JPH08221549A - Intrusion object recognizing device - Google Patents

Abstract

PURPOSE: To exactly detect the number of intrusion objects even when the images of plural intrusion objects are picked up while overlapping each other by recognizing the number of intrusion objects from an overlapped change area partitioned by overlapped edge line segments including a lot of edges among edge line segments extracted from edge straight lines inside the overlap change area. CONSTITUTION: A video interface circuit 10 converts a photographed image signal from a video camera 9 into a digital signal and supplys it to a background image memory circuit 1 or an input image memory circuit 2 and a difference processing circuit 3 calculates difference between picture elements from image signals from the image memory circuits 1 and 2. A change area division processing circuit 5 detects a straight line including an edge from the input image, divides this straight line into line segments, selects only the line segment expressing the overlap of intrusion objects and divides the changed area. Labeling processing 6 performs labeling to the divided changed areas and gives label numbers thereto, and a label count processing circuit 7 calculates the number of label numbers and gives this number to an output processing circuit 8 as the number of intrusion objects.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a device for recognizing an intruding object, which can be applied to a device for recognizing the number of persons when a plurality of persons are superposed on an image taken by a television camera or the like. .

[0002]

2. Description of the Related Art In recent years, research and development of a moving object recognition apparatus using a television camera have been performed for the purpose of monitoring an intruding object. As a proposal of a technique for detecting an intruding object in such a moving object recognition device, there is a technique as shown in the following document.

Reference: Proceedings of the Autumn Conference of the Institute of Electronics, Information and Communication Engineers, 1990, D-436, "Detection of Intruding Object by Difference between Frames of Differential Image".

According to the technique according to this document, as an object recognition method in an intruding object monitoring system, an input image for confirming the presence or absence of an intruding object and a background image are input from a television camera, and the input image and the background are input. The change area is obtained from the difference binary image from the image, the presence or absence of an intruding object is determined from the feature amount of the change area, and if there is an intruding object, for example, an alarm is issued.

Here, as a method of determining the presence or absence of an intruding object from the characteristic amount of the changing area, the presence or absence of an intruding object is determined using the area of the changing region and the circumscribed rectangle as parameters. In other words, the area of the change area is equal to or larger than a certain value, and the aspect ratio of the circumscribing rectangle (vertical length / horizontal length) is within the range. .

[0006]

However, in the conventional technique, when a plurality of intruding objects appear to overlap with each other, when the conventional difference processing or binarization processing is performed, these intruding objects are regarded as one change region. It had been extracted. Therefore, even if the subsequent processing is performed, the number of intruding objects may be erroneously detected.

From the above, there is a demand for an intruding object recognition apparatus which can accurately detect the number of intruding objects even when a plurality of intruding objects are imaged in an overlapping manner.

[0008]

Therefore, according to the present invention, an input is made from an image signal obtained by differential binarization of a background image signal picked up when an intruding object does not exist and an input image signal sequentially picked up. An intruding object recognition apparatus that detects an overlapping change area in an image and analyzes the overlapping change area to recognize an intruding object solves the above-described problem with the following characteristic configuration.

That is, the present invention detects an edge in the overlap change area from the input image, detects an edge straight line in the overlap change area from this edge, and recognizes an intruding object from the detected edge straight line. The effective edge line segment is extracted, and from this effective edge line segment, the overlapping edge line segment extraction means for extracting the edge line segment containing many edges in the overlapping change region as the overlapping edge line segment, and the overlapping edge line segment An intruding object number recognizing means for recognizing the number of intruding objects from the divided overlapping change areas is provided.

[0010]

According to the structure of the present invention, a portion in which the pixel value is abruptly changed between adjacent pixels in the overlap change region is detected as an edge. This is because, in general, a portion having different densities appears in a portion where intruding objects overlap. By performing statistical processing from the edges detected in this way,
The edge straight line is obtained along the horizontal and diagonal directions.

Since partial edges between intruding objects are not always extracted completely due to the influence of the photographing condition and noise, such a statistical processing is performed to obtain a probable edge straight line. There is a need.

Further, a line segment effective for discriminating the intruding object which overlaps from the detected edge straight line is extracted as an edge line segment. To extract this edge line segment, use the intersection of the edge straight lines or the intersection of the circumscribing polygon surrounding the overlap change area and the edge straight line, and connect the intersections to extract a valid edge line segment. is there.

Furthermore, the overlapping edge line segment for extracting the overlapping of the intruding object from the effective edge line segment and for dividing the overlapping change region is extracted. It is preferable that the overlapping edge line segment is a line segment that includes many edges in the overlapping change region from the edge line segments. Also, by connecting or connecting the edge line segments, it is possible to extract an appropriate overlapping edge line segment.

Since the obtained overlapping edge line segment divides the overlapping change region, it is possible to divide an intruding object that can be seen as overlapping in various overlapping states even if the edge of the intruding object is not clearly extracted. This makes it possible to accurately recognize the number of intruding objects.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a functional block diagram of an intruding object recognition apparatus according to an embodiment of the present invention. In FIG. 1, an intruding object recognition apparatus includes a background image memory circuit 1 and an input image memory circuit 2
A differential processing circuit 3, a binarization processing circuit 4, a change area division processing circuit 5, a labeling processing circuit 6, a label counting processing circuit 7, an output processing circuit 8, and a video camera 9.
And a video interface circuit 10.

A characteristic structure of this intruding object recognition apparatus is that it is provided with a changing area division processing circuit 5.

In FIG. 1, a video camera 9 is for picking up an intruding object and the like, and outputs a picked-up image as an analog signal and supplies it to the video interface circuit 10. When the video interface circuit 10 is supplied with the captured image signal, it converts the captured image signal into a digital signal and supplies it to the background image memory circuit 1 or the input image memory circuit 2. In particular, the input image memory circuit 2 is the video camera 9
The input image signals sequentially photographed by are stored as digital signals. On the other hand, the background image memory circuit 1 stores, as a digital signal, a background image signal captured when there is no intruding object (subject).

The difference processing circuit 3 is the input image memory circuit 2
Input image signal given from the background image memory circuit 1
The difference between the pixels (for example, the absolute value of the density difference) is obtained from the background image signal given by the above, and a difference image signal is generated and given to the binarization processing circuit 4. Binarization processing circuit 4
When the differential image signal is given, the binarization is performed with an appropriate threshold value to generate a differential binarized image signal and apply it to the change region division processing circuit 5. In this differential binarized image signal, the value of the pixel whose density value has changed from the background image signal in the input image signal is set to 1 and the value of the pixel which has not changed is set to 0.

The change area division processing circuit 5 divides the change area based on the internal edge in the input image in the range of the overlap change area. More specifically, a straight line including an edge is detected from the input image, the straight line is divided into line segments by using the intersections of the straight lines and the intersections of the straight lines and the circumscribed quadrangle, and within the line segment. By utilizing the ratio including edges and the connection relationship between line segments, only the line segment representing the overlap of intruding objects is selected, and the change region is divided by the line segment.

The labeling process 6 is to label the divided change regions, give a separate label number to each isolated change region, and give it to the label counting processing circuit 7.

The label counting processing circuit 7 calculates the number of label numbers given to the change area, and uses this number as the number of intruding objects and gives it to the output processing circuit 8. The output processing circuit 8 outputs the number of intruding objects as display information to a display device or the like.

(Operation of Entire Device): Video Camera 9
The analog image signal from is supplied to the video interface circuit 10, where it is converted into a digital image signal. The sequentially captured input image signals supplied from the video interface circuit 10 are stored in the input image memory 2, and the background image signal is stored in the background image memory circuit 1. The difference between the input image signal and the background image signal given from the background image memory circuit 1 is obtained by the difference processing circuit 3, and a difference image signal is generated. This difference image signal is given to the binarization processing circuit 4, binarized with an appropriate threshold value, a difference binarization image signal is generated, and given to the change region division processing circuit 5.

A straight line including an edge in the input image is detected from the difference binary image signal, and the straight line is divided into line segments by utilizing the intersections of the straight lines and the intersections of the straight lines and the circumscribed quadrangle. Then, only the line segment that represents the overlap of intruding objects is selected by using the ratio of edges in the line segment and the connection relationship between the line segments, and the change area is divided by the line segment.
It is given to the labeling process 6.

Labeling is performed on the divided change areas, and a separate label number is given to each of the isolated change areas, which is given to the label counting processing circuit 7. The label count processing circuit 7 calculates the number of label numbers given to the change area, and this number is given to the output processing circuit 8 as the number of intruding objects. The number of intruding objects is output as display information on a display device or the like.

(Details of change area division processing circuit 5):
FIG. 2 is a detailed functional block diagram of the change area division processing circuit 5. In FIG. 2, the change area division processing circuit 5 includes an inside-change-area edge detection processing circuit 51, a direction-specific edge projection processing circuit 52, an edge straight line detection processing circuit 53, an edge line segment detection processing circuit 54, and overlapping edges. It is composed of a line segment detection processing circuit 55 and an overlapping edge line segment inversion processing circuit 56.

The in-change-region edge detection processing circuit 51 performs an in-change-region edge detection process on the changed region obtained by the processing up to the binarization processing circuit 4.
This edge-in-change-region detection processing is processing for obtaining an edge of the input image in the change area, that is, a portion in which the pixel value rapidly changes between adjacent pixels, and is performed in the following procedure.

First, a range of a change area is cut out from the input image (hereinafter, this cut out area is referred to as an "input change area"), and the input change area is vertically, horizontally, or horizontally.
Edges in each diagonal direction (45 ° to the left and right) are detected.
The edge detection process here is performed by Sobel.
Perform using an operator such as an operator. By this processing, an “input change area edge image” in which an edge representing the contour of the change area and an edge inside the input change area are extracted is obtained. Thereafter, the image is binarized to obtain a "binarized input change region edge image" in which the pixel values at the edge portions are set to 1 and the other pixel values are set to 0.

Next, edge detection processing is performed on the changed area, edges representing the contours of the changed area are extracted, and binarization processing is similarly performed to obtain a "binarized changed area edge image". The difference between these images is calculated, and only the internal edges excluding the edges representing the contour in the input change area are extracted.

FIG. 3 is a diagram showing an example in which two person images are overlapped. FIG. 4 is a diagram showing a change area in the image in which the person images in FIG. 3 are overlapped. Furthermore, FIG.
FIG. 6 is a diagram in which edges are detected from the change area shown in FIG. 4 in each of vertical, horizontal, and diagonal (45 ° left and right) directions within the change area. FIG. 5A shows a vertical edge image in the input image in the change area. FIG. 5B shows a lateral edge image in the input image in the change area. FIG. 5C shows an oblique (right 45 °) direction edge image in the input image in the change area. Figure 5
(D) represents an oblique (left 45 °) direction edge image in the input image in the change area.

Further, the above "binarization change area edge image"
Performs edge detection on a change area that has already been binarized as shown in FIG. 4, so that an edge is detected only at the contour of the change area as shown in FIG. On the other hand, since the “binarization change area edge image” performs edge detection processing on the input image (human part in FIG. 3) in the range of the change area, as shown in FIG. Part where the concentration is different depending on the overlapping of (the overlapping part of the person in FIG. 7), and further the part where the concentration is different within the individual (neck part in FIG. 7)
An edge is detected at.

Therefore, when the difference between the two images is taken, the result shown in FIG.
As described above, due to the overlap between individuals, only the edges of the portions having different concentrations and the portions having different concentrations within the individual are detected. However, FIG. 6 to FIG. 8 are general edge images in which directionality is ignored, not edges for each direction. It should be noted that FIG. 8 is taken for each direction and becomes as shown in FIGS.

As can be seen from FIG. 5, the portion where the intruding object overlaps is extracted as an edge. This is because in general, the density of the overlapping portions of the intruding objects is different.
However, the partial edges between the intruding objects are not always extracted as connected edges due to the influence of the imaging situation and noise. Therefore, the direction-dependent edge projection processing is performed.

The direction-dependent edge projection processing circuit 52 performs a projection process on the edge in each direction obtained by the direction-dependent change area edge detection circuit 51 in each direction. The projection processing is to add pixels having a value of 1 representing the pixels in the edge portion in the direction of the edge as shown in FIGS. 9 and 10 and express the result of this addition in a histogram. FIG. 9A is a vertical edge projection histogram of FIG. 5A. FIG. 9B is the above FIG. 5B.
3 is a horizontal edge projection histogram of the. Figure 10
5A is an oblique (rightward 45 °) direction edge projection histogram of FIG. 5C. FIG. 10B shows the above-mentioned FIG.
It is a diagonal (left 45 degree) direction edge projection histogram of (d).

As can be seen from FIGS. 9 and 10, the vertical method,
The values of the histogram are high in the portions having edges in the horizontal direction, the 45 ° right direction, and the 45 ° left direction. Therefore, the edge straight line detection processing is next performed.

In the histogram in each direction, the edge straight line detection processing circuit 53 draws a straight line in each direction for coordinates whose histogram value is larger than a predetermined value (hereinafter, this straight line will be referred to as "edge"). "Straight line").

Since this edge straight line is a straight line containing many edges in each direction, it plays a role of connecting the interrupted edges in each direction as shown in FIG. Hereinafter, the formulas of the vertical, horizontal, and diagonal (right 45 ° direction, left 45 ° direction) straight lines are shown.

Y = xi (1) x = yj (2) y = x + yk (3) y = -x + yl (4) However, xi, yj, yk, and yl are constants.

FIG. 11 shows edge straight lines 11a to 11c obtained from the histograms of FIGS. 9 and 10 obtained from the overlapping human image of FIG.

The edge line segment detection processing circuit 54 extracts a line segment effective for separating an intruding object from an edge straight line. Therefore, first, the change area in FIG. 12 is surrounded by a circumscribed quadrangle, and the intersection of the circumscribed quadrangle and the edge straight line is
And the edge straight line is divided into line segments based on the intersection with the edge straight lines (hereinafter, this line segment is referred to as an edge line segment). 12A to 12N represent edge line segments.

The overlap edge line segment detection processing circuit 55 represents an overlap of an intruding object from among the edge line segments, and obtains an edge (hereinafter, referred to as an "overlap edge") for separating a change region. Therefore, there are various methods for this overlapping edge line segment detection processing, but the following method can be used here in consideration of the case where intruding objects do not overlap intricately.

First, a line segment including many edges in the change area is obtained from the edge line segments. For this purpose, from the total value C of the number of pixels of the change area and the total value E of the number of pixels of the edge in each line segment, the following equation (5) is used.
Then, the ratio R of the edges to the change area in each line segment is obtained.

R = E / C × 100 (%) (5) If this ratio R is a predetermined value or more, the line segment is regarded as a line segment including many edges, and, for example, a label 10 is attached. As an example, for the edge line segment in FIG. 12, B, D,
Labels 10 are attached to the edge line segments of H, I, K, and M. Labels 20 are attached to the edge line segments other than these, for example. Since the engine line segment labeled 10 by these processes includes many edges in the input change region, it is considered that there is a high possibility that the edge line segment represents an overlap, and when the overlapping edge line segment is extracted. Detect with priority. However, for the edge line segment where the value of C is 0,
R does not calculate but unconditionally gives the label 0.

Next, among the engine line segments labeled 0, 10 and 20, the edge line segment that intersects the circumscribing quadrangle is on the right side of the label number, that is, the ones digit is 1, and edges that do not intersect. Set the line segment to 2. Further, since the edge line segment with the label number of one digit, that is, the edge line segments with the label numbers 1 and 2 are the edge line segments that are not on the change area, they are deleted and are connected on the extension line of the edge line segment at that time. The label number of the edge line segment is changed to 1 at the first digit.

FIG. 13 is an explanatory diagram of labeled edge line segments. In FIG. 13, the edge line segment with the label number 11 includes many edges in the current image and represents one at the end of the change area, and the edge line segment with the label number 12 includes many edges in the current image, In addition, it represents something in the change region. The edge line segment with the label number 21 indicates that it does not include the edges in the current image and is located at the end of the change area, and the edge line segment with the label number 22 does not include many edges in the current image and the change area. It represents what is inside.

The edge line segments having these labels are connected by the following procedure to extract overlapping edge line segments for dividing the change area in the circumscribed quadrangle.

(Extraction of overlapping edge line segment candidates): First, overlapping edge line segment candidates are extracted. In this extraction method, the edge line segment with label number 11 in FIG. 13 includes label B, label I, and label M using the label numbers in FIG. Therefore, overlapping edge line segment candidates are extracted from these edge line segments.

Regarding the edge line segment of the label B, the labels C, D and E of the edge line segments to be connected are first compared. The label numbers in this case are 21 for label C, 12 for D, and 2 for E.
Therefore, the label D having the smallest label number is selected here. Next, regarding the labels G, K, and H of the edge line segments connected to the label D, the edge number of the label G is 2
2, K is 12, H is 12, and K and H are the smallest. Therefore, when comparing the interior angles of these two, K is 1
Since 80 ° and H are 45 °, the label K with a large interior angle
Select

The label numbers J, L, and M of the edge line segments connected to the label K have label numbers 22, 21, and 1, respectively.
Since it is 1, the label M is selected. Since the number on the right side of the label number of the label M is 1, this process is terminated.
Therefore, the overlapping edge line segment candidate in the label B is the one in which the labels B, D, K, and M are connected.

Regarding the edge line segment of label I, the label numbers of the edge line segments E, H, and N to be connected are respectively
Since it is 22, 12, and 21, the label H having the smallest number is selected. Edge line segments connected to label H are D, G, K
Since the label numbers are 12, 22 and 12, and the interior angles of the labels D and K with the label H are 45 ° and 135 °, respectively, the label K having the smallest label number and the largest interior angle is selected. For this label K, the label M is selected in the same manner as described above, and the processing ends. Therefore, the overlapping edge line segment candidates for label I are labels I, H, K,
M is connected.

Regarding the edge line segment of the label M, the label numbers of the edge line segments L, J and K to be connected are respectively
Since the labels are 21, 22, and 12, the label K having the smallest number is selected. The edge line segment connected to the label K is D,
The label numbers of G and H are 12, 22 and 12, respectively, and the interior angles of the labels D and H with the label K are 180 ° and 135 °, respectively, so that the label D having the smallest label number and the largest interior angle is selected.

The labels B, C, and E of the edge line segments connected to the label D have label numbers 11, 21, and 1, respectively.
Since the number is 2, the label B having the smallest number is selected and the process ends. Therefore, the overlapping edge line segment candidate in the label M is the one in which the labels M, K, D, and B are connected.

That is, in this example, there are two overlapping edge line segment candidates, one having labels M, K, D and B connected and one having labels I, H, K and M connected. , H, K, and M are connected to each other, the intruding objects are vertically separated by the labels I and H which are the horizontal edge line segments, and the intruding objects are rarely vertically overlapped. Since it violates the condition, it is not an appropriate overlapping edge line segment. Therefore, here, only the labels M, K, D, and B connected to each other are overlapped edge line segments.

As described above, in the case of extracting the overlapping edge line segment candidate, starting from the edge line segment of label number 11, the edge line segment connected to the line segment is most selected from the following "evaluation criteria". A line segment having a high evaluation value is selected and connected, and then the same processing is performed on the connected edge line segment. Further, when the edge line segment whose number on the right side of the label number is 1 is connected, this processing is ended.

As the evaluation criteria, (1) the smaller the number of the label number, the higher the evaluation value. (2) When the label numbers are the same, the evaluation value is higher as the interior angle formed by these two edge line segments is larger.

However, if there is the same evaluation value among the edge line segments connected to the line segment, the process is continued for each of them until the end, and overlapping edge line segment candidates are detected.

Further, from the extracted overlapping edge line segment candidates, unsuitable overlapping may occur due to physical conditions such as infrequent vertical overlap of intruding objects on the actual image, or conditions unique to the environment. Edge line segment candidates are excluded. FIG. 14 shows the overlapping edge line segment 14a obtained as described above.

The overlapping edge line segment inversion processing circuit 56 sets the value of the pixel representing the overlapping edge line segment in the change area to 0. As a result of this processing, the region representing the intruding object which is seen as overlapping as shown in FIG. 14 is divided into regions 14b and 14c corresponding to each intruding object. That is, it is possible to obtain divided areas according to the number of intruding objects.

(Operation of change area division processing circuit 5):
The change area inside edge detection processing circuit 51 performs the change area inside edge detection processing on the change area obtained by the processing up to the binarization processing circuit 4. By this processing, the edge of the input image within the change area, that is, the portion where the pixel value changes rapidly between the adjacent pixels is obtained.

Next, the change area edge detection circuit 51 for each direction
The edge-dependent edge projection processing circuit 52 performs projection processing (processing for histogram expression) on the edges in the respective directions (vertical, horizontal, and diagonal directions) obtained in (3). Next, in the histogram in each direction,
The edge straight line detection processing circuit 53 performs processing for drawing an edge straight line in each direction for coordinates whose histogram value is larger than a predetermined value.

Next, the edge line segment detection processing circuit 54 extracts an effective line segment for separating the intruding object from the edge straight line from the edge straight line in each direction.
From the extracted effective line segments, the overlapping edge line segment detection processing circuit 55 represents the overlapping of the intruding object from the edge line segments, and obtains the overlapping edge line segments for dividing the change region. Next, from the overlapping edge line segment, in the overlapping edge line segment inversion processing circuit 56, the value of the pixel representing the overlapping edge line segment is set to 0 in the change region, and the region representing the intruding object which was seen overlapping is invaded. It is divided into areas corresponding to the objects, and divided areas corresponding to the number of intruding objects are obtained.

After this, the divided change areas are labeled, and different label numbers are given to the isolated individual change areas, which are given to the label counting processing circuit 7. The number of label numbers assigned to the change area indicates the label counting processing circuit 7
The number of intruding objects is calculated and is given to the output processing circuit 8. The number of intruding objects is output as display information on a display device or the like.

FIG. 15 is a hardware configuration diagram of an example for realizing the intruding object recognizing device according to the embodiment of the present invention. In FIG. 15, the hardware configuration of the intruding object recognition apparatus includes, for example, a video camera 9, a workstation body 220, a CRT display 26, a keyboard 25, a mouse 24, and an external storage device 27. There is. The workstation body 220 is C
PU20, main memory 210, video board 28
It consists of and.

The image signal taken by the video camera 9 is
An analog signal (RGB signal or NTSC signal) is provided to the video board 28. This video board 28
Converts the input image signal from an analog signal into a digital signal, and stores the background image and the sequentially captured input image signal in the main memory 210. This main memory 2
In addition to storing the image signal from the video board 28, 10 also stores various program data for performing the above-mentioned person image recognition, various necessary work data, and the like. The CPU 20 processes these data in the main memory 210 to perform image recognition processing of an intruding object.

The external storage device 27 is a main memory 210.
In addition to being used for backing up program data, it stores various data necessary for image recognition processing of the number of intruding objects, and stores the main memory 21 as necessary.
It is given to 0. The result of the image recognition processing for the number of intruding objects is displayed and output on the CRT display 26. The mouse 24, the keyboard 25, etc. are used as input means for data selection and data input at the processing stage. By configuring the intruding object recognition device in this way, it is possible to perform image recognition of the number of intruding objects described above.

(Effects of the Embodiment) According to the above embodiments, a plurality of intruding objects originally appear to overlap each other, so that a change occurs when they are extracted as one change region by difference processing and binarization processing. Detect edges in the region,
Processing is performed to extract overlapping edge line segments that can separate the overlapping of intruding objects from the detected edges in various situations, and the overlapping change area is divided (divided) based on this edge line segment. Therefore, even when the edges are not clearly extracted between the overlapping of the intruding objects in various situations, it is possible to divide the intruding objects that appear to overlap with each other, and it is possible to accurately detect the number of the intruding objects.

Further, the edge straight line is subjected to histogram projection processing for each of the vertical edge, the horizontal edge, and the diagonal edge in the overlap change area to obtain a statistical amount, and from this statistical amount, a straight line including many edges is obtained. Since the detection is performed by subtracting, the edge straight line can be appropriately obtained regardless of the edge distribution in any direction.

Furthermore, a line segment obtained by enclosing an overlapping change area with a circumscribing polygon and connecting either the intersection of the circumscribing polygon and the edge straight line or the intersection of the edge straight lines with each other Therefore, it is possible to obtain the edge line segment suitable for the division of the overlap change region while conforming to the original edge distribution tendency.

Furthermore, since the overlapping edge line segment is extracted by connecting the effective edge line segments, it is possible to accurately and flexibly obtain the line segment even if complicated overlapping occurs.

Further, since the original shape of the intruding object can be cut out from the overlapping change areas almost accurately, it can be used not only for counting the intruding object but also for identifying the type of the intruding object.

Other Embodiments: (1) In the edge detection processing of the above embodiments, the Sobel (Sobe)
l) The operator was used, but in addition, for example, Robinson operator and Kirche (Kirsche)
It is also preferable to use a sch) operator or the like.

(2) When detecting an edge in the input image within the range of the change area, the difference image between the edge image of the background image (each direction) and the edge image of the input image (each direction) is calculated. The same effect as that of the above-described embodiment can be obtained by performing the binarization process.

(3) Further, in the process of obtaining the edge straight line, the same effect as that of the above-described embodiment can be obtained even if the edge straight line is obtained by using the Hough transform in addition to the process performed by using the edge projection histogram. You can

(4) Furthermore, although FIGS. 1 and 2 of the above-mentioned embodiment are shown as a functional configuration diagram of the intruding object recognizing device, they can be viewed as a flow chart showing a procedure (flow) of processing.

(5) Further, in FIGS. 12 to 14 described above, the circumscribed quadrangle is formed around the changing region, but depending on the shape of the changing region, a polygon (triangle, pentagon or more) is preferable. There are cases.

(6) Furthermore, the intruding object recognition device of the present invention is, for example, a video conference system, various monitoring camera devices, an entrance camera device, a monitoring device mounted on a moving body such as a vehicle or an aircraft. Can also be applied to.

[0076]

As described above, according to the present invention, the overlapping edge line segment extracting means detects the edge in the overlapping change area from the input image and detects the edge straight line in the overlapping change area from this edge. , Extract an effective edge line segment for recognizing an intruding object from the detected edge straight line, and extract an edge line segment containing many edges in the overlap change region as an overlapping edge line segment from the effective edge line segment, The intruding object number recognition means is configured to recognize the number of intruding objects from the overlapping change area divided by the overlapping edge line segments, so that even if a plurality of intruding objects are imaged in an overlapping manner,
It is possible to realize an intruding object recognition device that can accurately detect the number of intruding objects.

[Brief description of drawings]

FIG. 1 is a functional configuration diagram of an intruding object recognition apparatus according to an embodiment of the present invention.

FIG. 2 is a functional configuration diagram of a change area division processing circuit according to an embodiment.

FIG. 3 is an explanatory diagram of an intruding object in the case where the examples are seen to overlap each other.

FIG. 4 is an explanatory diagram of an overlap change area with respect to FIG.

5 is an explanatory diagram of edge images for each direction in the input image with respect to the overlapping change area in FIG. 4;

FIG. 6 is an explanatory diagram of a binarized change area edge image with respect to FIG. 3;

7 is an explanatory diagram of a binarized input change area edge image with respect to FIG. 3;

8 is an explanatory diagram of an image of a difference between the binarized change area edge image of FIG. 6 and the binarized input change area edge image of FIG. 7;

9 is an explanatory diagram of vertical and horizontal edge projection histograms for FIG.

10 is an explanatory diagram of a diagonal edge projection histogram for FIG.

11 is an explanatory diagram of an edge straight line with respect to FIGS. 9 and 10. FIG.

12 is an explanatory diagram of an edge line segment in FIG.

13 is an explanatory diagram of labeled edge line segments for FIG. 12;

FIG. 14 is an explanatory diagram of division of a change area with respect to FIG.

FIG. 15 is a hardware configuration diagram of an example of an intruding object recognition device according to an embodiment.

[Explanation of symbols]

1 ... Background image memory circuit, 2 ... Input image memory circuit, 3
... difference processing circuit, 4 ... binarization processing circuit, 5 ... change area division processing circuit, 6 ... labeling processing circuit, 7 ... label counting processing circuit, 8 ... output processing circuit, 9 ... video camera, 10
... video interface circuit, 11a to 11c ... edge straight line, 14a ... overlapping edge line segment, 51 ... change area edge detection processing circuit, 52 ... direction-dependent edge projection processing circuit,
53 ... Edge straight line detection processing circuit, 54 ... Edge line segment detection processing circuit, 55 ... Overlap edge line segment detection processing circuit, 56
... overlapping edge line segment inversion processing circuit, A to N ... edge line segment.

Claims (6)

[Claims] 1. An overlap change region in an input image is detected from an image signal obtained by differential binarization of a background image signal captured when an intruding object does not exist and an input image signal sequentially captured, In the intruder recognition device that recognizes the intruding object by analyzing this overlapping change area, the edge in the overlapping change area is detected from the input image,
The edge straight line in the overlap change area is detected from this edge, the effective edge line segment for recognizing the intruding object is extracted from the detected edge straight line, and many edges in the overlap change area are extracted from this effective edge line segment. An overlapping edge line segment extracting means for extracting the included edge line segment as an overlapping edge line segment, and an intruding object number recognizing means for recognizing the number of intruding objects from the overlapping change area divided by the overlapping edge line segment are provided. An intruding object recognition device. 2. The edge straight lines are vertical edges, horizontal edges, and
The intruding object recognition apparatus according to claim 1, wherein a histogram projection process is performed for each diagonal edge to obtain a statistic, and a straight line including many edges is drawn from the statistic to detect. 3. The intruding object recognition apparatus according to claim 1, wherein the edge straight line obtains a straight line in any of a vertical direction, a horizontal direction, and a diagonal direction by Hough transform. 4. The effective edge line segment is a line obtained by enclosing an overlapping change area with a circumscribed polygon and connecting either the intersection of the circumscribed polygon and the edge straight line or the intersection of the edge straight lines. The intruding object recognition device according to any one of claims 1 to 3, wherein the intruding object recognition device extracts the minute object. 5. The intruding object recognition device according to claim 1, wherein the overlapping edge line segment is extracted by connecting valid edge line segments. 6. The intruding object recognition device according to claim 1, wherein the intruding object is identified based on the shapes of the divided overlapping change regions.