects the result in the moving object history storage unit 6.
- the stationary person determination unit 10 determines whether or not a specified time has elapsed (S27). If the specified time has not elapsed, the process of S30 is performed. If the specified time has elapsed, it is determined whether or not the feature change amount FV (k) extracted by the feature change amount extraction unit 9 is within a specified range (S28).
- the moving body data of the still region B (k) is deleted from the moving body history storage unit 6, and the image of the still region B (k) is deleted from the still region image storage unit 8 (S29).
- the process proceeds to the process of the person information output unit 11 in S30. If it is within the prescribed range, it is determined that there is a person in the still area B (k), and the process of S30 is performed. As described above, when the still person determination unit 10 determines that there is no person in the still region input from the feature change amount extraction unit 9, the determination is performed by deleting the data of the still region from the moving body history storage unit 6. The data of the still person is reflected in the moving object history storage unit 6.
- FIG. 13 is a diagram showing a feature change amount by which the still person determination unit 10 determines a still person in the embodiment of the present invention.
- FIG. 13 shows feature change amounts 1301, 1302, and 1303.
- the feature change amount 1302 is within a specified range after a specified time.
- the still person determination unit 10 determines that there is a person in the still area of the feature change amount 1302.
- the feature change amount 1301 falls outside the specified range before the specified time. In this case, the still person determination unit 10 determines that there is no person in the still area of the feature change amount 1301.
- the feature change amount 1303 is outside the specified range when the specified time comes. Also in this case, the still person determination unit 10 determines that there is no person in the still area of the feature change amount 1303.
- the timing of updating the image of the same still region in the still region image storage unit 8 to the same still region of the latest image may be performed every time or when a certain condition is satisfied. You may implement.
- the latest image may be updated when the feature amount difference is equal to or greater than a specified value.
- the specified range is shown as a range in which the characteristic change amount is greater than or equal to a certain value, but the specified range may be a range from zero change amount including a state where there is no change amount. .
- the person determination based on the feature change amount will be described.
- a non-human object such as a chair or a desk with a caster also moves.
- the moving object extraction unit 4 and the moving object tracking unit 5 recognize a moving object other than a person as a moving object.
- the moving body tracking unit 5 even if the movement is not recognizable as a moving state, the person moves finely in a stationary state in daily life, and there are rarely people who do not move at all for a long time. .
- An object such as a chair or a desk has a very small change amount (for example, a feature change amount 1303 in FIG. 13). Therefore, when the amount of change is smaller than the specified value, the still region is determined to be an object other than a person.
- the TV or personal computer monitor when a person moves or stops near the TV or personal computer monitor, the TV or personal computer monitor, not the person, may be generated as a still area. In such a still region, the feature change amount of the image is increased by the screen display of the TV or personal computer monitor (for example, the feature change amount 1301 in FIG. 13). Therefore, when the amount of change is larger than the specified range, it can be determined that the object is an object such as a TV or a personal computer monitor, thereby preventing erroneous determination as a person.
- the still person determination unit 10 determines that there is no person, the still region is deleted from the moving object history storage unit 6 and the still region image storage unit 8, the feature change amount extraction unit 9
- the still person determination unit 10 repeats the determination of the presence / absence of a person by repeating the specific time, the specific number of times, and the presence / absence of a person, the determined still area is stored in the moving body history storage unit 6 and the still area image storage. You may delete from the part 8.
- the still region image stored in the still region image storage unit 8 is updated to a still region image cut out from the latest image again. May be. By doing so, other people pass near the person determined by the still image, the image of the moving body area of the other person overlaps the still image, and there are still people in the still area, Even if the image feature amount falls outside the threshold range, it is possible to detect a stationary person from a still area image cut out from the latest image.
- a plurality of still area images cut out from the latest image may be stored in the still area image storage unit 8.
- the still area determined to have no person is deleted from the moving object history storage unit 6.
- the position determined as the still area and the next extracted by the moving object extraction unit 4 When the distance from the position of the moving object region of the image is smaller than the set threshold value, it is determined that the moving object in the stationary region has transitioned to the moving object region, and the still region in the moving object history storage unit 6 is changed to the moving object region. Good.
- the person information output unit 11 acquires the moving body data of the moving body region and the stationary region stored in the moving body history storage unit 6 at predetermined time intervals or according to an instruction from an external device. For example, the number and position of the moving body data are obtained.
- the personal information is output to the external device (S30). Then, it is determined whether or not the detection of the person has been completed (S31). If the detection of the person is finished, the person detection process is finished. If the detection of the person has not been completed, the process returns to S1.
- the destination output by the person information output unit 11 may be a control device of an external device or an external system. Since the stationary person determination unit 10 deletes the stationary region determined to have no person from the moving body history storage unit 6, the moving body region and the stationary region stored in the moving body history storage unit 6 are regions where a person is present.
- the position of the person as the person information is, for example, the center position of the moving object area or the stationary area, or the center of gravity position of the moving object area or the stationary area.
- optimal energy saving control can be performed according to the presence / absence, position, and number of people.
- the lighting device since the human sensor using infrared rays does not react to a stationary person, the lighting device may be turned off even if there is a person.
- even a moving person does not know where the person is in the sensor range, so when many lighting devices such as offices and factories are installed, which lighting device should be dimmed to obtain the optimal brightness I do n’t know if it ’s going to happen.
- a still region is determined based on the distance between the moving object region extracted using the image and the moving object region of the previous frame, and the person is stopped using the amount of change in the image feature amount for this still region. Therefore, it is possible to detect the presence / absence, position, and number of persons including a stationary person. Therefore, it is possible to turn on or turn off the lighting device including a stationary person. Even when many lighting devices such as offices or factories are installed, dimming control of appropriate lighting devices can be performed according to the presence / absence, position, and number of people.
- a human sensor using infrared rays needs to narrow the sensor range and install many sensors, which increases the cost.
- the cost is reduced. Can be suppressed.
- it can be applied to various control systems such as optimal operation of an elevator that shortens waiting time, or crime prevention / safety management based on detection of congestion and entry prohibition.
- the person detection device and the person detection method according to the present invention include the position of the first moving object region extracted from the image and the second moving object region extracted from the image before the image of the first moving object region.
- the still region is determined based on the distance from the position.
- the person information including the stationary person for example,
- the lighting or air conditioning control device can be controlled. Or it can apply to the management system of various apparatuses, such as the operation
- 1 person detection device 2 camera, 3 image acquisition unit, 4 moving object extraction unit, 5 moving object tracking unit, 6 moving object history storage unit, 7 still region storage unit, 8 still region image storage unit, 9 feature change amount extraction unit, 10 Still person determination unit, 11 person information output unit, 12 moving object region extraction unit, 13 still region determination unit.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Automation & Control Theory (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
- Burglar Alarm Systems (AREA)
Abstract
Description
フレーム間差分法で静止物体を検出する従来手法としては、画像中の動いている物体を追尾して、動きがなくなった位置で物体が静止したと認識するトラッキング法などがある。
そこで、画像を用いたトラッキング法により静止している人を検出する方法として、予め記憶した静止している物体の画像領域の画像特徴量と、特定の物体の画像特徴量とを比較することで、静止した物体を特定する方法がある。(例えば、特許文献1。)
また、事前に人の画像特徴量を登録しておき、登録した画像特徴量と類似したときに人と認識する方法も考えられるが、服装や髪形によって特徴量は異なるため、結局は事前に膨大な画像特徴量の収集及び登録が必要となる。
図1は、この発明の実施の形態における、人物検出装置1の構成図である。
図1において、人物検出装置1は、特定のエリアを撮影するカメラ2から画像データを取得して、特定エリアの人の数や位置を検出し、人物情報として出力する。
本実施の形態では、カメラ2は、天井に設置されているとして説明するが、カメラはどこに設置されていてもよい。また、人物検出装置1とカメラ2は、別構成でも一体化した構成でもよい。例えば、カメラ2を天井に設置して、人物検出装置1をフロア、またはビルの制御室に設置し、カメラ2と人物検出装置1をLAN(Local Area Network)ケーブル、または同軸ケーブルで接続する。そして、人物検出装置1が、複数のカメラ2から画像データを取得して、人物検出を行ってもよい。
画像取得部3は、カメラ2からカメラ2で撮影された画像データを取得する。そして、取得した画像データから最新の1フレームの画像を抽出する。そして、抽出した最新の1フレームの画像データを一次記憶装置(図示せず)に記憶する。
尚、静止人物判定部10に、静止領域画像保存部7または特徴変化量抽出部9の処理を含めてもよい。
以上が人物検出装置1の各構成の説明である。
図2~6は、この発明の実施の形態における、人物検出装置1の処理を示すフローチャートである。
まず、図2を用いて、画像取得部3と動体抽出部4の処理について説明する。
画像取得部3は、カメラ2で撮影された画像データをカメラ2から取得し、取得した画像データから最新の1フレームの画像を取得する(S1)。
次に、動体抽出部4は、一次記憶装置に記憶されている最新の1フレームの画像から動体領域を抽出する。
図7は、この発明の実施の形態における、動体抽出部4の処理を説明する画像の図である。
図7の画像701と画像702は、一次記憶装置に記憶されている画像であり、画像701は、画像取得部3で取得された最新の画像702の1フレーム前の画像である。そして、画像701には物体703、画像702には物体704が撮影されている様子を示している。
そして、フレーム間の差分が、予め設定された閾値以上の画像ブロック705を動きブロック706として抽出する(S4)。
画像ブロック705は、画像を均等に分割した領域であり、分割数は画像サイズに応じて予め決めた数である。また、前記閾値は、事前に決めた規定値でもよいし、動的に変更してもよい。
そして、グループ化した動きブロックを包含する矩形を動体領域707とする。このとき、グループ化した動きブロックの数が、予め決められた規定範囲内の場合のみ動体領域としてもよい。そして、動体領域707の位置、例えば動体領域の中心位置または重心位置を算出し(S6)、算出した位置と領域の座標の情報を、動体領域707の動体データとして動体追跡部5に出力し記憶する。
動体追跡部5は、動体履歴記憶部6に記憶されている過去の動体領域の動体データを用いて、動体抽出部4から入力された最新の動体領域の追跡を行う。
まず、動体追跡部5は、1フレーム前の動体領域の位置と最新の動体領域の位置との最短距離Dminに閾値M、最短距離Dminである過去の動体領域の識別番号Minに-1、動体抽出部4から入力される最新の動体領域の識別番号iに1、および動体履歴記憶部6に記憶されている過去の動体領域の識別番号jに1を設定し初期化する(S7)。
動体履歴記憶部6から取得した1フレーム前の動体領域B(j)の位置PB(j)を前位置として取得する(S8)。
次に、動体抽出部4から入力された最新の動体データである動体領域Aの動体の位置PAとPB(j)の距離D(j)を算出する(S9)。
算出した距離D(j)が予め設定したDminよりも小さければ、算出した距離D(j)をDminに設定し、Minに距離を算出した動体領域の識別番号jを設定する(S11)。
算出した距離D(j)が予め設定したDmin以上であれば、S12の処理を行う。
j+1が、動体履歴記憶部6から取得した過去のすべての動体領域の数よりも大きいか否かを判定する(S12)。
j+1が、過去のすべての動体領域の数よりも大きければS13の処理を行う。
j+1が、過去のすべての動体領域の数以下であれば、jにj+1を設定してS8の処理に戻り、次の過去の動体の動体領域に対して、最新の動体領域との距離を算出する。
j+1が、過去のすべての動体領域の数よりも大きければ、Dminが閾値Mよりも小さいか否かを判定する(S13)。
Dminが閾値Aよりも小さければ、PB(Min)をPA(i)とし動体履歴記憶部6の動体データの位置を更新する(S15)。
そして、i+1が動体抽出部4から入力された最新の動体領域の数よりも大きいか否かを判定し(S16)、i=i+1としてS8の処理に戻る。
そして、1フレーム前の動体領域の位置PB(j)が更新されていないかを判定する(S18)。
1フレーム前の動体領域の位置PB(j)が、動体抽出部4から入力された動体領域PA(i)の位置に更新されていなければ、動体領域B(j)の動体が静止している状態である静止領域B(k)として動体履歴記憶部6に記憶する(S19)。
1フレーム前の動体領域の位置PB(j)が、動体抽出部4から入力された動体領域PA(i)の位置に更新されていれば、S21の処理を行う。
また、静止領域判定部13は、動体抽出部4で抽出された動体領域の動体と、1フレーム前の動体領域の動体が同じかどうかを判定した後の処理として説明したが、この動体追跡部5の処理(1フレーム前の動体領域の動体と同じか否かを判定する処理)を行ってもよい。
図8の画像801は、一次記憶装置に記憶されている最新の画像の1フレーム前の画像に、動体追跡部5が動体履歴記憶部6から取得した1フレーム前の動体データを示している。また、画像802は、一次記憶装置に記憶されている最新の画像に、動体抽出部4から入力された動体データを示している。
また、画像801には、動体履歴記憶部6に記憶されている動体データから、動体804の位置の履歴807、808、809、810と、動体803の位置811を示している。
また、画像802には、動体抽出部4から入力された動体803の位置812と、動体804の位置813、および動体履歴記憶部6から取得した動体データの動体の位置の履歴807、808、809と1フレーム前の動体の位置810、811を示している。
そして、動体抽出部4から入力された最新の動体データである動体領域806の動体804の位置813と動体履歴記憶部6から取得した動体領域の位置810との距離、および動体抽出部4から入力された最新の動体データである動体領域806の動体804の位置813と、動体履歴記憶部6から取得した動体領域の位置811との距離を算出する。
図9は、この発明の実施の形態における、動体追跡部5の静止領域判定部13の動作を説明する画像の図である。
図9の画像901は、一次記憶装置に記憶された最新の画像の1フレーム前の画像に、動体追跡部5が動体履歴記憶部6から取得した動体データの追跡位置を示している。また、画像902は、一次記憶装置に記憶された最新の画像に、動体抽出部4から入力された動体領域を示している。
また、画像901は、動体履歴記憶部6から取得した動体903aを追跡した位置(905、906、907、908a)と動体903aの動体領域909を示している。画像902は、動体903bを追跡した位置(905、906、907、908b)と、動体抽出部4から入力された動体領域910の位置911を示している。
図10に示すように、動体履歴記憶部6には、例えば、動体を識別する動体ID、動体領域の位置(中心位置または重心位置)、動体領域の座標、および動体の状態情報(移動状態または静止状態)が記憶されている。
静止領域画像保存手段7は、動体履歴記憶部6に静止領域のデータが入力されると(動体データに静止状態のデータが入力されると)、一次記憶装置に記憶された最新の画像から、静止領域B(k)の画像を切り出し、静止領域画像記憶部8に出力して、静止領域B(k)の最新の画像を記憶する(S20)。
静止領域画像記憶部8には、例えば、動体ID1101、画像1102、更新回数1103が記憶されている。
動体ID1101は、動体履歴記憶部6に記憶された動体IDと同じIDが記憶されている。また、画像1102は、静止領域だけを切り出した最新の画像である。静止領域画像記憶部8には、同じ動体IDに対する画像は何度更新してもよいため、更新回数1103は、同じ動体IDに対する画像を更新した回数を示している。
特徴変化量抽出部9は、静止領域画像記憶部8に静止画像の最新の画像が記憶されると、以下の処理を行う。まず、画像の特徴変化量の初期化(FV(k)=0)を行い、静止領域の識別番号を次の識別番号にする(k=k+1)(S21)。
そして、動体履歴記憶部6に記憶された過去の動体領域の数まで静止領域の画像を取得したか否かを判定する(S22)。
過去の動体領域の数まで取得していなければ、S18の処理に戻る。
過去の動体領域の数まで取得していれば、以下の処理を行う。
そして、特徴変化量抽出部9は、動体履歴記憶部6に静止領域の情報が記憶されると(動体データに静止状態の情報が記憶されると)、静止領域画像記憶部8に記憶された静止領域B(k)の画像と、一次記憶装置に記憶された最新の画像の静止領域の画像との画像の特徴量の差分FD(k)を算出して(S24)、画像の特徴変化量を抽出する。
そして、特徴変化量FV(k)=FV(k)+FD(k)として、FV(k)に特徴変化量を加算して特徴変化量の総和を算出する(S25)。
静止領域の数以下であれば、次の静止領域の識別子を設定してS24に戻る。
静止領域の数より大きければ、算出した特徴変化量を静止人物判定部10に出力し、S27以降の静止人物判定手段10の処理を行う。
図12は、この発明の実施の形態における、特徴変化量抽出部9の動作を説明する画像の図である。
図12では、静止領域画像記憶部8に記憶されている静止領域画像データ1201と、画像取得部3が取得して一次記憶装置に記憶されている最新の画像1202を示している。そして、静止領域画像データ1201には、動体ID3に対応した静止領域の画像1203と、動体ID4に対応した静止領域の画像1204が記憶されていることを示している。
特徴量の差分の算出は、画像間のテンプレートマッチングによる差分、またはHOG特徴量の差分、または輝度ヒストグラムの差分などを利用する。この差分を画像特徴量の変化量としてもよいし、上記の処理を規定の時間あるいは規定の回数だけ繰り返し特徴量差分の総和を画像特徴量の変化量としてもよい。また、このような処理を繰り返すとき、静止領域画像記録部8の静止領域の画像を最新の画像の静止領域に更新するタイミングは、毎回実施してもよいし、ある条件が成立したときに実施してもよい。例えば、特徴量差分が規定値以上になった場合に最新の画像に更新する。
以上が、特徴変化量抽出部9の動作についての画像例での説明である。
静止人物判定部10は、特徴変化量抽出部9から画像の特徴変化量が入力されると、静止領域中の人の有無を判定し、その結果を動体履歴記憶部6に反映する。
まず、静止人物判定部10は、規定時間経過したか否かを判定する(S27)。
規定時間経過していなければ、S30の処理を行う。
規定時間経過していれば、特徴変化量抽出部9で抽出した特徴変化量FV(k)が、規定の範囲内か否かを判定する(S28)。
規定の範囲内であれば、静止領域B(k)に人が居ると判定し、S30の処理を行う。
このように、静止人物判定部10は、特徴変化量抽出部9から入力された静止領域に人が居ないと判定すると、動体履歴記憶部6から静止領域のデータを削除することで、判定した静止している人物のデータを動体履歴記憶部6に反映する。
図13は、この発明の実施の形態における、静止人物判定部10が静止人物を判定する特徴変化量を示す図である。
図13には、特徴変化量1301、1302、1303が示されている。
例えば、特徴変化量1302は、規定の時間後に、規定の範囲内にある。この場合、静止人物判定部10は、特徴変化量1302の静止領域に人が居ると判定する。
また、特徴変化量1301は、規定の時間前に規定の範囲外になる。この場合、静止人物判定部10は、特徴変化量1301の静止領域に人は居ないと判定する。
以上で説明した処理を繰り返すとき、静止領域画像記憶部8の同じ静止領域の画像を最新の画像の同じ静止領域に更新するタイミングは、毎回実施してもよいし、ある条件が成立したときに実施してもよい。例えば、特徴量差分が規定値以上になった場合に最新の画像に更新してもよい。
尚、図13では、規定範囲は、特徴変化量がある値以上の範囲であるとして示しているが、規定範囲は、変化量なしの状態を含む変化量ゼロからの範囲を規定範囲としてもよい。
人が移動する際に、キャスター付きの椅子または机など人以外の物体も移動するケースが多い。このとき、動体抽出部4と動体追跡部5では、人以外の移動した物体も動体として認識してしまう。しかし、動体追跡部5では、移動状態として認識できる程の動きではなくても、人は日常生活の中では静止状態で微細な動きをするものであり、長時間全く動かない人は滅多にいない。
また、静止人物判定部10で人が居ないと判定し続けたときに、改めて最新の画像から切り出した複数の静止領域の画像を静止領域画像記憶部8に記憶しておいてもよい。
また、本実施の形態では、人が居ないと判定した静止領域を動体履歴記憶部6から削除するとして説明したが、例えば、静止領域と判定した位置と、動体抽出部4で抽出された次の画像の動体領域の位置との距離が設定された閾値よりも小さい場合、静止領域の動体が動体領域に遷移したと判定し、動体履歴記憶部6の静止領域を動体領域に変更してもよい。
人物情報出力部11は、所定の時間毎または外部装置からの指示により、動体履歴記憶部6に記憶されている動体領域と静止領域の動体データを取得し、例えばこの動体データの数と位置を人物情報として外部装置に出力する(S30)。
そして、人物の検出が終了しているか否かを判定する(S31)。
人物の検出が終了していれば、人物の検出処理を終了する。
人物の検出が終了していなければ、S1の処理に戻る。
静止人物判定部10は、人が居ないと判定した静止領域を動体履歴記憶部6から削除するため、動体履歴記憶部6に記憶されている動体領域と静止領域が人の居る領域となる。人物情報とする人の位置は、例えば動体領域または静止領域の中心位置、あるいは動体領域または静止領域の重心位置とする。
照明機器の制御では、赤外線を利用した人感センサでは、静止した人には反応しないため、人がいるのに照明機器が消灯してしまうことがある。また、動いている人でも、センサ範囲のどの位置に人が居るのか分からないため、オフィスや工場など多くの照明機器が設置されている場合には、どの照明機器を調光すれば最適な明るさになるのかが分からない。
また、省エネ以外では、待ち時間を短縮するエレベータを最適に運行、または混雑度や立ち入り禁止検知による防犯・安全管理など、様々な制御システムに適用することができる。
Claims (6)
- 撮影された画像を取得する画像取得部と、
該画像取得部で取得した画像から抽出した第1の動体領域の位置と、該第1の動体領域の画像より前の画像から抽出した第2の動体領域の位置との距離に基づいて、前記第2の動体領域の動体が静止状態である静止領域と判定する静止領域判定部と、
前記静止領域の画像の特徴量の変化量を用いて人物の有無を判定する静止人物判定部とを備えたことを特徴とする人物検出装置。 - 前記静止領域判定部は、前記第1の動体領域の位置と、前記第2の動体領域の位置との距離が予め設定された閾値よりも大きく、且つ前記第1の動体領域の画像から抽出された他のすべての動体領域の位置と、前記第2の動体領域の位置との距離が予め設定された閾値よりも大きい場合、前記第2の動体領域を静止領域と判定することを特徴とする請求項1記載の人物検出装置。
- 前記静止人物判定部は、前記静止領域の特徴量の変化量が設定した範囲の場合、静止領域に人が居ると判定することを特徴とする請求項1または請求項2の何れか1項に記載の人物検出装置。
- 前記動体領域の情報を動体の情報として記憶する動体履歴記憶部と、
前記動体履歴記憶に記憶された動体の情報から人物の情報を外部に出力する人物情報出力手段を更に備え、
前記静止人物判定手段は、人物有りと判定した静止領域の情報を動体の情報として前記動体履歴記憶部に記憶し、
前記人物情報出力手段は、前記動体履歴記憶部に記憶された動体の情報から、人の位置と人の数を人物の情報として出力することを特徴とする請求項1~3の何れか1項に記載の人物検出装置。 - 前記静止人物判定は、人物無しと判定した前記静止領域に対して、前記第1の動体領域を抽出した画像より後の画像から新たに抽出した当該静止領域の情報を用いて人物有りと判定すると、当該静止領域の情報を動体の情報として前記動体履歴記憶部に記憶することを特徴とする請求項4に記載の人物検出装置。
- 撮影された画像を用いて人物を検出する人物検出装置の人物検出方法において、
撮影された画像を取得する画像取得ステップと、
該画像取得ステップで取得した画像から抽出した第1の動体領域の位置と、該第1の動体領域の画像より前の画像から抽出した第2の動体領域の位置との距離に基づいて、前記第2の動体領域の動体が静止状態である静止領域と判定する静止領域判定ステップと、
前記静止領域の画像の特徴量の変化量を用いて人物の有無を判定する静止人物判定ステップとを備えたことを特徴とする人物検出方法。
Priority Applications (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2016507283A JP6266088B2 (ja) | 2014-03-11 | 2015-01-28 | 人物検出装置および人物検出方法 |
CN201580012286.8A CN106104631B (zh) | 2014-03-11 | 2015-01-28 | 人物检测装置及人物检测方法 |
US15/118,420 US10026019B2 (en) | 2014-03-11 | 2015-01-28 | Person detecting device and person detecting method |
GB1613448.8A GB2537780B (en) | 2014-03-11 | 2015-01-28 | Person detecting device and person detecting method |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2014047281 | 2014-03-11 | ||
JP2014-047281 | 2014-03-11 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2015136828A1 true WO2015136828A1 (ja) | 2015-09-17 |
Family
ID=54071291
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2015/000356 WO2015136828A1 (ja) | 2014-03-11 | 2015-01-28 | 人物検出装置および人物検出方法 |
Country Status (5)
Country | Link |
---|---|
US (1) | US10026019B2 (ja) |
JP (1) | JP6266088B2 (ja) |
CN (1) | CN106104631B (ja) |
GB (1) | GB2537780B (ja) |
WO (1) | WO2015136828A1 (ja) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP6573744B2 (ja) * | 2017-03-06 | 2019-09-11 | 三菱電機株式会社 | 物体追跡装置及び物体追跡方法 |
CN109812934B (zh) * | 2019-01-29 | 2021-05-25 | 广东美的制冷设备有限公司 | 送风系统、送风方法及计算机可读存储介质 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2010199701A (ja) * | 2009-02-23 | 2010-09-09 | Fujitsu Ltd | 画像処理装置、画像処理方法、及び、画像処理プログラム |
JP2011198244A (ja) * | 2010-03-23 | 2011-10-06 | Hiromitsu Hama | 対象物認識システム及び該システムを利用する監視システム、見守りシステム |
WO2013187047A1 (ja) * | 2012-06-13 | 2013-12-19 | パナソニック株式会社 | 物体検出装置 |
Family Cites Families (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH05205175A (ja) | 1992-01-24 | 1993-08-13 | Hitachi Ltd | 物体検出装置 |
JP3230509B2 (ja) | 1998-06-26 | 2001-11-19 | 日本電気株式会社 | 動画像処理装置 |
US7133537B1 (en) * | 1999-05-28 | 2006-11-07 | It Brokerage Services Pty Limited | Method and apparatus for tracking a moving object |
JP2002197463A (ja) | 2000-12-26 | 2002-07-12 | Matsushita Electric Ind Co Ltd | 挙動検出装置および挙動検出システム |
US7167576B2 (en) * | 2001-07-02 | 2007-01-23 | Point Grey Research | Method and apparatus for measuring dwell time of objects in an environment |
JP2004094518A (ja) | 2002-08-30 | 2004-03-25 | Oki Electric Ind Co Ltd | 人物追跡装置、人物追跡方法及びプログラム |
CN100474339C (zh) * | 2004-08-03 | 2009-04-01 | 松下电器产业株式会社 | 人物判定装置 |
US20080074496A1 (en) * | 2006-09-22 | 2008-03-27 | Object Video, Inc. | Video analytics for banking business process monitoring |
FR2916562B1 (fr) * | 2007-05-22 | 2010-10-08 | Commissariat Energie Atomique | Procede de detection d'un objet en mouvement dans un flux d'images |
JP5154461B2 (ja) | 2009-01-27 | 2013-02-27 | セコム株式会社 | 移動物体追跡装置 |
JP5214533B2 (ja) | 2009-05-21 | 2013-06-19 | 富士フイルム株式会社 | 人物追跡方法、人物追跡装置および人物追跡プログラム |
JP5570176B2 (ja) * | 2009-10-19 | 2014-08-13 | キヤノン株式会社 | 画像処理システム及び情報処理方法 |
WO2012014430A1 (ja) * | 2010-07-27 | 2012-02-02 | パナソニック株式会社 | 移動体検出装置および移動体検出方法 |
JP2012054810A (ja) * | 2010-09-02 | 2012-03-15 | Olympus Corp | 画像処理装置、画像処理方法、撮像装置、および画像処理プログラム |
JP5758165B2 (ja) | 2011-03-30 | 2015-08-05 | セコム株式会社 | 物品検出装置および静止人物検出装置 |
JP5669648B2 (ja) | 2011-03-30 | 2015-02-12 | セコム株式会社 | 異常検知装置 |
CN102496024B (zh) * | 2011-11-25 | 2014-03-12 | 山东大学 | 智能监控中以特征帧触发的事件检测方法 |
US9247211B2 (en) * | 2012-01-17 | 2016-01-26 | Avigilon Fortress Corporation | System and method for video content analysis using depth sensing |
US9330468B2 (en) * | 2012-02-29 | 2016-05-03 | RetailNext, Inc. | Method and system for analyzing interactions |
JPWO2013187037A1 (ja) * | 2012-06-14 | 2016-02-04 | パナソニックIpマネジメント株式会社 | ミキサ、ミキサ装置、およびこれらを備えた無線受信機 |
JP6276519B2 (ja) * | 2013-05-22 | 2018-02-07 | 株式会社 日立産業制御ソリューションズ | 人数計測装置および人物動線解析装置 |
-
2015
- 2015-01-28 WO PCT/JP2015/000356 patent/WO2015136828A1/ja active Application Filing
- 2015-01-28 US US15/118,420 patent/US10026019B2/en not_active Expired - Fee Related
- 2015-01-28 JP JP2016507283A patent/JP6266088B2/ja active Active
- 2015-01-28 GB GB1613448.8A patent/GB2537780B/en not_active Expired - Fee Related
- 2015-01-28 CN CN201580012286.8A patent/CN106104631B/zh active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2010199701A (ja) * | 2009-02-23 | 2010-09-09 | Fujitsu Ltd | 画像処理装置、画像処理方法、及び、画像処理プログラム |
JP2011198244A (ja) * | 2010-03-23 | 2011-10-06 | Hiromitsu Hama | 対象物認識システム及び該システムを利用する監視システム、見守りシステム |
WO2013187047A1 (ja) * | 2012-06-13 | 2013-12-19 | パナソニック株式会社 | 物体検出装置 |
Also Published As
Publication number | Publication date |
---|---|
JP6266088B2 (ja) | 2018-01-24 |
GB2537780A (en) | 2016-10-26 |
US10026019B2 (en) | 2018-07-17 |
JPWO2015136828A1 (ja) | 2017-04-06 |
CN106104631A (zh) | 2016-11-09 |
CN106104631B (zh) | 2019-03-01 |
US20170169312A1 (en) | 2017-06-15 |
GB2537780B (en) | 2019-04-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
KR101337060B1 (ko) | 화상 처리 장치 및 화상 처리 방법 | |
CN107958258B (zh) | 用于追踪限定区域中的对象的方法和系统 | |
JP5369175B2 (ja) | 映像によるエレベータドアの検出装置および検出方法 | |
US20160142680A1 (en) | Image processing apparatus, image processing method, and storage medium | |
KR102376479B1 (ko) | 인공지능 기반 객체 자동 인식을 위한 cctv 제어 방법, 장치 및 시스템 | |
WO2018096787A1 (ja) | 人物行動監視装置および人物行動監視システム | |
US8165348B2 (en) | Detecting objects crossing a virtual boundary line | |
JPWO2017122258A1 (ja) | 混雑状況監視システム及び混雑状況監視方法 | |
US20070229663A1 (en) | Image processing apparatus, monitoring camera, and image monitoring system | |
US9336444B2 (en) | System and method for occupancy detection using differential image sensing and optical markers | |
JP2006221355A (ja) | 監視装置及び監視システム | |
US10373015B2 (en) | System and method of detecting moving objects | |
TWI438702B (zh) | 影像環境邊界之動態設定方法及人員活動內容之即時判定方法 | |
US20130293721A1 (en) | Imaging apparatus, imaging method, and program | |
KR20150051630A (ko) | 사람들의 움직임 불규칙성을 이용한 폭력 행위 검출 장치 및 방법 | |
WO2019220589A1 (ja) | 映像解析装置、映像解析方法、及びプログラム | |
JP6266088B2 (ja) | 人物検出装置および人物検出方法 | |
KR101826669B1 (ko) | 동영상 검색 시스템 및 그 방법 | |
JP5758165B2 (ja) | 物品検出装置および静止人物検出装置 | |
KR20200056498A (ko) | 무인 매장 내 안전사고 예방을 위한 딥러닝 기반 실시간 잔류자 감지 방법 및 이를 이용한 잔류자 감지 시스템 | |
JP6253950B2 (ja) | 画像監視システム | |
KR102464196B1 (ko) | 빅데이터 기반 영상 감시시스템 | |
JP2009044526A (ja) | 撮影装置、撮影方法、人物認識装置および人物認識方法 | |
KR102270858B1 (ko) | 객체 추적을 위한 cctv 카메라 시스템 | |
KH et al. | Smart CCTV surveillance system for intrusion detection with live streaming |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 15762442 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 2016507283 Country of ref document: JP Kind code of ref document: A |
|
ENP | Entry into the national phase |
Ref document number: 201613448 Country of ref document: GB Kind code of ref document: A Free format text: PCT FILING DATE = 20150128 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1613448.8 Country of ref document: GB |
|
WWE | Wipo information: entry into national phase |
Ref document number: 15118420 Country of ref document: US |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 15762442 Country of ref document: EP Kind code of ref document: A1 |