JP2012088861A - Intrusion object detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an intrusion object detection device capable of detecting an intrusion object moving in front of vegetation existing in a monitoring area and suppressing erroneous detection caused by swinging of the vegetation.SOLUTION: An intrusion object detection device 100 includes an imaging part 10 for acquiring a monitoring image of a monitoring area, a storage part 20 for storing a background image which is an image of the monitoring area in a state in which no intrusion object exists in the monitoring area and a vegetation area showing an area on an image of vegetation, change area extraction means 31 for extracting a change area where a luminance value has been changed by a difference between the monitoring image and the background image, and determination means 36 for calculating for the change area an intrusion object attribute value showing a degree of an intrusion object certainty which becomes lower in a case that the change area overlaps with the vegetation area than in a case that there is no overlap and determining whether the change area appears due to an intrusion object or not based on the intrusion object attribute value.

Description

  The present invention relates to an intruding object detection apparatus that detects an intruding object from an image obtained by imaging a monitoring area.

  Conventionally, an intruding object detection apparatus that detects an intruding object that has entered a monitoring area by analyzing an input image obtained by capturing the monitoring area has been widely used. For example, when detecting an intruder as an intruding object, such an intruding object detection device holds a background image obtained by photographing a monitoring area in a situation where there is no intruder to be detected, and between the input image and the background image. A difference in luminance value is obtained for each corresponding pixel. Then, the intruding object detection device uses a feature amount such as a size, an aspect ratio, and a luminance feature for a pixel collection region (hereinafter referred to as a change region) whose difference is equal to or greater than a predetermined threshold value. , And the presence / absence of an intruder is detected by determining whether or not the attribute value is within a predetermined range.

  However, when the outdoor object is monitored using the intruding object detection device, an object whose shape changes due to wind or the like, such as planting, may be included in the monitoring region. In such a case, when the planting is shaken by the wind, the above attribute value falls within a predetermined range for detecting an intruder, and the intruding object detection device displays a change area caused by planting in the input image. There was a problem that it was erroneously detected that it was caused by an intruder and a remote security center was notified that the intruder was detected (that is, a false alarm was issued).

  For example, a method of suppressing such erroneous detection by setting the predetermined range so as to be difficult to determine as an intruder can be considered. However, simply setting the predetermined range so narrow that it is difficult to determine an intruder may increase the risk of intruder detection omission. For this reason, it is very difficult to set a range suitable for detection of an intruder, and it may be difficult to appropriately suppress the erroneous detection.

  On the other hand, Patent Document 1 discloses a technique for preventing misinformation caused by planting. The monitoring image processing apparatus disclosed in Patent Document 1 extracts a region (temporal edge block) created by planting shake or the like. When the extracted area and the change area overlap by a predetermined amount or more, the monitoring image processing apparatus determines that the change area is not a detection target and does not issue a notification.

JP 2008-181321 A

  However, in the technique described in Patent Document 1, when an intruder moves before planting, a change area caused by the intruder overlaps the temporal edge block, and the intruder is determined not to be a detection target. In other words, there is a problem that there is a high possibility that intruders will miss detection.

  Therefore, an object of the present invention is to provide an intruding object detection device that can detect an intruding object that moves in front of planting existing in a monitoring area and that can suppress erroneous detection due to shaking of the planting.

  The present invention for solving such a problem provides an intruding object detection device for detecting an intruding object that has entered a monitoring area. In such an intruding object detection device, an imaging unit that acquires a monitoring image obtained by photographing a monitoring area, a background image that is an image obtained by photographing the monitoring area in a situation where no intruding object exists in the monitoring area, and planting are imaged. A storage unit that stores a planting region indicating a region on the image, a change region extraction unit that extracts a change region whose luminance value has changed according to a difference between the monitoring image and the background image, and the change region is planted with respect to the change region Calculate the intruding object attribute value that represents the degree of intruding object likelihood that makes the intruding object less likely than the intruding object when it overlaps the area, and whether or not the change area is due to the intruding object based on the intruding object attribute value Determining means for determining.

  Further, in the intruding object detection device according to the present invention, when the change area does not overlap with the planting area, the determining means calculates an intruding object attribute value based on the first feature amount that is a characteristic of the intruding object, When the change region overlaps the planting region, the intruding object attribute is based on the second feature amount that is a feature other than the first feature amount that is significantly different between the intruding object and the planting in addition to the first feature amount. It is preferable to calculate the value.

  Further, in the intruding object detection device according to the present invention, each time a monitoring image is acquired by the imaging unit, the latest change area extracted from the latest monitoring image and the monitoring image acquired immediately before are extracted. A tracking processing means for associating an object in the latest change area with a change area immediately before the same object as the object in the latest change area, and the judging means for planting the latest change area If the change area in the monitoring image immediately before or a predetermined number of times before being associated with the latest change area does not overlap with the planting area, the intruding object attribute value is obtained by the first process. When the latest change area does not overlap with the planting area, and the change area in the monitoring image immediately before or a predetermined number of times corresponding to the latest change area overlaps with the planting area First process It is preferable to calculate the intruding object attribute value by the second process of a value not like intruder intruding objects attribute values.

  Furthermore, in the intruding object detection device according to the present invention, it is preferable that the determination means sets the intruding object attribute value to a value that is less likely to be an intruding object as the ratio of the area that overlaps the planting area in the change area increases.

  The intruding object detection device according to the present invention can detect an intruding object that moves in front of planting existing in the monitoring area, and can suppress an erroneous detection due to shaking of the planting.

It is a schematic block diagram of the intruding object detection apparatus which is one Embodiment of this invention. It is a figure showing the relationship between an input image, a background image, a change area | region, and a planting area | region. (A)-(d) is a figure showing the relationship between the motion of a change area | region, and a planting area | region. (A) is a figure which shows an example of the function F1 which normalizes the magnitude | size T1 of an intruding object, (b) is a figure which shows an example of the function F2 which normalizes aspect ratio T2, (c) is an edge It is a figure which shows an example of the function F3 which normalizes the change rate T3, (d) is a figure which shows an example of the function F4 which normalizes the movement amount T4. (A) is a figure which shows an example of function F1 'which normalizes size T1 of an intruding object, (b) is a figure which shows an example of function F2' which normalizes aspect ratio T2, (c) Is a diagram showing an example of a function F3 ′ for normalizing the edge change rate T3, (d) is a diagram showing an example of a function F4 ′ for normalizing the movement amount T4, and (e) is a straight traveling degree T5. FIG. 6D is a diagram illustrating an example of a function F6 that normalizes the shape stability T6. It is a flowchart which shows the intruding object detection operation | movement by an image process part.

Hereinafter, an intruding object detection apparatus according to an embodiment of the present invention will be described with reference to the drawings.
This intruding object detection device detects an intruding object that has entered the monitoring area from a monitoring image obtained by photographing the monitoring area. In particular, the intruding object detection device calculates the degree of likelihood of an intruding object using the feature amount of the intruding object for an area where the luminance value of the monitoring image has changed, and based on the calculated degree of intruding object likelihood, Determine presence or absence. And when the intruding object detection device overlaps the region where the brightness value has changed with the region where the planting exists, the intruding object detection device determines that the change is likely due to the shaking of the planting, This makes it more difficult to determine that the object is an intruding object than when it does not overlap with the area to be operated. As a result, the intruding object detection device detects an intruding object that moves before planting, and suppresses erroneous detection of the shaking of the planting as an intruding object.

  Note that the intruding object detected by the intruding object detection device may be anything such as a person or a car, but in this embodiment, a case where a person is detected as an intruding object will be described as an example.

  FIG. 1 is a diagram showing a schematic configuration of an intruding object detection device 100 according to an embodiment of the present invention. As illustrated in FIG. 1, the intruding object detection device 100 includes an imaging unit 10, a storage unit 20, an image processing unit 30, and an output unit 40.

The imaging unit 10 includes a two-dimensional detector composed of a photoelectric converter having sensitivity to visible light or near-infrared light, such as a CCD element or a C-MOS element, and an image of a monitoring region on the two-dimensional detector. An image forming optical system that forms an image and an electric circuit that amplifies an electric signal output from the two-dimensional detector and performs analog / digital (A / D) conversion, and the like.
The imaging unit 10 captures images at regular time intervals (for example, 1/5 second). Then, the imaging unit 10 generates a monitoring image obtained by capturing the monitoring area as digital image data having, for example, horizontal 320 pixels × vertical 240 pixels and each pixel having a luminance value of 0 to 255. Then, the imaging unit 10 transmits the generated digital image data (hereinafter referred to as an input image) to the image processing unit 30.

  The storage unit 20 includes a nonvolatile semiconductor memory such as a flash memory (registered trademark), a volatile semiconductor memory, a magnetic disk (HDD), an optical disk drive such as a CD-ROM and a DVD-RAM, and a recording medium thereof. And the memory | storage part 20 memorize | stores the planting area | region which shows the background image which is the image which image | photographed the monitoring area | region when the intruding object does not exist in the monitoring area | region, and the area | region on the image where planting was imaged. Further, the storage unit 20 stores a program executed on the image processing unit 30, various setting parameters used by the program, a calculation value or an image obtained as a result of image processing or an intermediate result, and the like. The storage unit 20 stores the calculated values or images according to a control signal from the image processing unit 30 or outputs various stored information to the image processing unit 30.

  The image processing unit 30 includes a built-in microprocessor unit and its peripheral circuits, and controls the entire intruding object detection apparatus 100. The image processing unit 30 detects an intruding object that has entered the monitoring area from the input image received from the imaging unit 10. For this purpose, the image processing unit 30 includes a change area extraction unit 31, a tracking processing unit 32, a feature amount calculation unit 33, an overlap determination unit 34, a departure determination unit 35, a determination unit 36, and a background image update unit. 37 and planting area update means 38. Each of these units included in the image processing unit 30 is implemented as, for example, a functional module of a program that operates on the microprocessor unit.

  The change area extraction unit 31 extracts an area where an object moving in the monitoring area (hereinafter referred to as a moving object) may be reflected in the latest input image. Therefore, the change area extraction unit 31 obtains a luminance difference between corresponding pixels between the latest input image and the background image, and calculates a difference image in which the pixel value of each pixel is represented by the absolute value of the luminance difference. create. The change area extraction unit 31 binarizes the difference image so that each pixel of the difference image has a pixel whose pixel value is equal to or greater than a predetermined threshold Th1 as a variation pixel and a pixel whose pixel value is less than the threshold Th1 as a background pixel. A binarized image is created. For example, the threshold value Th1 can be an average value of the pixel values of the difference image.

  Further, the change area extraction unit 31 extracts, as a change area, a region in which the variable pixels are connected to each other by using a known labeling method or the like for the created difference binary image (hereinafter referred to as a difference area). The area of the input image and the area of the background image corresponding to the change area of the binarized image are also referred to as change areas). Note that the number of change areas extracted simultaneously is not limited to one, and a plurality of change areas may be extracted. In addition, the labeling process may be performed by 4-neighbor concatenation in which a region in which variable pixels are connected in the vertical and horizontal directions is extracted as one change region, or a region in which the vertical, horizontal, and diagonal directions are connected to one region. You may carry out by the 8-neighbor connection extracted as a change area.

  In addition, the change area extraction unit 31 determines that the adjacent change area is one change area when the rectangular width and height when the adjacent change areas are integrated are within a range corresponding to the width and height of the actual person. It may be integrated into. Alternatively, the change area extraction unit 31 determines that the adjacent change areas are one when the distances in the horizontal direction and the vertical direction on the image between the adjacent change areas are determined to be one object. You may integrate into a change area.

  The change area extraction unit 31 attaches a label number to each extracted change area, and creates a label image that has the same size as the difference binary image and uses the pixel value as the label number of the corresponding change area. Then, the change area extraction unit 31 stores the created label image in the storage unit 20.

The tracking processing unit 32 assumes that the same moving object is captured by using a known tracking process for each of the changed regions extracted from the input images of a plurality of frames continuously acquired by the imaging unit 10. The associated change area is detected.
The tracking processing unit 32 acquires information about the change area from the label image stored in the storage unit 20. Then, the tracking processing unit 32 includes a change area extracted from the latest input image (hereinafter referred to as the current frame) and a moving object identical to the moving object shown in the change area. The change area extracted from the input image (hereinafter referred to as the previous frame) is associated. The tracking processing unit 32 displays the change area of the current frame associated with the change area of the previous frame, the change area of the previous frame associated with the change area of the previous frame, and the past frame in which the change area of the previous frame is associated. Are stored in the storage unit 20 in association with the change area.

  The change area that could not be associated with the change area of the previous frame is assumed to be a new moving object, and is newly subject to tracking processing. In addition, the change area of the past frame that could not be associated with the change area of the current frame is excluded from the target of the subsequent tracking process.

  For example, for each change area in which the same moving object is captured, the tracking processing unit 32 calculates the change area of the previous frame and the center of gravity position of the change area of the previous frame corresponding to the change area. By calculating the difference, the traveling direction and moving speed of the moving object on the image are determined. Then, the tracking processing means 32 estimates the position where the moving object is reflected in the current frame from the traveling direction and the moving speed, and the estimated position is the change area closest to the center of gravity position of the change area of interest in the current frame. A change area in which the same moving object as the moving object shown in the target change area is shown. Alternatively, the tracking processing unit 32 calculates an average value of the size of the change area up to the previous frame for each change area that is the target of the tracking process and includes the same moving object, and the calculated average value is the current frame. The change area closest to the size of the target change area may be the change area in which the same moving object as the moving object shown in the target change area is shown. Alternatively, the tracking processing unit 32 obtains the correlation between the change area of the previous frame and the target change area of the current frame for each change area in which the same moving object is captured, and the highest correlation is obtained. The change area may be a change area in which the same moving object as the moving object shown in the target change area is shown.

  The feature amount calculation means 33 acquires information relating to the association between the label image stored in the storage unit 20 and each change region, and calculates a feature amount representing the feature of the intruding object in the change region for each change region. Then, the feature amount calculation unit 33 stores the calculated feature amount in the storage unit 20.

FIG. 2 is a diagram illustrating an example of a relationship among an input image, a background image, a change area, and a planting area. In FIG. 2, an image 200 represents an input image. In the input image 200, planting 201, 202, 203 and people 204, 205 are captured. An image 210 shows a background image. The background image 210 is the same as the plants 201, 202, and 203, and the plants 211, 212, and 213 captured in the past are shown.
An image 220 shows a change area extracted from the input image 200 and the background image 210. In the image 220, change areas 224 and 225 due to movement of the people 204 and 205 are extracted, and change areas 222 and 223 due to shaking of the planting 202 and 203 are also extracted.
The image 230 represents a planting area stored in the storage unit 20. In the image 230, planting regions 231, 232, and 233 are shown as planting regions corresponding to the planting 201, 202, and 203, respectively.

As shown in an image 220, an intruding object detection apparatus that detects an intruding object by analyzing a monitoring image may extract not only a person who is a detection target but also planting swayed by a wind or the like as a change region. In the images 220 and 230, the change region 223 due to the shaking of the planting 203 overlaps the planting region 233. Therefore, even if it is the prior art which excludes the change area | region which overlaps a planting area | region as a detection target, it can prevent misdetecting the planting 203 with an intruder. However, in that case, since the change area 224 by the person 204 moving in front of the planting 201 overlaps with the planting area 231, the person 204 cannot be detected. Moreover, since the change area 222 due to the shaking of the planting 202 does not overlap with the planting area 232, the planting 202 may be erroneously detected as an intruder even if the change area overlapping the planting area is excluded from the detection target.
Therefore, the feature amount calculation unit 33 according to the present embodiment calculates the feature amount of the intruding object in order to distinguish each change region extracted by the change region extraction unit 31 from a person and a planting.

The feature amount calculating means 33 calculates the size T1, the aspect ratio T2, the edge change rate T3, and the movement amount T4 as the feature amounts of the intruding object. The feature amounts T1 to T4 are feature amounts for determining whether or not the change area seems to be an intruding object.
Further, the feature amount calculating means 33 calculates the straightness of movement T5 and the shape stability T6 as the feature amounts of the intruding object. The feature amounts T5 to T6 are feature amounts for distinguishing an intruding object from planting.
Further, the feature amount calculation means 33 calculates the planting region overlap rate T7 and the planting region overlap count T8 as the feature amount of the intruding object. The feature amounts T7 to T8 are feature amounts based on the positional relationship between the change region and the planting region.

  The size T1 is the size of the moving object. In general, on the image, a moving object that is present at a position close to the imaging unit 10 (photographed downward on the image) appears large, and a moving object that is located at a position far from the imaging unit 10 (photographed upward on the image) is small. It is reflected. Therefore, for example, the intruding object detection apparatus 100 captures a moving object at a predetermined distance from the imaging unit 10 in advance, and determines the relationship between the number of pixels on the image of the captured moving object and the actual size of the moving object. It is obtained for each lower end position (horizontal line position) on the image. Then, the feature amount calculating unit 33 calculates the number of pixels included in the change area, and converts the calculated number of pixels into an actual size according to the lower end position (horizontal line position) of the change area as a size T1. To do.

  The aspect ratio T2 is the aspect ratio of the moving object. The feature amount calculating unit 33 calculates the aspect ratio of the moving object based on the ratio of the width and height of the circumscribed rectangle of the change area. The aspect ratio T2 represents the shape feature of the moving object.

The edge change rate T3 is the change rate of the edge compared to the background image in the change area of the input image, and is an index for determining whether the change in the change area is due to an intruding object, or due to light or shadow. Become. If the variation in the change region is due to light or shadow, the texture in the change region does not change, so the edge distribution does not change roughly. On the other hand, if the variation in the change area is due to an intruding object, the background image in the change area represents the background texture of the monitoring area, and the input image represents the texture of the intruding object. Distribution changes greatly. Therefore, the feature amount calculation unit 33 performs an inter-pixel calculation using an edge detection filter such as a sobel filter or a prewitt filter on the input image and the background image. Then, the feature amount calculation unit 33 sets a pixel having an edge strength, which is an absolute value of a difference value obtained as a result of the calculation between neighboring pixels, to be larger than a predetermined threshold value as an edge pixel. In addition, this threshold value can be made into the average value of the absolute difference value obtained about each pixel of the image which performed the calculation between adjacent pixels, for example. Then, the feature amount calculating unit 33 calculates the edge change rate T3 by the following equation, for example.
T3 = (number of edge pixels among corresponding pixels in the change area of the input image and the background image) / (total number of pixels in the change area) (1)
That is, the numerator of equation (1) is not an edge pixel in the background image, but the number of pixels that are edge pixels in the corresponding pixel of the input image, and an edge pixel in the corresponding pixel of the input image. It is the sum of the number of pixels that are not. As a result, the feature amount calculating unit 33 performs not only when the edge that did not exist in the background image appears in the input image due to the movement of the intruding object, but also when the edge that existed in the background image disappears in the input image. The rate of change T3 can be increased.

  The movement amount T4 is the distance moved from the place where the change area in which the same moving object appears is first appearing. When the change area is caused by an intruding object, the intruding object that has entered the monitoring area is likely to move toward the purpose. On the other hand, if the change area is caused by light or shadow, the change area is generated by light and shadow swaying on the spot, so the change may appear to move over a long distance. Is low. Therefore, the feature amount calculation means 33 obtains the movement amount from the distance between the center of gravity of the change region that first appears among the change regions in which the same moving object is reflected and the center of gravity of the change region of the current frame, and enters the intrusion. Used as a feature value for determining an object.

The figure for demonstrating the relationship between the motion of a change area | region and a planting area | region to Fig.3 (a)-(d) is shown.
In FIG. 3A, an image 300 represents the relationship between the movement of the change area 224 and the planting area 231 shown in FIG. The change area 301 is a change area by the person 204 extracted two frames before the input image 200, and the change area 302 is a change area by the person 204 extracted one frame before the input image 200.
In FIG. 3B, an image 310 represents the relationship between the movement of the change area 222 and the planting area 232 shown in FIG. The change area 311 is a change area by the planting 202 extracted two frames before the input image 200, and the change area 312 is a change area by the planting 202 extracted one frame before the input image 200.
In FIG. 3C, an image 320 represents the relationship between the movement of the change area 223 shown in FIG. 2 and the planting area 233. The change area 321 is a change area by the planting 203 extracted two frames before the input image 200, and the change area 322 is a change area by the planting 203 extracted one frame before the input image 200.
In FIG. 3D, an image 330 represents the movement of the change area 225 shown in FIG. The change area 331 is a change area by the person 205 extracted two frames before the input image 200, and the change area 332 is a change area by the person 205 extracted one frame before the input image 200.

  As shown in FIGS. 3A and 3D, when the change area is caused by a person, the person who has entered the monitoring area is likely to move linearly toward the purpose, and the moving direction of the change area (The direction of the arrow) tends not to change. On the other hand, as shown in FIGS. 3B and 3C, when the change area is due to planting, the change area is generated by shaking of the planting and the moving direction tends to change greatly. is there. Therefore, the feature amount calculating means 33 calculates a straight traveling degree T5 that is an index for determining whether or not the change area is moving linearly as a feature amount that distinguishes an intruding object from planting.

For example, the feature amount calculating unit 33 uses the information of the change areas of the last three frames in which the same moving object is shown, and calculates the straight movement degree T5 by the following formula to distinguish between the intruding object and the planting. It is used as a feature amount.
T5 = (X1 x X2 + Y1 x Y2) / (D1 x D2) (2)
However, D1 is the position of the center of gravity of the change region of the current frame (X _t, Y _t) (provided that, X _t, Y _t is the horizontal coordinate of the pixels in each of the image data represents a vertical coordinate. Hereinafter the same) as before This is the distance between the center of gravity position (X _t-1 , Y _t-1 ) of the change area of the frame, and D2 is the position of the center of gravity (X _t-1 , Y _t-1 ) of the change area of the previous frame. This is the distance between the barycentric positions (X _t-2 , Y _t-2 ) of the change area of the frame. X1 is the horizontal component of the vector from the centroid position of the change area of the previous frame to the centroid position of the change area of the current frame, and X2 is the centroid position of the change area of the previous frame from the centroid position of the change area of the previous frame. The horizontal component of the vector to the position, Y1 is the vertical component of the vector from the centroid position of the change area of the previous frame to the centroid position of the change area of the current frame, and Y2 is the centroid of the change area of the previous frame This is a vertical component of the vector from the position to the center of gravity position of the change area of the previous frame, and is expressed by the following equations.
_{X1 = X t -X t-1} (3)
X2 = X _t-1 -X _t-2 (4)
Y1 ＝ Y _t −Y _t-1 (5)
Y2 = Y _t-1 -Y _t-2 (6)
That is, the numerator of equation (2) is a vector from the center of gravity of the change area of the previous frame to the center of gravity of the change area of the previous frame and the center of gravity of the change area of the previous frame to the center of gravity of the change area of the current frame. The travel straightness T5 is 1 when the directions of the two vectors are the same (0 °), becomes smaller as they move away, and becomes 0 when the directions are opposite (180 °). That is, the straight travel degree T5 represents the same degree of movement direction of the center of gravity position of each corresponding change area.

  Even when the moving object is planted, the change area in which the moving object is captured moves in the same direction for three frames, and the straight movement degree T5 may increase. Therefore, the feature amount calculation unit 33 may use the average value of the values calculated by the expression (2) for a plurality of frames acquired within a predetermined period as the straight travel degree T5. In this case, the predetermined period may be optimized as appropriate according to the environment in which the intruding object detection device 100 is installed, the purpose of detecting a person, and the like.

  In addition, as shown in FIGS. 3B and 3C, when the change region is due to planting, the change region is changed by the shape of the branches or leaves reflected in the input image due to the planting shaking. Due to the occurrence, the change in shape tends to be larger than the change region by the person shown in FIGS. 3 (a) and 3 (d). Therefore, the feature amount calculation means 33 calculates a shape stability T6, which is an index representing the degree of stability of the shape of the change area, as a feature amount that distinguishes an intruding object from planting.

The feature amount calculation means 33 calculates the shape stability T6 by the following formula using information on the change area of the latest three frames in which the same moving object is shown, for example, and distinguishes the intruding object from planting It is used as a feature amount.
T6 = (Rectangle width change rate of the current frame change area) × (Rectangle height change ratio of the current frame change area) + (Rectangle width change ratio of the previous frame change) × (Previous frame change) (Rectangle height change rate) (7)
However, the change rate of the rectangular width of the change area of the current frame is the ratio of the smaller value to the larger value of the rectangular width of the change area of the current frame and the rectangle width of the change area of the previous frame. Similarly, the change rate of the rectangular height of the change area of the current frame is the ratio of the smaller value to the larger value of the rectangular height of the change area of the current frame and the rectangular height of the change area of the previous frame. Similarly, the change rate of the rectangular width of the change area of the previous frame is the ratio of the smaller value to the larger value of the rectangular width of the change area of the previous frame and the rectangle width of the change area of the previous frame. Similarly, the change rate of the rectangular height of the change area of the previous frame is the ratio of the smaller value to the larger value of the rectangular height of the change area of the previous frame and the rectangular height of the change area of the previous frame. .
That is, the shape stability T6 becomes smaller as the change in the rectangular shape of the change region is larger, and becomes larger as the change in the rectangular shape of the change region is smaller.

  Even when the moving object is planted, the shape of the rectangle of the change area in which the moving object is captured does not change for three frames, and the shape stability T6 may increase. Therefore, the feature amount calculating unit 33 may use the average value of the values calculated by the equation (7) for a plurality of frames acquired within a predetermined period as the shape stability T6. In this case, the predetermined period may be optimized as appropriate according to the environment in which the intruding object detection device 100 is installed, the purpose of detecting a person, and the like.

The overlapping rate T7 of the planting region represents the degree of overlap between the change region and the planting region that is stored in the storage unit 20 and indicates the region on the image where planting is captured. There is a high possibility that the change area generated at the position overlapping the planting area is due to the planting. Therefore, the feature quantity calculating means 33 calculates the overlapping ratio T7 of the planting area by using the planting area, for example, by the following equation.
T7 = (number of pixels overlapping the planting area in the change area) / (total number of pixels in the change area) (8)
That is, the overlapping ratio T7 of the planting area becomes a larger value as the ratio of the area overlapping the planting area in the change area is larger, and becomes a smaller value as the ratio is smaller.

The planting area overlap count T8 represents the number of times that the change area overlaps the planting area in the previous frame. In order to suppress false detection due to shaking of the planting, set the planting area large so that it covers not only the area where the planting is imaged but also all the areas where the planting may shake It is desirable. However, if the planting area is set to be large, the area that is difficult to determine as an intruder becomes large, so in general, the area where planting fluctuation is likely to be observed, that is, on the image where the planting is imaged. Only the area is set as the planting area. Therefore, when the planting is greatly shaken by the influence of the wind or the like, or when the plant grows and becomes large, a change region due to the shaking of the planting may be extracted beyond the planting region. Therefore, even if the change area that shows the same moving object does not overlap the planting area in the current frame, if the overlapping area overlaps the planting area in the previous frame, the moving object may be part of the planting. Is expensive.
Therefore, the feature quantity calculating means 33 assigns a counter to each change area associated with the same moving object as a reflection, and the change area of the current frame where the same moving object is reflected is the planting area. The counter is incremented when it overlaps with, and is initialized to 0 when the change area of the current frame does not overlap with the planting area. Then, the feature amount calculating means 33 holds the value of the counter before updating, that is, the number of times that the state in which the change region overlaps the planting region in the previous frame is continued as the planting region overlap count T8.

  Note that the feature amount calculation unit 33 may increment the counter when the change region overlaps with the planting region even a little. In the present embodiment, the feature amount calculating means 33 increments the counter when the number of pixels where the change area and the planting area overlap is 1 pixel or more, or when the overlapping ratio of the change area and the planting area is greater than 0%. To do.

  The overlap determination unit 34 determines, for each change region, whether or not it overlaps the planting region using the planting region overlap rate T7 calculated by the feature amount calculation unit 33. In the present embodiment, the overlap determination unit 34 determines that the change region and the planting region overlap when the overlapping rate T7 of the planting region is equal to or greater than a predetermined threshold. In addition, when the change area | region has overlapped with the planting area even if it is a little, it is preferable to determine with the change area | region and planting area | region overlapping. Therefore, the predetermined threshold is preferably 0.01, for example. Then, the overlap determination unit 34 notifies the separation determination unit 35 and the determination unit 36 of the determination result that determines whether or not the change region and the planting region overlap.

The leaving determination means 35 leaves the planting area using the overlapping number T8 of the planting area calculated by the feature amount calculating means 33 for the change area determined not to overlap the planting area by the overlap determining means 34. It is determined whether or not In the present embodiment, the leaving determination means 35 determines that the changing area has left the planting area when the overlapping number T8 of the planting area for the changing area is 1 or more, and overlaps the planting area. When the number of times T8 is 0, it is determined that the change area is not a departure from the planting area.
That is, for example, when the change area overlaps the planting area in the previous frame, the leaving determination unit 35 determines that the change area has left the planting area, and the change area becomes the planting area in the previous frame. When it does not overlap, it determines with the change area | region not having left | separated from the planting area | region.
Alternatively, when the change area overlaps the planting area in the monitoring image up to a predetermined number before, the leave determination unit 35 determines that the change area has left the planting area, and the change area is a predetermined number of times ago. If the monitoring image does not overlap with the planting region, the change region may be determined not to have left the planting region. In that case, when the change area does not overlap with the planting area, the feature amount calculating means 33 does not initialize the counter for calculating the planting area overlap count T8 to 0, but the counter is 0 or more. What is necessary is just to decrement in the range which takes a value. Thereby, even if it is the change area | region which has left | separated from the planting area for a long time, the leaving | separation determination means 35 determines with the possibility of planting being high when it overlaps with the planting area for a long time before leaving. Thus, it can be difficult to detect an intruding object.
Alternatively, when the change area has overlapped with the planting area even once in the past, the leaving determination unit 35 determines that the change area has left the planting area, and the change area has been once planted in the past. If the area does not overlap, the change area may be determined not to have left the planting area. In that case, the feature amount calculation means 33 does not initialize or decrement the counter for calculating the number of times T8 of planting areas overlaps, even if the change area does not overlap with the planting area, and changes it. Do not do that. Thereby, the leaving determination means 35 determines that there is a high possibility of planting even if it is a change region that has left the planting region for a long time and has overlapped with the planting region even once in the past. This makes it difficult to detect an intruding object.
And the leaving | separation determination means 35 notifies the determination means 36 of the determination result which determined whether the change area | region was what left | separated from the planting area | region.

The determination unit 36 is a unit that determines whether or not the change area is caused by an intruding object based on the feature amount stored in the storage unit 20.
The change area 225 that does not overlap the planting area as shown in FIG. 3D is highly likely to be caused by an intruding object. On the other hand, there is a high possibility that the change area 223 overlapping the planting area as shown in FIG. However, as shown to Fig.3 (a), even if the change area | region which overlaps with the planting area | region, the change area | region may be a thing by the person who moves before planting. In addition, as shown in FIG. 3B, even in a change area that does not overlap with the planting area, if it overlaps with the planting area in the immediately preceding frame, there is a high possibility that it is due to planting.
Therefore, the determination unit 36 determines that the change region is determined to be a planting region that has been determined not to overlap the planting region by the overlap determination unit 34 and that has not been separated from the planting region by the separation determination unit 35. It is determined that there is a high possibility that it is not related to trees, and the presence or absence of an intruding object is determined by normal determination processing. On the other hand, the determination means 36 has the change area determined to be overlapped with the planting area by the overlap determination means 34 or the change area determined to be detached from the planting area by the separation determination means 35. Although the change area may be due to an intruding object, it is determined that the change area is likely due to planting. In this case, the presence / absence of an intruding object is determined by a special determination process that makes it difficult to determine that the change area is due to an intruding object than in the normal determination process.

The normal determination process will be described below.
In the normal determination process, the determination unit 36 uses the size T1 of the intruding object, the aspect ratio T2, the edge change rate T3, and the movement amount T4, which are feature amounts for determining whether or not the change area is a person. Then, an intruding object attribute value representing the degree of intrusion object likelihood of the change area is calculated. Then, the determination unit 36 determines whether the calculated intruding object attribute value is equal to or greater than a predetermined threshold Th3. When the intruding object attribute value is equal to or greater than the threshold Th3, the determination unit 36 determines that the change area is due to the intruding object, and when it is less than the threshold Th3, determines that the change area is not due to the intruding object. The threshold value Th3 is appropriately optimized according to the environment where the intruding object detection device 100 is installed, the purpose of detecting a person, and the like.

The intruding object attribute value is calculated, for example, by multiplication of values obtained by normalizing the intruding object size T1, aspect ratio T2, edge change rate T3, and movement amount T4.
FIGS. 4A to 4D show examples of functions F1, F2, F3, and F4 for normalizing the size T1, aspect ratio T2, edge change rate T3, and movement amount T4 of the intruding object, respectively.

  In FIG. 4A, a graph 401 represents the relationship between the input value and the output value of the function F1, the horizontal axis represents the size T1 of the intruding object, and the vertical axis represents the output value of the function F1. In this example, the output value of the function F1 increases linearly from 0 to 1 when 0 ≦ T1 <a, and becomes 1 when a ≦ T1. The threshold value a is determined based on a person's height and physique, and is determined in advance by experiment or experience so that a person can be distinguished from small animals such as cats, mice, and insects.

  In FIG. 4B, a graph 402 represents the relationship between the input value and output value of the function F2, the horizontal axis represents the aspect ratio T2, and the vertical axis represents the output value of the function F2. In this example, the output value of the function F2 increases linearly from 0 to 1 when 0 ≦ T2 <b1, becomes 1 when b1 ≦ T2 <b2, and linearly from 1 to 0 when b2 ≦ T2 <b3. Decreases and becomes 0 when b3 ≦ T2. Like the threshold value a, the threshold values b1 to b3 are determined based on human height and physique, and are determined in advance by experiment or experience so that a person can be distinguished from small animals such as cats, mice, and insects. It is done.

  In FIG. 4C, a graph 403 represents the relationship between the input value and the output value of the function F3, the horizontal axis represents the edge change rate T3, and the vertical axis represents the output value of the function F3. In this example, the output value of the function F3 increases linearly from 0 to 1 when 0 ≦ T3 <c, and becomes 1 when c ≦ T3. The threshold value c is determined in advance experimentally or empirically so that the variation caused by the intruding object can be distinguished from the variation caused by light or shadow.

  In FIG. 4D, a graph 404 represents the relationship between the input value and the output value of the function F4, the horizontal axis represents the movement amount T4, and the vertical axis represents the output value of the function F4. In this example, the output value of the function F4 increases linearly from 0 to 1 when 0 ≦ T1 <d, and becomes 1 when d ≦ T4. The threshold value d is determined in advance by experiment or experience so that the amount of movement of the intruding object can be distinguished from the amount of movement due to light or shadow.

As shown in FIGS. 4A to 4D, the output values of the functions F1, F2, F3, and F4 are such that the output value becomes closer to 1 as the change region is more human, and the output value becomes smaller as the change region is less human. This function is close to 0. At this time, the intruding object attribute value is calculated by, for example, the following equation using the functions F1, F2, F3, and F4.
(Intruding object attribute value) = F1 x F2 x F3 x F4 (9)
That is, the intruding object attribute value becomes closer to 1 as the change area is more human, and closer to 0 as the change area is less human.

The intruding object attribute value may be calculated by an average value or a weighted sum of F1, F2, F3, and F4. When the average value of F1, F2, F3, and F4 is used, the intruding object attribute value is calculated by the following equation.
(Intruding object attribute value) = (F1 + F2 + F3 + F4) / 4 (10)
When the weighted sum of F1, F2, F3, and F4 is used, the intruding object attribute value is calculated by the following equation.
(Intruding object attribute value) = (F1 × α1 + F2 × α2 + F3 × α3 + F4 × α4) (11)
However, α1 to α4 are weighting factors determined depending on which one of F1 to F4 is important, and the sum (Σαi (i = 1 to 4)) is 1. For example, when equalizing the weights of F1 to F4, all the weighting coefficients are set to the same value, and when some of the feature quantities of F1 to F4 are not used, the corresponding weighting coefficient is set to 0. How to set the specific value of each weighting factor is appropriately optimized according to the environment in which the intruding object detection device 100 is installed, the purpose of detecting a person, and the like.

Below, the special determination process applied about the change area | region which overlaps with the planting area | region, or the change area | region which has left | separated from the planting area | region is demonstrated.
In the special determination process, the determination means 36 is a feature amount for determining whether or not the change area is human, in addition to the intruding object size T1, aspect ratio T2, edge change rate T3, and movement amount T4. The intruding object attribute value is calculated using the straightness of movement T5 and the shape stability T6, which are feature quantities for distinguishing between a person and planting. Then, the determination unit 36 determines whether or not the calculated intruding object attribute value is greater than or equal to a threshold value Th3. When the intruding object attribute value is equal to or greater than the threshold Th3, the determination unit 36 determines that the change area is due to the intruding object, and when it is less than the threshold Th3, determines that the change area is not due to the intruding object.

The intruding object attribute value is calculated by, for example, multiplying values obtained by normalizing the intruding object size T1, aspect ratio T2, edge change rate T3, movement amount T4, movement straightness T5, and shape stability T6.
FIGS. 5A to 5F show functions F1 ′ and F2 for normalizing the size T1, the aspect ratio T2, the edge change rate T3, the movement amount T4, the straight movement degree T5, and the shape stability T6, respectively, of the intruding object. Examples of ', F3', F4 ', F5 and F6 are shown.

  In FIG. 5A, a graph 501 represents the relationship between the input value and output value of the function F1 ′, the horizontal axis represents the size T1 of the intruding object, and the vertical axis represents the output value of the function F1 ′. In this example, the output value of the function F1 ′ increases linearly from 0 to 1 when 0 ≦ T1 <a ′, and becomes 1 when a ′ ≦ T1. However, when the relationship of a <a ′ is established with respect to a shown in FIG. 4A and the input value T1 is the same, the output value of the function F1 ′ is less than or equal to the output value of the function F1. . That is, normalizing using the function F1 ′ is narrower than the normalizing using the function F1, and the range allowing the size of the intruding object is narrower, and it is difficult to determine the intruding object.

  In FIG. 5B, a graph 502 represents the relationship between the input value and the output value of the function F2 ′, the horizontal axis represents the aspect ratio T2, and the vertical axis represents the output value of the function F2 ′. In this example, the output value of the function F2 ′ is 0 when 0 ≦ T2 <b4, increases linearly from 0 to 1 when b4 ≦ T2 <b1 ′, and becomes 1 when b1 ′ ≦ T2 <b2 ′. , Linearly decreases from 1 to 0 when b2 ′ ≦ T2 <b3 ′, and becomes 0 when b3 ′ ≦ T2. However, the relations 0 <b4, b1 <b1 ′, b2> b2 ′, b3> b3 ′ are established with respect to b1, b2, and b3 shown in FIG. 4B, and the input value T2 is the same. In this case, the output value of the function F2 ′ is less than or equal to the output value of the function F2. That is, normalization using the function F2 ′ is narrower than the normalization using the function F2, and the range allowing the aspect ratio that is likely to be an intruding object is narrow, and it is difficult to determine the intruding object.

  In FIG. 5C, a graph 503 represents the relationship between the input value and the output value of the function F3 ′, the horizontal axis represents the edge change rate T3, and the vertical axis represents the output value of the function F3 ′. In this example, the output value of the function F3 ′ increases linearly from 0 to 1 when 0 ≦ T3 <c ′, and becomes 1 when c ′ ≦ T3. However, when the relationship c <c ′ is established with respect to c shown in FIG. 4C and the input value T3 is the same, the output value of the function F3 ′ is less than or equal to the output value of the function F3. . In other words, normalization using the function F3 ′ is narrower than the normalization using the function F3, and the range allowing the edge ratio that is likely to be an intruding object is narrower, and it is difficult to determine that the object is an intruding object.

  In FIG. 5D, a graph 504 represents the relationship between the input value and the output value of the function F4 ′, the horizontal axis represents the movement amount T4, and the vertical axis represents the output value of the function F4 ′. In this example, the output value of the function F4 ′ increases linearly from 0 to 1 when 0 ≦ T1 <d ′, and becomes 1 when d ′ ≦ T4. However, when d <d ′ is established with respect to d shown in FIG. 4D and the input value T4 is the same, the output value of the function F4 ′ is equal to or less than the output value of the function F4. . That is, normalizing using the function F4 ′ is narrower than the normalizing using the function F4, and the range allowing the amount of movement that seems to be an intruding object is narrower, making it difficult to be determined as an intruding object.

  In FIG. 5E, a graph 505 represents the relationship between the input value and the output value of the function F5, the horizontal axis represents the straight travel degree T5, and the vertical axis represents the output value of the function F5. In this example, the output value of the function F5 increases linearly from 0 to 1 when 0 ≦ T5 <e, and becomes 1 when e ≦ T5. That is, as the degree of straight travel T5 is small, that is, the change area is not linear, the change area is more likely to be planted, so the output value of the function F5 is reduced. Note that the threshold value e is determined in advance by experiment or experience so that movement by an intruding object and movement by shaking of planting can be distinguished.

  In FIG. 5F, a graph 506 represents the relationship between the input value and output value of the function F6, the horizontal axis represents the shape stability T6, and the vertical axis represents the output value of the function F6. In this example, the output value of the function F6 increases linearly from 0 to 1 when 0 ≦ T6 <f, and becomes 1 when f ≦ T6. That is, the smaller the shape stability T6, that is, the greater the change in the rectangular shape of the change area, the higher the possibility that the change area is due to planting, and thus the output value of the function F6 is reduced. Note that the threshold value f is determined in advance by experiment or experience so that a change in the rectangular shape by a person and a change in the rectangular shape by planting can be distinguished.

As shown in FIGS. 5A to 5D, the functions F1 ′, F2 ′, F3 ′, and F4 ′ have an output value closer to 1 as the change area is more human, and the change area is not as human. This is a function whose output value is close to 0. Further, as shown in FIGS. 5E to 5F, the functions F5 and F6 are so human that the change area is not likely to be planted, and the output value is close to 0, and the change area is not likely to be planted. Is a function whose output value is close to 1. At this time, the intruding object attribute value is calculated by, for example, the following equation using the functions F1 ′, F2 ′, F3 ′, F4 ′, F5, and F6.
(Intruding object attribute value) = F1 'x F2' x F3 'x F4' x F5 x F6 (12)
That is, the intruding object attribute value becomes closer to 1 as the change area is more human, and closer to 0 as the change area is less human. As described above, since the output values of the functions F5 and F6 are 0 or more and 1 or less, the intruding object attribute value when the functions F5 and F6 are multiplied is the intrusion when the functions F5 and F6 are not multiplied. The value is less than or equal to the object attribute value. Therefore, when the feature amounts T1 to T4 are the same, the intruding object attribute value calculated by Expression (12) is a value equal to or smaller than the intruding object attribute value calculated by Expression (9) in the normal determination process. Thereby, the determination means 36 can make it difficult to determine an intruding object in the special determination process than the normal determination process, and the change area that overlaps the planting area or the change area that is separated from the planting area. In addition to detecting an intruding object that moves in front of planting, it is possible to suppress false detection due to shaking of planting.

The threshold values a ′, b1 ′, b2 ′, b3 ′, b4, c ′, d ′, e, f (hereinafter referred to as special determination threshold values) shown in FIGS. The larger the overlapping rate T7 of the planting area, the smaller the output values of the functions F1 ′, F2 ′, F3 ′, F4 ′, F5 and F6, that is, it is difficult to determine that the changing region is due to humans. You may adjust so that it may become. For this purpose, for example, the special determination threshold values a ′, c ′, d ′, e, and f are increased as the overlapping ratio T7 of the planting area is increased, and the special determination threshold values b1 ′, b2 ′, and b3 ′ are increased. B4 is set so that the width between b1 ′ and b2 ′ and the width between b4 and b3 ′ become smaller as the overlapping ratio T7 of the planting area is larger. In that case, for example, the special judgment threshold when the overlapping ratio T7 of the planting area is less than 50% is a ₁ ', b1 ₁ ', b2 ₁ ', b3 ₁ ', b4 ₁ , c ₁ ', d ₁ ' , E ₁ , f ₁ and T7 is 50% or more and less than 75%, and the special judgment threshold values are a ₂ ′, b1 ₂ ′, b2 ₂ ′, b3 ₂ ′, b4 ₂ , c ₂ ′, d ₂ ', E ₂ , f ₂ and T7 is 75% or more, the special judgment thresholds are a ₃ ', b1 ₃ ', b2 ₃ ', b3 ₃ ', b4 ₃ , c ₃ ', d ₃ ', e _{As 3} and f ₃ , values that satisfy the conditions of the following equations are determined in advance.
a <a ₁ '<a ₂ '<a ₃ '(13)
b1 <b1 ₁ '<b1 ₂ '<b1 ₃ '(14)
b2 <b2 ₁ '<b2 ₂ '<b2 ₃ '(15)
b3> b3 ₁ '> b3 ₂ '> b3 ₃ '(16)
b4> b4 ₁ > b4 ₂ > b4 ₃ (17)
c <c ₁ '<c ₂ '<c ₃ '(18)
d <d ₁ '<d ₂ '<d ₃ '(19)
e ₁ <e ₂ <e ₃ (20)
f ₁ <f ₂ <f ₃ (21)
Thereby, the determination means 36 narrows the range which accept | permits that it is the feature-value like an intruding object, so that the ratio of the area | region where the change area | region and the planting area overlapped is large. Therefore, the determination means 36 can improve the discrimination accuracy between planting and people, can suppress erroneous detection due to shaking of the planting, and can detect an intruder that moves before planting.

In addition, the determination means 36 adjusts each special determination threshold value using the overlapping ratio T7 of the planting area in the current frame for the change area determined by the overlap determination means 34 to overlap the planting area. Good.
On the other hand, the determination unit 36 determines that the overlapping determination unit 34 has not overlapped with the planting area and the change determination area 35 has determined that the planting area has left the planting area. Each special determination threshold value may be adjusted using the overlapping rate T7 of the planting area when it overlaps. However, since the overlapping rate T7 of the planting region immediately before the change region leaves the planting region may be extremely low, the determination unit 36 calculates the overlapping rate T7 of the planting region calculated within the predetermined period. You may make it adjust each special determination threshold value using an average value.

Alternatively, the determination unit 36 calculates the change area determined by the overlap determination unit 34 as not overlapping with the planting region and determined to have left the planting region by the separation determination unit 35 by the separation determination unit 35. Each special determination threshold value may be adjusted using the number of overlaps T8 of the planted areas. In that case, for example, the special judgment threshold values when the number of overlapping T8 of the planting area is less than 5 are set as a ₁ ′, b1 ₁ ′, b2 ₁ ′, b3 ₁ ′, b4 ₁ , c ₁ ′, d ₁ ′. , E ₁ , f ₁ and T8 is 5 times or more and less than 10 times, and the special judgment threshold values are a ₂ ′, b1 ₂ ′, b2 ₂ ′, b3 ₂ ′, b4 ₂ , c ₂ ′, d ₂ ′ , E ₂ , f ₂ and T8 is 10 times or more, the special judgment threshold values are a ₃ ′, b1 ₃ ′, b2 ₃ ′, b3 ₃ ′, b4 ₃ , c ₃ ′, d ₃ ′, e ₃ , F ₃ .
As a result, the determination means 36 determines that the change area where the same moving object is reflected overlaps the planting area, and that the change area is more likely to be due to the planting. Narrow the range that allows object-like features. Therefore, the determination means 36 can improve the discrimination accuracy between planting and people, can suppress erroneous detection due to shaking of the planting, and can detect an intruder that moves before planting.

Further, in the special determination process, the determination means 36 uses only the size T1, the aspect ratio T2, the edge change rate T3, and the movement amount T4 of the intruding object without using the straight movement degree T5 and the shape stability T6. An attribute value may be calculated. In that case, in the special determination process, the determination unit 36 calculates the intruding object attribute value using only the functions F1 ′, F2 ′, F3 ′, and F4 ′, for example, by the following equation.
(Intruding object attribute value) = F1 'x F2' x F3 'x F4' (22)
When the feature amounts T1 to T4 are the same, the intruding object attribute value calculated by the equation (22) is calculated by the equation (9) in the normal determination process as in the case of the calculation by the equation (12). The value is less than or equal to the intruding object attribute value. Thereby, the determination means 36 can make it difficult to determine an intruding object in the special determination process than the normal determination process, and the change area that overlaps the planting area or the change area that is separated from the planting area. In addition to detecting an intruding object that moves in front of planting, it is possible to suppress false detection due to shaking of planting. Further, the determination means 36 can easily adjust the difficulty of determining an intruding object by changing the special determination threshold values of the functions F1 to F4 for these change areas.

Alternatively, in the special determination process, the determination unit 36 uses the same functions F1 to F4 as the normal determination process as functions for normalizing the size T1, the aspect ratio T2, the edge change rate T3, and the movement amount T4 of the intruding object. May be. In that case, the determination means 36 may calculate the intruding object attribute value using the functions F1, F2, F3, F4, F5, and F6, for example, according to the following equation in the special determination processing.
(Intruding object attribute value) = F1 x F2 x F3 x F4 x F5 x F6 (23)
When the feature amounts T1 to T4 are the same, the intruding object attribute value calculated by the equation (23) is calculated by the equation (9) in the normal determination process as in the case of the calculation by the equation (12). The value is less than or equal to the intruding object attribute value. Thereby, the determination means 36 can make it difficult to determine an intruding object in the special determination process than the normal determination process, and the change area that overlaps the planting area or the change area that is separated from the planting area. In addition to detecting an intruding object that moves in front of planting, it is possible to suppress false detection due to shaking of planting. Furthermore, since the determination means 36 discriminate | determines a person and planting based on the characteristic for distinguishing an intruding object and planting about these change areas, it can improve the discrimination | determination precision of a person and planting, and planting While being able to suppress false detection due to shaking, it is possible to detect intruders that move before planting.

  Alternatively, the determination unit 36 may calculate the intruding object attribute value using exactly the same function in the normal determination process and the special determination process, and may change only the threshold value to be compared with the intruding object attribute value. In this case, the threshold value used in the special determination process and compared with the intruding object attribute value may be set larger than the threshold value Th3 used in the normal determination process. Thereby, the determination means 36 can make it difficult to determine an intruding object in the special determination process than the normal determination process, and the change area that overlaps the planting area or the change area that is separated from the planting area. In addition, it is possible to suppress erroneous detection due to shaking of planting.

Moreover, when a change area | region has left | separated from the planting area | region, it can also be judged that the change area | region is highly likely to be a person from the case where it overlaps with a planting area | region now. Therefore, the determination means 36 makes it easier to determine that the change area determined to be separated from the planting area is due to a person than the change area determined to overlap the planting area. Also good. For example, the determination means 36 has functions F1 ′, F2 ′, F3 ′, and F4 ′ in the change region determined to be separated from the planting region than the change region determined to overlap the planting region. , F5 and F6 may be adjusted so that the output values become large, that is, the intruding object attribute value becomes large. Alternatively, the determination means 36 determines the intruding object attribute value by the expression (22) using only the functions F1 ′, F2 ′, F3 ′, and F4 ′ for the change region determined to be separated from the planting region. For the change area that is calculated and determined to overlap the planting area, the intruding object attribute value is calculated by the formula (12) using the functions F1 ′, F2 ′, F3 ′, F4 ′, F5, and F6. May be. Alternatively, the determination means 36 calculates the intruding object attribute value by the formula (23) using the functions F1, F2, F3, F4, F5, and F6 for the change area determined to be separated from the planting area. However, for the change area determined to overlap the planting area, the intruding object attribute value is calculated by the equation (12) using the functions F1 ′, F2 ′, F3 ′, F4 ′, F5 and F6. Also good.
Accordingly, the determination unit 36 makes it difficult to determine that the change region is due to a person, when the change region is separated from the planting region, than when the change region has not overlapped with the planting region in the past. It can be made easier to determine that the change area is due to a person than when the area overlaps with the planting area. Therefore, the determination means 36 can improve the discrimination accuracy between planting and people, can suppress erroneous detection due to shaking of the planting, and can detect an intruder that moves before planting.

The intruding object attribute value may be calculated by an average value or a weighted sum of F1 ′, F2 ′, F3 ′, F4 ′, F5, and F6, as in the normal determination process. When the average value of F1 ′, F2 ′, F3 ′, F4 ′, F5, and F6 is used, the intruding object attribute value is calculated by the following equation.
(Intruding object attribute value) = (F1 '+ F2' + F3 '+ F4' + F5 + F6) / 6 (24)
When the weighted sum of F1 ′, F2 ′, F3 ′, F4 ′, F5, and F6 is used, the intruding object attribute value is calculated by the following equation.
(Intruding object attribute value) = (F1 'x β1 + F2' x β2 + F3 'x β3 + F4' x β4 + F5 x β5 + F6 x β6) (25)
However, β1 to β6 are weighting factors determined depending on which of F1 ′ to F4 ′ and F5 to F6 is important, and the sum (Σβi (i = 1 to 6)) is 1. For example, when equalizing the weights of F1 ′ to F4 ′ and F5 to F6, all the weighting factors are set to the same value, and some of the feature amounts of F1 ′ to F4 ′ and F5 to F6 are not used. The corresponding weighting factor is set to zero. How to set the specific value of each weighting factor is appropriately optimized according to the environment in which the intruding object detection device 100 is installed, the purpose of detecting a person, and the like.

  The determination unit 36 determines whether or not the intruding object attribute value calculated in either the normal determination process or the special determination process is greater than or equal to a predetermined threshold value. When the intruding object attribute value is equal to or greater than a predetermined threshold, the determining unit 36 determines that the change area is due to the intruding object, and outputs an abnormality detection signal indicating that the intruding object has been detected in the monitoring area. 40. This threshold value is set to a value indicating that the change area is caused by an intruding object. For example, this threshold value is set to an intruding object attribute value that can be determined to be a possible intruding object, even a little, obtained by experiments.

  Further, when there is no change area determined to be due to the intruding object, the determining unit 36 determines that there is no intruding object in the input image, and the background stored in the storage unit 20 in the background image update unit 37. A background image update request for updating the image is transmitted. Alternatively, the determination unit 36 may transmit a background image update request to the background image update unit 37 to update the background image when the change area itself does not exist.

  Upon receiving the background image update request from the determination unit 36, the background image update unit 37 updates the background image stored in the storage unit 20 by replacing it with the input image. In this way, by always keeping the background image representing the latest state of the monitoring area, the image processing unit 30 can reduce false detection of an intruding object caused by weather changes, solar diurnal variations, and the like. .

  The background image update unit 37 may perform the background image update process at regular intervals (for example, every 10 minutes) rather than when the determination unit 36 determines that there is no intruding object. Good. In this case, for example, a time during which the average luminance value of the input image changes by a predetermined value or more due to sunshine fluctuations with respect to the monitoring area may be obtained by experiment, and the shortest time among the times obtained by the experiment may be set as the cycle for performing the update process.

  Alternatively, the background image update unit 37 may perform the above-described background image update process when there is illumination variation. In this case, for example, the background image update unit 37 may calculate the average luminance of the input image periodically and perform the background image update process when the average luminance changes to a predetermined threshold value or more. Note that an appropriate value can be determined as the threshold value by, for example, measuring the average luminance value of the input image when an illumination variation is actually generated.

  Alternatively, the background image update unit 37 may perform the background image update process using a moving average. In this case, for example, the background image update unit 37 may generate the background image so that the average value or the weighted average value of the luminance values of the corresponding pixels of the most recent predetermined number of input images is the pixel value.

  The planting region update unit 38 is a unit that updates the planting region stored in the storage unit 20. A method for extracting a planting region is disclosed in, for example, Japanese Patent Application Laid-Open No. 2007-249270. Specifically, the planting region update unit 38 obtains an edge pixel by performing an inter-pixel calculation using an edge detection filter such as a sobel filter or a prewitt filter on the input image. And the planting area | region update means 38 counts the frequency | count that the presence or absence of an edge pixel fluctuated for every pixel about the input image acquired within the fixed time, and the pixel position where the counted frequency | count is more than predetermined number is edge variation position. And Then, the planting area updating unit 38, in the change area of the current frame, when the ratio that the position of the edge pixel matches the edge fluctuation position is a predetermined value or more, that is, many of the edge pixels frequently repeat the edge fluctuation. In this case, the change area is set as a planting area.

  Alternatively, the planting area may be specified by an operator of the intruding object detection apparatus 100 or the like. In that case, the intruding object detection device 100 is provided with an input unit (not shown) such as a keyboard, a mouse, or a touch panel that receives an operation instruction from the operator and passes it to the image processing unit 30, and a planting area update unit 38 should just make the area | region designated from the operator via the input part the planting area | region. As a method for specifying the planting region, the operator may create image data in which the region where the planting is imaged and the region where the planting is not imaged are distinguished, or planting in the image data The operator may be allowed to input the coordinates of the pixel corresponding to the area in which the image is captured.

  The planting area update unit 38 may perform the planting area update process every time an input image is acquired, or may be performed at regular intervals (for example, every 10 minutes). . Or you may make it the planting area | region update means 38 perform the update process of a planting area, when there exists a lighting fluctuation | variation. Alternatively, when there is a change area that is not determined to be an intruding object, the planting area update unit 38 is likely to have occurred due to a change in planting, that is, the planting area may have changed. It may be determined that the planting area is high and the updating process of the planting area may be performed.

The output unit 40 includes a communication interface connected to a communication network such as a local area LAN or a public line network and a control circuit thereof. When the output unit 40 receives the abnormality detection signal from the determination unit 36, the output unit 40 outputs an alarm to the security device or the monitoring center device connected to the intruding object detection device 100 via the communication network.
The security device or the monitoring center device that has received the warning from the intruding object detection device 100 notifies the monitoring person that an intruding object has been detected in the monitoring area by voice or image.

  Next, the intruding object detection operation of the intruding object detection device 100 will be described with reference to the flowchart shown in FIG. The intruding object detection operation described below is controlled by the image processing unit 30 and is repeatedly executed at an image acquisition interval by the imaging unit 10.

First, an input image obtained by imaging the monitoring area is generated by the imaging unit 10 and transmitted to the image processing unit 30 (step S1).
When the image processing unit 30 receives the input image, the change region extraction unit 31 of the image processing unit 30 extracts the change region from the received input image and the background image stored in the storage unit 20 (step S2). Then, the change area extraction unit 31 attaches a label number to each extracted change area, and creates a label image having the pixel value as the label number of the corresponding change area (step S3).
When the change area extraction unit 31 extracts the change area and creates a label image, the tracking processing unit 32 changes the previous frame change area in which the same moving object is shown in the change area of the current frame. Are associated with each other (step S4).

When the tracking processing unit 32 associates the change region of the current frame with the change region of the previous frame, the feature amount calculation unit 33 calculates the feature amounts T1 to T8 of the intruding object in the change region for each change region (Step S1). S5).
When the feature amount calculation unit 33 calculates the feature amounts T1 to T8 of the intruding object, the overlap determination unit 34 uses the planting region overlap rate T7 calculated by the feature amount calculation unit 33 to change the change region into the planting region. (Step S6).
When it is determined by the overlap determination unit 34 that the change region does not overlap the planting region, the leaving determination unit 35 uses the planting region overlap count T8 calculated by the feature amount calculation unit 33 to determine whether the change region is a change region. It is determined whether or not the planted area has left (Step S7).

If it determines with the change | decrease area | region not having left | separated from the planting area | region by the detachment | determination determination means 35, the determination means 36 will determine by a normal determination process whether a change area | region is an intruding object (step S8). When the determination unit 36 determines that the change area is caused by an intruding object, the determination unit 36 transmits an abnormality detection signal to the output unit 40.
On the other hand, when it is determined by the overlap determination means 34 that the change area overlaps the planting area, or when the change determination area 35 determines that the change area has left the planting area, the determination means 36 Whether or not the change area is caused by an intruding object is determined by a special determination process (step S9). When the determination unit 36 determines that the change area is caused by an intruding object, the determination unit 36 transmits an abnormality detection signal to the output unit 40.

  When the determination unit 36 determines whether or not the change area is caused by an intruding object by the normal determination process or the special determination process, the determination unit 36 determines whether or not the intruding object determination process has been performed for all the change areas (step S10). . If there is a change area where the intruding object determination process is not performed, the determination unit 36 returns the control to step S5 and repeats the processes of steps S5 to S8.

On the other hand, when the determination process of the intruding object is performed for all the change areas, the determination unit 36 determines whether or not there is a change area determined to be an intruding object, and there is no change area determined to be an intruding object. In this case, a background image update request for causing the background image update means 37 to update the background image is transmitted. When receiving the background image update request from the determination unit 36, the background image update unit 37 updates the background image (step S11).
When the background image update unit 37 updates the background image, the planting region update unit 38 performs a planting region update process (step S12).
When the planting region update unit 38 performs the planting region update process, the output unit 40 determines whether an abnormality detection signal has been received from the determination unit 36. When the output unit 40 receives an abnormality detection signal from the determination unit 36, the output unit 40 outputs an alarm to the security device or the monitoring center device. (Step S13).

  Note that the details of the processing in each of the above steps have been described in detail in the description of each part of the intruding object detection device 100, and thus the description thereof is omitted here.

As described above, the intruding object detection device according to the present invention calculates an intruding object attribute value representing the degree of intruding object based on the feature amount of the intruding object for each changing region, and the changing region is planted. When overlapping with the planting area, it is difficult to detect the changed area as an intruding object. As a result, the intruding object detection device can suppress erroneous detection of the change area caused by the shaking of the planting as an intruding object, and further intrudes if it is a change area caused by the intruder moving before planting. It can be detected as an object.
Further, the intruding object detection device according to the present invention adjusts the calculation criterion of the intruding object attribute value based on whether or not the change area overlaps the planting area, and therefore the intruding object attribute value based only on the characteristics of the intruding object. It is possible to determine the likelihood of an intruding object with higher accuracy than calculating.

Furthermore, the intruding object detection device according to the present invention is a feature amount for distinguishing an intruding object from planting in addition to the intruding object feature amounts T1 to T4 when the change region overlaps the planting region. The intruding object attribute value is calculated based on the straight travel degree T5 and the shape stability T6. Therefore, the intruding object detection device can determine the likelihood of an intruding object with high accuracy for an intruder moving before planting. Misjudgments can be suppressed.
Furthermore, in the intruding object detection device according to the present invention, even when the change area does not overlap with the planting area, if the change area overlaps with the planting area in the past frame, the change area overlaps with the planting area. The presence or absence of an intruding object is determined by the same method. Therefore, the intruding object detection device can determine an object that has left the planting region, determine that the object is likely to be planted, and can suppress erroneous detection of planting shake as an intruding object.

  As described above, those skilled in the art can make various modifications in accordance with the embodiment to be implemented within the scope of the present invention.

DESCRIPTION OF SYMBOLS 10 Imaging part 20 Storage part 30 Image processing part 31 Change area extraction means 32 Tracking processing means 33 Feature amount calculation means 34 Overlap determination means 35 Departure determination means 36 Determination means 37 Background image update means 38 Planting area update means 40 Output section 100 Intruder detection device

Claims (4)

An intruding object detection device for detecting an intruding object that has entered a monitoring area,
An imaging unit that acquires a monitoring image obtained by imaging the monitoring region;
A storage unit that stores a background image that is an image obtained by capturing the monitoring area in a situation where the intruding object does not exist in the monitoring area, and a planting area that indicates an area on an image in which planting is captured,
A change area extraction means for extracting a change area in which a luminance value has changed according to a difference between the monitoring image and the background image;
For the change area, calculate an intrusion object attribute value indicating the degree of intrusion object likelihood that the intrusion object likelihood is lower than when the change area overlaps the planting area, and the intrusion object attribute value Determining means for determining whether the change area is caused by an intruding object based on
An intruding object detection device comprising:   When the change area does not overlap the planting area, the determination unit calculates the intruding object attribute value based on a first feature amount that is a characteristic of the intruding object, and the change area is the planting area. The intruding object attribute value is calculated based on a second feature amount that is a feature other than the first feature amount that is significantly different between the intruding object and planting in addition to the first feature amount. The intruding object detection device according to claim 1. Further, each time the monitoring image is acquired by the imaging unit, the latest change region extracted from the latest monitoring image and the latest change region extracted from the monitoring image acquired immediately before are acquired from the latest change region. Tracking processing means for associating the object reflected in the change area with the immediately preceding change area in which the same object is reflected,
The determination unit is configured so that the latest change area does not overlap the planting area, and the change area in the monitoring image immediately before or a predetermined number of times associated with the latest change area is the planting. If the area does not overlap, the intruding object attribute value is calculated by a first process, the latest change area does not overlap the planting area, and the latest change area is associated with the latest change area If the change area in the monitoring image immediately before or a predetermined number before is overlapped with the planting area, the intrusion is performed by the second process in which the intruding object attribute value is not a value that is not an intruding object than the first process. The intruding object detection device according to claim 1, wherein the object attribute value is calculated.   The intruding object detection apparatus according to claim 2, wherein the determination unit sets the intruding object attribute value to a value that is less likely to be an intruding object as the ratio of the area that overlaps the planting area in the change area increases.

Images