JP2001148011A - Method and device for identifying small animal by image recognition - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a small animal detecting method and a small animal detecting device which can judge an intruder, a small animal, an insect, etc., by an image recognizing means on the basis of image data picked up by a camera. SOLUTION: This device consists of an image picking-up means 1 such as a camera, an image storing means 3, an image region setting means 4 setting a recognition object image region, a difference operating means 5 for operating and outputting differential image data between background image data 31 and currently monitored image data, an abnormal shape extracting means 6 for extracting an abnormal shape on the basis of a preliminarily registered and stored threshold, a bottom part grounding position detecting means 7 for detecting the grounding position to floor of an abnormal shape, a real shape extracting means 9 correctively operating shape dimensions and extracting real shape data, an intruding object deciding means 11 deciding the type of an intruding object according to the real shape data and the acceptance/rejection data of bottom part ground-contact position detection and a judging condition memory 10 storing the judging conditions of the intruding object deciding means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an intruding object recognizing method and an intruding object recognizing apparatus for judging small animals, insects and the like by recognizing an image taken by a camera by image recognition.

[0002]

2. Description of the Related Art An intruding object recognizing apparatus for extracting an intruder, an intruding object, and the like from image data captured by a surveillance camera by image processing means is widely known. Methods for recognizing an intruding object are roughly divided into two types. The first is a method of directly recognizing the shape and movement of an intruding object by image processing means from difference image data between background image data registered in advance and current monitoring image data. This method is not only expensive because it involves image processing in units of pixels, but also for humans, small animals,
It is difficult to judge insects and the like over the entire monitoring area. Second, the difference image data set as the recognition target area and the average value or peak value of the image in the area are compared with a pre-registered threshold value for determining an abnormality, and the calculation is performed. This is a simple method for determining the presence or absence of an intruding object.
This method is small and can be manufactured at low cost, but the image size in the monitoring screen greatly varies depending on the distance from the camera to the monitoring target object, and it is not possible to make a detailed determination of a person, a small animal, an insect, or the like.

The intruder detection / warning device according to the prior art cannot perform detailed judgment of trees such as humans, small animals, insects, and the like. There are problems such as erroneous determination and issuing a false alarm. There is a need for a small animal identification method and a small animal identification device that are simple and capable of recognizing humans and small animals over the entire monitoring area.

[0004]

SUMMARY OF THE INVENTION In view of the problems and problems described above, the present invention provides a simple image recognizing means for recognizing image data picked up by a television camera by an intruder.
An object of the present invention is to propose a small animal identification method and a small animal identification device capable of judging small animals and insects.

[0005]

SUMMARY OF THE INVENTION An object identification method for capturing an object to be monitored by an image capturing means such as a camera, converting an image signal into digital image data, and identifying a person or a small animal by image recognition. Image storage means for temporarily storing data, section setting means for setting a recognition target area of image data stored in the image storage means, background image data and current monitoring image registered and stored in the image storage means Difference calculation means for calculating and outputting difference image data from the data, abnormal shape extraction means for extracting an abnormal shape based on a threshold value previously registered and stored from the difference image data, and a contact position of the lower end portion of the abnormal shape to the floor surface And a real shape for extracting the real shape data by correcting the shape and dimension of the abnormal shape based on the distance from the camera or the position of the demarcation area. The intruding object judging means for judging the type of the intruding object from the actual shape data and the data on whether or not the lower end contact position can be detected, and a judgment condition memory storing judgment conditions of the intruding object judging means. The background image data and the current monitoring image data in a state where there is no intruding object are temporarily stored in the image storage means, and the difference calculating means converts the background image data and the current monitoring image data of the recognition target area set by the area setting means. The difference image data is read out from the image storage means and the difference image data between them is computed and output. The abnormal shape extraction means extracts the abnormal shape data from the difference image data based on a threshold value registered and stored in advance. The bottom shape of the floor is detected, and data on whether or not ground contact is made with the floor surface is detected. Is calculated and the distance correction data is used to extract, from the abnormal shape data, actual shape data of a normalized actual size that does not depend on the distance from the camera, and the intruding object determination means performs the actual shape data and the From the propriety data, based on the judgment condition stored in advance in the judgment condition memory, a person, a small animal,
Judgment and output of the type of intruding object such as insects.

[0006] The area setting means sets the area to be recognized at the same size and at equal intervals, sets the area to be recognized at an arbitrary size and along an equidistant line from the camera, or The section is set to have an arbitrary size inversely proportional to the distance from the camera and along an equidistant line from the camera.

The abnormal area of the abnormal shape extracting means is extracted by using a threshold value of the hue average value in the same area or the same section, which is extracted by using a threshold value of an average value or a sum value in the same area or the same area. Extraction is performed using a threshold value of the number of abnormal sections in the same section, or extraction is performed using a threshold value of the number of abnormal pixels in the same section.

[0008] The detection of the lower end contact position of the lower end contact position detecting means detects a singularity or discontinuity of continuity between the lower end of the abnormal shape and the floor surface, and detects the presence or absence of a shadow at the lower end of the abnormal shape. Alternatively, an auxiliary lighting is additionally installed to detect the presence or absence of a shadow at the lower end of the abnormal shape.

The distance correction data of the actual shape extraction means is generated based on binary imaging data of an auxiliary optical system additionally provided in the imaging means.

[0010] The type determination of the intruding object by the intruding object determining means is performed based on the area value of the actual shape data, based on the shape and dimension value of the actual shape data, the presence or absence of the flying object is determined based on the availability data, or abnormal. The determination is made based on a combination condition of the shape data and the distance data or the area data.

[0011] In an intruding object discriminating apparatus for capturing an object to be monitored by an imaging means such as a camera, converting an image signal into digital image data, and identifying a person or a small animal by image recognition, a camera and an image captured by the camera A / D for converting the converted image signal into digital image data, a background image memory for storing background image data free of an intruding object, an image memory for cyclically storing real-time image data, and recognition of the image data A zone setting unit that sets a target zone, a zone setting function memory that stores an equal interval setting function required when the zone setting unit sets a recognition target zone, and a zone setting unit that reads out A difference calculation unit for calculating a difference between the background image data and the image data and calculating and outputting the difference image data; and a threshold data of an average value for extracting an abnormal area or an abnormal area. A threshold memory storing data,
An abnormal segment extracting unit that extracts an abnormal segment or abnormal segment from the difference image data based on the threshold data, an approximate shape extracting unit that extracts an approximate shape from the abnormal segment or abnormal segment data, and temporarily stores the approximate shape An approximate shape memory, a lower end detecting unit that detects a lower end of the approximate shape, a grounding position on the floor surface of the lower end or a grounding position detecting unit that detects whether or not grounding is possible, and a camera based on the grounding position data. A distance calculation unit for calculating and outputting distance data, a distance correction calculation unit for performing shape correction calculation of the approximate shape based on the distance data, a real shape extraction unit for extracting real shape data from the shape correction calculation result, and an intruding object A judgment condition memory for registering and storing judgment conditions such as shape and area value for judging the type;
It is composed of an intruding object judging unit for judging and outputting the type of intruding object such as a person, a small animal, and an insect based on judgment conditions, actual shape data, data on whether or not it can touch the ground.

[0012] An equidistant area setting function for setting the area to be recognized in an arbitrary size and along an equidistant line from the camera is additionally stored in the area setting function memory, or the area to be recognized is set from the camera. Further, a distance weight setting function having an arbitrary size inversely proportional to the distance and set along an equidistant line from the camera is additionally stored.

[0013] A shadow detecting means for detecting a shadow formed at a lower end portion of the intruding object is additionally provided in the contact position detecting section.

[0014] A shadow detecting means for detecting a shadow formed at the lower end of the intruding object and an auxiliary lighting for facilitating the shadow at the lower end of the intruding object are additionally installed in the ground position detecting section.

A binary imaging unit having an auxiliary optical system in the camera, a difference calculation unit that calculates and outputs a difference time between the image data of the binary shooting, and a distance data from the camera to the imaging target is converted and output based on the difference time. A distance data generation unit having a distance conversion unit and the like is additionally installed.

An auxiliary optical system having an auxiliary lens and a half mirror or a prism;
It is composed of a liquid crystal shutter for selectively inputting an optical image from the main lens and the auxiliary lens system.

The threshold memory additionally stores an abnormal section number threshold for extracting an abnormal area based on the number of abnormal sections in the same area. An abnormal pixel extracting an abnormal area based on the number of abnormal pixels in the same area. Additional storage of several thresholds, additional storage of abnormal shape extraction, and infrared image threshold of the average value of infrared or near-infrared images in the same area or section, or extraction of abnormal shape in the same area or section The hue threshold value of the average hue value is additionally stored.

[0018] The shape and size conditions for judging the type of the intruding object based on a combination of the vertical and horizontal dimensions are additionally stored in the judgment condition memory. The condition is additionally stored, or an imaging failure determination condition for determining an abnormal image due to camera occlusion, partial illumination occlusion, camera failure, or the like is additionally stored.

[0019]

FIG. 1 is a schematic diagram showing a first embodiment of a small animal identification apparatus according to the present invention, and a bird's-eye view of a monitoring area. FIG. 2 is a diagram showing an example of a captured image display on a monitor and a configuration conceptual diagram of a recognition target area according to the present invention. FIG. 3 is a flowchart of a method for recognizing an intruding object according to the present invention. FIG.
FIG. 3 is a block diagram of a main part of a second embodiment of the small animal identification device according to the present invention. FIG. 5 is a flowchart illustrating a method for determining an intruding object according to the present invention. Using FIG. 1 to FIG.
A method for recognizing an intruding object of the small animal identification device according to the present invention will be described. The small animal identification device according to the first embodiment shown in FIG. 1B includes an imaging unit 1 such as a camera for imaging an object to be monitored.
A / D2 for converting an image signal into digital image data,
Image storage means 3 for temporarily storing digital image data, area setting means 4 for setting a recognition target area for image data stored in image storage means 3, background image data 31 registered and stored in the image storage means Calculating means 5 for calculating and outputting the difference image data between the current and the surveillance image data, abnormal shape extracting means 6 for extracting the abnormal shape based on the threshold value registered and stored in advance from the difference image data, the floor at the lowermost end of the abnormal shape A bottom contact position detecting means 7 for detecting a contact position on the surface,
The actual shape extracting means 9 for correcting the abnormal shape based on the distance from the camera 1 or the position of the area and calculating the actual shape data, and extracting the actual shape data. The standard configuration includes an intruding object determining means 11 for determining and a determining condition memory 10 storing the determining conditions of the intruding object determining means.

The intruder monitoring environment shown in FIG. 1A is a room about 7 m wide × about 11 m deep, and the camera 1 is installed at a position about 3 m high on the rear wall W 4. The imaging field of view covers from a floor position of about 1 m from the rear wall surface W4 to a position of about 2.7 m in height of the front wall surface W2. Rear wall W4
More than 1.5m near human H1 and cat A1, butterfly B in the air
People H2 and cat A2 around 1,3m, Butterflies F2,7 in the air
m, cat A3, butterfly F3 in the air, 1
There is a human H4 and a cat near 0m, and a butterfly F4 in the A4 air.
FIG. 2A shows an image captured by the camera 1 displayed on a monitor. In this figure, the bottom of the screen is the nearest field of view of the camera and represents the floor 1 m from the rear wall W4, and the top of the screen is the farthest field of the camera and the height of the front wall W2 is 2.7 m.
Of the wall. The side walls W1 and W3 of the room are displayed on the left and right sides of the screen. The display image of the person or small animal to be monitored is displayed small in inverse proportion to the distance from the camera 1. The butterfly F1 closest to the camera 1 is displayed larger than the head of the person H1 and the size of the cat A1. Further, the butterfly F2 near 3 m is displayed much larger than the human H4 and the cat A4 near 10 m. In this state, the type of the object cannot be determined based on the size or area of the image.

With reference to FIG. 3, an outline of a method for recognizing and determining an intruding object type according to the present invention will be described. An image of the monitoring target area is captured by the camera 1 in ST1. The image signal from the camera 1 is converted into digital image data by the A / D 2 in ST2. The background image data and the current monitoring image data in the state where there is no intruding object are temporarily stored in the image storage means 3 and ST3, and the difference calculating means 5 sets the background image data 31 of the recognition target area set by the area setting means 4 in ST4 and ST5 in which the current monitoring image data is read from the image storage means 3 and the difference image data between the two is calculated and output. The abnormal shape extraction means 6 extracts the approximate shape together with ST6 which extracts the abnormal shape data based on the threshold value registered and stored in advance from the difference image data ST7. The lower end contact position detecting means 7 detects the lower end portion of the abnormal approximation shape and detects the contact position data on the floor surface or the data on whether or not the contact is possible with the floor (ST8). The actual shape extracting means 9 uses the distance correction data together with ST9 for calculating and outputting distance correction data from the ground contact position data, and uses the distance approximation data from the abnormal approximation shape data to provide normalized actual shape data having actual dimensions independent of the distance from the camera. ST10 for extracting. The intruding object judging means 11 judges and outputs the type of the intruding object such as a person, a small animal, or an insect based on the actual shape data and the availability data based on the ST13 judgment condition stored in the judgment condition memory 10 in advance (ST14).

Next, a detailed description will be given for each element of the present invention. A first embodiment of setting a recognition target area is shown in FIG.
As shown in the figure, the divisions P of the same size are arranged at equal intervals in the XY direction. One plot area is vertical mx horizontal n (m = 16,
n = 12) = 192, and one section is vertical y
X section x (y = 10, x = 10) = 100 sections K, and one section is composed of 8 × 8 = 64 pixels G. This means that the image of the camera 1 has 960 effective horizontal scanning lines and 1280 effective pixels in the horizontal direction.

In the second embodiment, equidistant lines r2 (2m) and r3 (3) from the camera 1 indicated by dotted lines in FIG.
m), r5 (5 m), r7 (7 m), r10 (10 m)
In addition, though not shown, in the third embodiment, the size of one section is reduced in inverse proportion to the distance. Zones surrounded by these dotted lines are Z1, Z2,
.. Can be expressed as Z3, Z4, Z5,.

An abnormal area or an abnormal area of the abnormal shape extracting means 6 is extracted based on a threshold registered and stored in advance. In the first embodiment, the threshold value of the average value or the sum value of the levels in the same area or the same section is set, and in the second embodiment, the threshold value of the average hue value in the same area or the same section is set in the third embodiment. In the example, the threshold for the number of abnormal sections in the same area is extracted, and in the fourth embodiment, the threshold is used for the number of abnormal pixels in the same area. The first and second embodiments are analog methods. In the third and fourth embodiments,
As in the first and second embodiments, the detection of an abnormal pixel is determined by an analog value. Thereafter, a digital method such as the number of abnormal pixels and the number of abnormal sections is used, and image processing after extraction can be easily performed.

In the first embodiment, the lower end contact position detection means 7 detects the lower end contact position in the singular point of the continuous image of the extracted lower end portion of the abnormal shape and the floor surface (meaning the contact boundary line).
Alternatively, a method of detecting discontinuity (meaning that the object is floating above the floor) is detected. In the second embodiment, a method of detecting the presence or absence of a shadow at the lower end of the abnormal shape is used. In the third embodiment, A method of arranging auxiliary lighting (not shown) on the ceiling and detecting the presence or absence of a shadow at the lower end of the abnormal shape is taken,
Auxiliary lighting makes it easier to cast shadows on the floor.

The detailed configuration of the second embodiment of the small animal identification device according to the present invention will be described with reference to FIG. The small animal identification device according to the second embodiment includes a camera 1a for capturing an image of an object to be monitored, and an A / A for converting an image signal into digital image data.
D2, a background image memory 3b for storing background image data in which no intruding object exists, an image memory 3a for cyclically storing real-time image data, an area setting unit 4a for setting an area to be recognized for the image data, A zone setting function memory 4b storing a zone setting function required when the zone setting unit 4a sets a recognition target zone, a difference operation is performed between the background image data read by the zone setting unit and the image data, and the difference is calculated. A difference calculation unit 5a for calculating and outputting image data, a threshold memory 6b storing threshold data of an average value for extracting an abnormal area or an abnormal section, and an abnormal area or an abnormal area based on the threshold image data. An abnormal area extracting unit 6a for extracting an abnormal area, an approximate shape extracting unit 6c for extracting an approximate shape from abnormal area or abnormal area data, and an approximate form for temporarily storing the approximate shape. A memory 6d, a lower end detecting unit 7a for detecting a lower end of an approximate shape, a grounding position detecting unit 7b for detecting whether or not the lower end touches a floor surface, or whether or not a touchdown is made; distance data from the camera 1 based on the grounding position data , A distance-calculation section 9a that stores a table of the distance between the area and the camera, and a distance correction calculation section 9a that performs a shape correction calculation of the approximate shape based on the distance data. A real shape extraction unit 9d for extracting real shape data from the result of the shape correction calculation, a judgment condition memory 10 for registering and storing judgment conditions for judging the type of intruding object, judgment conditions, actual shape data, groundability data, etc. hand,
An intruding object judging unit 11A for judging and outputting an intruding object type such as a person, a small animal, an insect, a system memory 13 storing an operation sequence program of each unit in the apparatus, and issuing a control command to each unit in the apparatus according to the operation sequence program. The control unit 14 is configured as a standard.

Further, shadow detection means (not shown) for detecting a shadow formed at the lower end of the intruding object, shadow detection means (not shown) for detecting a shadow formed at the lower end of the intruding object, Auxiliary illumination (not shown) for facilitating shadowing at the lower end of the object, an auxiliary optical system having an auxiliary lens and a half mirror or a prism, and a liquid crystal shutter for selectively inputting an optical image from the main lens and the auxiliary lens system. A distance having a binary imaging unit 1b, a difference calculation unit 8a for calculating and outputting a difference time between image data of the binary imaging, and a distance conversion unit 8b for converting and outputting distance data from the camera 1a to the imaging target based on the difference time. Data generating means 8 and the like are additionally provided.

The detailed operation will be described along the signal flow. The image of the monitoring target area is captured by the camera 1a. An image signal from the camera 1a is converted into digital image data by A / D2. Background image data without an intruding object is stored in the background image memory 3b. The current monitoring image data is temporarily stored in the image data memory 3a, and the difference calculation unit 5a stores the background image data and the current monitoring image data of the recognition target area set by the area setting unit 4a in the background image memory 3b and the image data memory 3a. The data is read out and the difference image data between the two is output. The setting of the recognition target area is performed based on a setting function registered in advance in the area setting function memory 4b.

The area setting function pre-registered in the area setting function memory 4b is an equal distance setting function, an equidistant area for setting a recognition target area with an arbitrary size and along an equidistant line from the camera. A setting function, a distance weight setting function for setting the recognition target area in an arbitrary size inversely proportional to the distance from the camera and along an equidistant line from the camera, and the like are selected.

The abnormal area extracting unit 6a extracts an abnormal area from the difference image data based on a threshold value previously registered and stored in the threshold value memory 6b. The threshold value registered and stored is the threshold value of the average value or the sum value in the same area or the same area,
A threshold value of a similar hue average value, a threshold value of the number of abnormal sections in the same image section, a threshold value of the number of abnormal pixels in the same section, a threshold value of an average or a sum of infrared or near infrared images, and the like are selected.
The infrared or near-infrared image is captured as needed using an auxiliary infrared pass filter (not shown), and detects infrared and near-infrared images emitted by the body temperature of the animal.

The approximate shape extracting unit 6c extracts an approximate external shape (approximate shape) of the abnormal segment from the extracted abnormal segment data and temporarily stores the same in the approximate shape memory 6d. The lower end detector 7a detects the lowermost end of the approximate shape and inputs the same to the contact position detector 7b. The role of the contact position detector 7b is
The purpose is to detect where the intruding object is on the floor surface to calculate the distance from the camera 1a to the intruding object, and to detect whether the intruding object is a flying object that does not touch the floor.
The ground contact position detects the singularity or discontinuity of the continuity between the lower end of the abnormal shape and the floor, detects the presence or absence of a shadow at the lower end of the abnormal shape, or is additionally installed at the lower end of the abnormal shape. A technique such as detecting the presence or absence of a shadow formed by auxiliary lighting is used. The method of detecting a shadow can effectively detect a grounding portion under an environment such as a nightlight at night or lighting provided on a camera. In the case of an aerial flying object, a shadow appears but is generated at a discontinuous position with respect to the flying object, so that it is easy to determine that the object is not on the ground.

The distance calculation unit 9b includes a ground position detection unit 7b
The distance data is calculated and output from the area-distance table 9c on the basis of the position data from. The distance correction calculation unit 9a
The approximate shape data is read from the approximate shape memory 6d, and a correction operation is performed using the distance data to calculate corrected approximate shape data. The actual shape extraction unit 9d extracts actual shape data such as a shape dimension and an area that does not depend on the distance from the camera.

A binary image pickup means for additionally inputting an optical image from the auxiliary optical system 1 b composed of a half mirror, a prism and the like to an image pickup device such as a CCD of a main camera unit through a liquid crystal optical shutter is provided. , The two-dimensional image data slightly shifted in time is generated. The difference time calculation unit 8 calculates and outputs the time difference, and the distance conversion unit 8b converts and outputs distance data to the target object from the distance between the main optical system and the auxiliary optical system. The distance data is input to the distance correction calculator 9a. This method is somewhat expensive, but can detect accurate distance data.

The intruding object determination unit 11A determines and outputs the type of intruding object such as a person, a small animal, or an insect based on the actual shape data and the propriety data based on the judgment conditions stored in the judgment condition memory 10 in advance. Judgment conditions stored in advance include shape conditions such as shape and area values, shape and dimension conditions determined by a combination of vertical and horizontal dimensions, and abnormal flying object determination determined by a combination of an airborne object and a critical area. There are conditions, such as an imaging failure determination condition for determining an abnormal image due to camera occlusion, partial illumination occlusion, camera failure, and the like. The shape and size conditions determined based on the combination of the vertical and horizontal dimensions are effective for determining whether an intruder crawls in, an infant, or a buckled pedestrian.

An operation flow for determining the type of the intruding object will be described with reference to FIG. First, a detection result is input to one terminal of the intruding object determination unit 11A from the ground position detection unit 7b in ST14a. When the contact position cannot be detected (no), it is determined that the object is a flying object such as an insect. ST14b where the actual shape data is input to the other input terminal when the detection of the contact position is detected (yes). If the actual shape size is larger than S1 (yes), ST14C determined to be an intruder. further,
If smaller than S1 and larger than S2 (yes), it is determined to be a small animal (1) such as a dog or a cat (ST14d).・ ・
... If the value is smaller than S3 and larger than S4 (yes), ST14f is determined to be an insect such as a cockroach. If it is smaller than S4 (no), it is ignored.

[0036]

The present invention is embodied in the manner described above, and has the following effects. The difference image data is read out from the image storage means and the difference image data between the background image data and the current image data is calculated and output. The abnormal shape extraction means extracts the abnormal shape data from the difference image data based on a threshold value registered and stored in advance, and the lower end is grounded. The position detecting means detects the lower end portion of the abnormal shape, and also detects the position of the ground contact with the floor surface or the data of whether or not the ground contact is made. The actual shape extracting means computes and outputs distance correction data from the ground contact position data, and uses the distance correction data to extract, from the abnormal shape data, normalized actual shape data having actual dimensions independent of the distance from the camera. . By judging and outputting the types of invading objects such as humans, small animals, and insects based on the judgment conditions pre-stored in the judgment condition memory based on the actual shape data and the availability data, image data captured by the camera can be simplified. A small animal identification method and a small animal identification device capable of judging invaders, small animals, insects, and the like by image recognition means have been proposed.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a first embodiment of a small animal identification device according to the present invention and an overhead view of a monitoring area.

FIG. 2 is a conceptual diagram illustrating a configuration of a captured image display example on a monitor and a recognition target area according to the present invention.

FIG. 3 is a flowchart of a method for recognizing an intruding object according to the present invention;

FIG. 4 is a main part block diagram of a second embodiment of the small animal identification device according to the present invention.

FIG. 5 is a flowchart showing a method for determining an intruding object according to the present invention.

[Explanation of symbols]

 Reference Signs List 1 imaging means 1a camera 1b auxiliary optical system 2 A / D 3 image storage means 3a image memory 3b background image memory 4 area setting means 4a area setting section 4b area setting function memory 5 difference operation means 5a difference operation section 6 abnormality Shape extracting means 6a Abnormal area extracting part 6b Threshold memory 6c Approximate shape extracting part 6d Approximate shape memory 7 Lower end contact position detecting means 7a Lower end detecting part 7b Contact position detecting part 8 Distance data generating means 8a Difference calculating part 8b Distance converting part 9 Actual shape extraction unit 9a Distance correction operation unit 9b Distance operation unit 9c Area-distance table 9d Real shape extraction unit 10 Judgment condition memory 11 Intruding object judgment unit 11A Intrusion object judgment unit 13 System memory 14 Control unit

Continued on the front page F-term (reference) CA24 FA69 FA70 GA08 GA51 HA01 9A001 HH23 LL09

Claims (30)

[Claims] An object identification method for capturing an object to be monitored by an imaging means such as a television camera (hereinafter referred to as a camera), converting an image signal into digital image data, and identifying a person or a small animal by image recognition. , An image storage unit for temporarily storing the digital image data, a division setting unit for setting a recognition target division of the image data stored in the image storage unit, and registration and storage in the image storage unit Difference calculating means for calculating and outputting difference image data between the background image data and the current monitoring image data,
An abnormal shape extracting means for extracting an abnormal area (hereinafter referred to as an abnormal shape) from the difference image data based on a threshold value registered and stored in advance, and a bottom contact point for detecting a contact position of the lowermost portion of the abnormal shape to the floor surface Position detecting means, actual shape extracting means for calculating the shape and size of the abnormal shape based on the distance or area position from the camera and extracting the actual shape data, and the actual shape data and data on whether the lower end contact position can be detected. An intruding object determining means for determining the type of the intruding object;
A judgment condition memory for storing judgment conditions of the intruding object judgment means, wherein the background image data and the current monitoring image data in a state where there is no intruding object are temporarily stored in the image storage means, and the difference calculation is performed. The means reads the background image data and the current monitoring image data of the recognition target area set by the area setting means from the image storage means, and calculates and outputs difference image data between the two. The abnormal shape data is extracted from the image data based on a threshold value registered and stored in advance, and the lower end contact position detecting means detects the lower end portion of the abnormal shape and detects the contact position data on the floor surface or the data on whether or not the contact is possible. The actual shape extracting means calculates and outputs distance correction data from the contact position data and uses the distance correction data to calculate the abnormal shape data. The intruded object determining means extracts the normalized actual shape data of the actual dimensions independent of the distance from the camera, and determines the in advance object stored in the judgment condition memory from the actual shape data and the propriety data. Based on the conditions,
A small animal identification method based on image recognition, characterized in that the type of an intruding object such as a small animal or an insect is determined and output. 2. The small animal identification method by image recognition according to claim 1, wherein the area setting unit sets the recognition target areas at the same size and at equal intervals. 3. The small animal identification method by image recognition according to claim 1, wherein the area setting unit sets the area to be recognized in an arbitrary size along an equidistant line from a camera. . 4. The area setting unit according to claim 1, wherein the area setting unit sets the area to be recognized in an arbitrary size inversely proportional to a distance from the camera and along an equidistant line from the camera. Of small animals by image recognition. 5. The method according to claim 1, wherein the abnormal area of the abnormal shape extracting means is extracted using a threshold of an average value or a sum value of levels in the same area or the same area. The small animal identification method by the image recognition of the description. 6. The image recognition method according to claim 1, wherein the abnormal area of the abnormal shape extracting means is extracted using a threshold value of a hue average value in the same area or the same area. Small animal identification method. 7. The small animal identification by image recognition according to claim 1, wherein the abnormal area of the abnormal shape extracting means is extracted using a threshold value of the number of abnormal areas in the same area. Method. 8. The small animal identification method according to claim 1, wherein the abnormal section of the abnormal shape extracting means is extracted using a threshold value of the number of abnormal pixels in the same section. 9. The method according to claim 1, wherein the detection of the lower end contact position of the lower end contact position detecting means is a method of detecting a singularity or discontinuity of the continuity between the lower end of the abnormal shape and the floor surface. 3. The small animal identification method by image recognition according to 1. 10. The small animal according to claim 1, wherein the bottom contact position detection means of the bottom contact position detection means detects the presence or absence of a shadow at the lower end of the abnormal shape. Identification method. 11. The method according to claim 1, wherein the bottom contact position detection means of the bottom contact position detection means detects the presence or absence of a shadow at a lower end portion of the abnormal shape by additionally installing auxiliary lighting. 3. The small animal identification method by image recognition according to 1. 12. The distance correction data of the actual shape extracting means is transferred to an auxiliary optical system additionally installed in the imaging means.
2. The small animal identification method by image recognition according to claim 1, wherein the method is generated based on original imaging data. 13. The small animal identification method by image recognition according to claim 1, wherein the type determination of the intruding object by the intruding object determining means is performed based on an area value of the actual shape data. 14. The small animal identification method by image recognition according to claim 1, wherein the type determination of the intruding object by the intruding object determining means is performed based on a shape dimension value of the actual shape data. 15. The small animal identification method according to claim 1, wherein the type of the intruding object is determined by the intruding object determining means based on the availability data. 16. The method according to claim 1, wherein the type determination of the intruding object by the intruding object determining means is performed based on a combination condition of the abnormal shape data and the distance data or the area data. Small animal identification method by image recognition. 17. An intruding object discriminating apparatus which picks up an object to be monitored by an image pickup means such as a camera, converts an image signal into digital image data, and identifies a person or a small animal by image recognition. A / D (Analog Digital converter) that converts image signals captured by a camera into digital image data
ter), a background image memory for storing background image data in which no intruding object is present, an image memory for cyclically storing real-time image data, and an area setting unit for setting an area to be recognized for the image data. The difference between the background image data and the image data read by the zone setting unit, and the zone setting function memory storing the equal interval setting function required when the zone setting unit sets the recognition zone. A difference calculation unit that calculates and outputs difference image data, a threshold memory that stores threshold data of an average value for extracting an abnormal partition or an abnormal partition, and a difference memory based on the threshold data. An abnormal area extracting unit that extracts an abnormal area or abnormal area; an approximate shape extracting unit that extracts an approximate shape from the abnormal area or abnormal area data; A similar shape memory, a lower end detecting unit for detecting a lower end portion of the approximate shape, a grounding position detecting unit for detecting whether or not the lower end touches a floor surface, and a camera based on the grounding position data A distance calculation unit for calculating and outputting distance data from a distance, a distance correction calculation unit for performing shape correction calculation of the approximate shape based on the distance data, and an actual shape extraction unit for extracting real shape data from the shape correction calculation result And a judgment condition memory for registering and storing judgment conditions such as a shape and an area value for judging an intruding object type, and a person, a small animal, an insect, etc., based on the judgment condition, the actual shape data, grounding availability data, and the like. A small animal identification device comprising: an intruding object determining unit that performs determination output of an intruding object type. 18. An equidistant area setting function for setting the recognition target area to an arbitrary size and along an equidistant line from a camera is additionally stored in the area setting function memory. Item 18. The small animal identification device according to Item 17. 19. A distance weighting setting function for setting the recognition target area in an arbitrary size inversely proportional to the distance from the camera and for setting the recognition target area along an equidistant line from the camera in the area setting function memory. The small animal identification device according to claim 17, wherein: 20. The small animal identification device according to claim 17, wherein a shadow detecting means for detecting a shadow formed at a lower end portion of the intruding object is additionally provided in the contact position detecting section. 21. A shadow detecting means for detecting a shadow formed at a lower end portion of an intruding object, and an auxiliary lighting for facilitating shadowing at a lower end portion of the intruding object are additionally installed in the ground position detecting section. 18. The small animal identification device according to 17. 22. Binary imaging means having an auxiliary optical system in the camera, a difference calculation unit for calculating and outputting a difference time between the image data of the binary imaging, and distance data from the camera to the object to be imaged based on the difference time. 18. The small animal identification device according to claim 17, further comprising a distance data generation unit having a distance conversion unit for converting and outputting the distance data. 23. The binary image pickup means comprising an auxiliary optical system having an auxiliary lens and a half mirror or a prism, and a liquid crystal shutter for selectively inputting an optical image from the main lens and the auxiliary lens system. 23. The small animal identification device according to claim 22, wherein 24. The small animal identification device according to claim 17, wherein the threshold memory further stores a hue threshold value of an average hue value in the same area or the same area for extracting an abnormal shape. 25. The small animal identification apparatus according to claim 17, wherein the threshold memory further stores an abnormal section number threshold value for extracting an abnormal section based on the number of abnormal sections in the same section. 26. The small animal identification apparatus according to claim 17, wherein an abnormal pixel number threshold value for extracting an abnormal section based on the number of abnormal pixels in the same section is additionally stored in the threshold memory. 27. The threshold value memory according to claim 17, wherein an abnormal shape extraction is additionally stored with an infrared image threshold value of an average value of infrared or near-infrared images in the same area or the same area. Small animal identification device. 28. The apparatus according to claim 1, wherein said determination condition memory additionally stores a shape and size condition for determining the type of the intruding object by a combination of vertical and horizontal dimensions.
8. The small animal identification device according to 7. 29. The small animal identification apparatus according to claim 17, wherein an extraordinary flying object judging condition for judging by a combination of an aerial flying object and a limit area is additionally stored in the judgment condition memory. 30. The small animal according to claim 17, wherein said judgment condition memory additionally stores an imaging obstruction judgment condition for judging an abnormal image due to a camera occlusion, a partial occlusion of illumination, a camera failure, or the like. Identification device.