JP5318664B2 - Object detection device - Google Patents

Description

  The present invention relates to an object detection apparatus that extracts a region having a characteristic of a detection object from a monitoring image and determines the presence of the detection object.

  In order to detect an intruder or a suspicious person, a human region that is a detection target is extracted from the monitoring image. For the extraction of this region, background difference processing, search processing by a classifier, or the like is used. At this time, a region other than the detection target object may be erroneously extracted, which causes a false alarm.

  In this regard, the moving object tracking device described in Patent Document 1 excludes the region from the processing target when the features of the shape such as the size and aspect ratio of the region extracted by the background difference processing are not suitable for monitoring. This is intended to reduce extraction errors.

  Moreover, in the target object detection apparatus described in Patent Literature 2, a window having the size of the target object to be detected is sequentially set in the input image, and scanning using a discriminator is performed to reduce erroneous extraction. .

JP 2008-250772 A JP 2005-157679 A

  However, in the prior art, it is not possible to exclude the erroneously extracted region, and there still remains a problem that the presence of the detection target is determined in the erroneously extracted region.

  For example, in the moving object tracking device described in Patent Document 1 that uses background difference processing, a partial area of the background that is swayed by a vegetation or illuminated by a headlight happens to have a human-like size and aspect ratio. In such a case, these areas are erroneously extracted. Since there are a wide variety of changes that occur in the background, it is difficult to eradicate false extractions simply by adjusting the dimensions and aspect ratio criteria.

  Further, for example, in the object detection device described in Patent Document 2 using a discriminator, this region is erroneously extracted when a partial region of the background happens to have a size and characteristics similar to a human being. Since backgrounds vary, it is difficult to eradicate erroneous extraction even if the learning data of the discriminator is increased.

  The present invention has been made to solve the above problems, and an object of the present invention is to provide an object detection apparatus that can detect the presence of an object with high accuracy by eliminating erroneous background extraction.

  An object detection device according to the present invention detects a detection object based on a determination target image, and extracts a feature area having features of the detection target from the determination target image. And an image dividing unit that divides the determination target image into a plurality of image fragments each having a pixel value having a predetermined similarity, and an indeterminate fragment extending between the inside and outside of the feature region is detected among the image fragments. The inner evaluation value that increases in accordance with the number of pixels inside the feature area in the uncertain fragment is obtained, and it is determined whether the detection target exists in the feature area using the inner evaluation value. Presence / absence determining means.

  In another object detection apparatus according to the present invention, the presence / absence determination unit obtains the inner evaluation value and the outer evaluation value that increases according to the number of pixels outside the feature region for each indeterminate fragment, The determination is made using the inner evaluation value for the uncertain fragment whose outer evaluation value is greater than the inner evaluation value.

  In the object detection device according to another aspect of the invention, the presence / absence determination unit includes the inner evaluation value for the uncertain fragment whose outer evaluation value is larger than the inner evaluation value, and the outer evaluation value is the inner evaluation value. The determination is made using the total value of the uncertain fragment and the outer evaluation value as follows.

  In the target object detection apparatus according to the present invention, the presence / absence determination unit adds the weights that increase in accordance with the distance between the pixels constituting the indeterminate fragment and the outline of the feature region, thereby calculating the inner evaluation value or The outer evaluation value may be calculated.

  A preferred aspect of the present invention includes a fragment articulation calculation unit that obtains a fragment intelligibility according to a predetermined difference between the pixel values inside and outside the indeterminate fragment, wherein the presence / absence determination unit is configured to perform the determination. The object detection device is characterized in that the inner evaluation value of the uncertain fragment is weighted by the fragment intelligibility of the uncertain fragment.

  According to the present invention, it is possible to determine the presence of an object with high accuracy by eliminating erroneous background extraction.

1 is a schematic block diagram of an image monitoring apparatus according to an embodiment of the present invention. It is a schematic diagram for demonstrating a fragment intelligibility calculation process. It is a schematic diagram for demonstrating the extraction correctness determination process of a presence determination means. It is a typical graph of the function F which defines the weight according to the distance of a pixel and the outline of a characteristic area. It is explanatory drawing explaining the change of the function F along the straight line which crosses a characteristic area | region. It is a schematic diagram for demonstrating the phenomenon which arises when a fragment | piece clarity is low. It is a schematic flowchart of an image monitoring process. It is a general | schematic flowchart of extraction correct / incorrect determination processing. It is a schematic diagram which shows the example of the process result of the extraction of the feature area in a monitoring image, and segmentation. It is a schematic diagram which shows the positional relationship of the misextracted characteristic area | region and an image fragment. It is a schematic diagram which shows the positional relationship of the characteristic area and image fragment which were correctly extracted.

  Hereinafter, an image monitoring apparatus 1 according to an embodiment of the present invention (hereinafter referred to as an embodiment) will be described with reference to the drawings. The image monitoring apparatus 1 detects an intruder by extracting a feature region having human characteristics as a detection target in a monitoring image obtained from the monitoring space. When an intruder is detected, the image monitoring apparatus 1 outputs an abnormal signal. FIG. 1 is a schematic block diagram of an image monitoring apparatus 1 according to the embodiment. The image monitoring apparatus 1 includes an image input unit 2, a storage unit 3, a control unit 4, and an output unit 5.

  The image input unit 2 is a monitoring camera, is installed in the monitoring space, images the monitoring space at a predetermined time interval, and displays a monitoring image (determination target image) in which each pixel is expressed by a multi-tone pixel value. Output. In this embodiment, the monitoring image is a color image. For example, the pixel value of each pixel is represented by a set of R, G, and B values of 256 gradations. Further, the monitoring image may be a grayscale image represented by a luminance value. In this case, the pixel value is expressed by a luminance value of 256 gradations, for example. In the image monitoring apparatus 1, the monitoring camera is configured to perform imaging under installation conditions (installation position, etc.) and imaging conditions (view angle, etc.) such that at least the detection target and the surrounding background are included in the monitoring image. Has been. More preferably, the monitoring camera is configured to perform imaging under installation conditions and imaging conditions such that the background area occupied in the monitoring image is equal to or larger than the area of the detection object when the detection object is imaged. . The captured monitoring images of the monitoring space are sequentially output to the control unit 4.

  The storage unit 3 is a storage device such as a ROM (Read Only Memory), a RAM (Random Access Memory), and a hard disk, and stores programs and data used by the control unit 4. The storage unit 3 inputs and outputs these programs and data to and from the control unit 4. The data stored in the storage unit 3 includes extraction information 30.

  The extraction information 30 is information for extracting a region of the object, and is information that is generated prior to the extraction process and stored in the storage unit 3. In the present embodiment in which the object is extracted by the background difference process, the extraction information 30 is specifically an image in which the object does not exist in the monitoring space, and the background image is based on the monitoring image prior to the extraction process. Is generated as In another embodiment in which a region of an object is extracted by a discriminator, a parameter representing an identification plane that separates the object from the other in the feature space is stored as the extraction information 30. The parameters representing the identification plane include the feature values (contrast, edge direction, etc.) of a large number of sample images in which the object is imaged and a large number of sample images in which the object is not imaged. It is learned beforehand by applying to.

  The control unit 4 is configured using an arithmetic device such as a DSP (Digital Signal Processor) or MCU (Micro Control Unit), and is connected to the image input unit 2, the storage unit 3, and the output unit 5. The control unit 4 reads out and executes the program from the storage unit 3 and executes various means described later (a feature region extraction unit 40, an image division unit 41, a fragment articulation calculation unit 42, an existence determination unit 43, and an abnormality detection unit 44). Function as.

  The feature region extraction unit 40 extracts a feature region having the feature of the detection target from the monitoring image by comparing the monitoring image with the extraction information 30, and outputs the extracted feature region information to the existence determination unit 43. To do. Specifically, the feature area extraction unit 40 uses the background image stored as the extraction information 30 to perform a difference process between the monitoring image and the background image, and collects pixels having a difference larger than the difference threshold. Are extracted as feature regions. In this case, an area having a pixel value different from that of the background is extracted as the feature area.

  In another embodiment using a discriminator, the feature area extraction unit 40 sequentially sets candidate areas having a predetermined size and a predetermined shape on the monitoring image, analyzes the image of the candidate area, and analyzes the feature amount (contrast, edge direction, etc.). ) Is extracted, and the extracted feature value is compared with the parameter of the identification surface stored as the extraction information 30 to derive an identification result (whether or not it is an object). The search in which the candidate area is set may be repeated a plurality of times by changing the size and shape of the candidate area. In this case, a region having a feature amount similar to the object is extracted as a feature region.

  The image dividing means 41 extracts a plurality of image fragments by grouping adjacent pixels having similar pixel values in the monitoring image. By this extraction process, the monitoring image is divided into a plurality of image fragments including pixels whose pixel values have a predetermined similarity. The extracted image fragment information is output to the fragment articulation calculation means 42 and the existence determination means 43. Such an image fragment is generally referred to as a segment, and the process of extracting a segment is generally referred to as segmentation. Since such image fragments are composed of adjacent pixels having similar pixel values, they are extracted for each road, building, person, and the like.

  Typical segmentation methods include the k-average method and the average shift method. The segmentation procedure based on these will be described.

[Segmentation based on k-means]
(Procedure 1) k pixels are arbitrarily selected from the monitoring image, and the pixel value of the selected pixel is set as a seed.
(Procedure 2) Each pixel in the monitoring image is associated with the most similar seed.
(Procedure 3) For each seed, the average pixel value of the associated pixel is calculated, and the amount of movement from the seed to the average pixel value is obtained.
(Procedure 4) If the movement amount is less than or equal to the preset threshold value ε, the process proceeds to the next procedure 5. Otherwise, the calculated average pixel value is set as a seed and the process returns to procedure 2.
(Procedure 5) Among the pixels associated with the same seed, adjacent pixels are grouped into one segment.

[Segmentation based on average shift method]
(Procedure 1) A local region having a predetermined size centering on each pixel in the monitoring image is initialized.
(Procedure 2) The density gradient vector of the local region is calculated, and shifting the local region in the vector direction (in the direction of increasing density) using the hill-climbing method is repeated until the shift amount converges.
(Procedure 3) A pixel value at the convergence point (meaning the most frequent pixel value) is set for each pixel in the monitoring image.
(Procedure 4) In the image generated by the procedure 3, pixels whose pixel value difference between adjacent pixels is equal to or less than a specified value are grouped into one segment.

As a reference for the mean shift method, Dorin Comaniciu, Peter Meer: Mean
Shift: A Robust Approach Toward Feature Space Analysis. IEEE Trans. Pattern
Anal. Mach. Intell. 24 (5): 603-619 (2002).

  As another segmentation method, an image can be regarded as a graph, and a minimum cut or normalized cut can be obtained.

  The fragment articulation calculation unit 42 calculates the fragment articulation degree according to a predetermined difference between the inside and outside pixel values of each image fragment extracted by the image dividing unit 41 and outputs it to the presence / absence determination unit 43. An image fragment that is significantly different from the surrounding image of the image fragment can be clearly distinguished from the surrounding image, and its contour is the contour of the target object, the contour of the target component constituting the target object, or the contour of the background constituent constituting the background. There is a high possibility that it matches the true contour. On the other hand, an image fragment with a small difference is highly likely to be an image fragment in which only a part of the target component or the background component is extracted due to the influence of a shadow or the like.

For example, the fragment clarity is the sum of differences (mutual differences) between each edge pixel existing at the edge of the target fragment and surrounding pixels existing outside the target fragment and adjacent to the edge pixel. Is calculated by Specifically, the pixel value c _i of the edge _pixel, expressed as a pixel value c _j of the surrounding pixels adjacent to each edge pixel, the calculation formula of fragments clarity L _s of the image fragments s is the following ( 1) It can be defined by the formula.
_{L s = G (1 / Σ} {α · exp (-‖c i -c j ‖ ^{2 / β)}) ··· (} 1)

In equation (1), G (x) is an increasing function that becomes 0 when x = 0 and becomes 1 when x = ∞, and this G (x) normalizes the range of fragment clarity L _s to [0, 1). It becomes. In this embodiment, the pixel values c _i and c _j are vectors composed of RGB values, and ‖x‖ represents the norm of the vector x. Α and β are constants set in advance. Σ represents the total sum for the set of pixel values c _i and c _j .

Here, since the number of edge pixels and the distribution of pixel values of edge pixels differ for each image fragment, calculating the fragment clarity only from the relationship between the edge pixels and surrounding pixels, the fragment clarity between image fragments There may be disparities in the reliability of Therefore, the fragment articulation calculation means 42 calculates the difference (self-difference) between the pixels in the target fragment, and normalizes the mutual dissimilarity based on the self-difference. The self-difference is calculated by summing the dissimilarities between adjacent pixels in the target fragment, and pixel values of two adjacent pixels in the image fragment s are expressed as c _h1 and c _h2 , respectively. A formula for calculating the normalized fragment clarity L _s of the image fragment s is given by equation (2).
_{L s = G (Σ {exp} (-‖c h1 -c h2 ‖ ^{2 / β)} / Σ {} exp (-‖c i -c j ‖ ^{2 / β)}) ··· (} 2)

In the equation (2), the first Σ represents the sum of the pixel values c _h1 and c _h2 , and the second Σ represents the sum of the pixel values c _i and c _j . By performing the normalization process in this way, it is possible to prevent a decrease in fragment clarity reliability caused by a difference between image fragments.

FIG. 2A is a schematic diagram of the target fragment s and the surrounding image, and shows the positions of the edge pixels (pixel value c _i ) in the target fragment s and the surrounding pixels (pixel value c _j ) outside the target fragment s. Has been. FIG. 2B shows an example of the function G, where the horizontal axis is the argument x and the vertical axis is the function value G (x).

  The presence / absence determination unit 43 determines whether or not a detection target exists in the feature region extracted by the feature region extraction unit 40, and outputs the determination result to the abnormality detection unit 44. Hereinafter, the presence / absence determination unit 43 will be described.

  When the feature region extraction result and the segmented image are superimposed, if the target image is perfectly extracted as the feature region, the image fragment group corresponding to the target image is a feature like a puzzle piece. It is expected to fit into the area. On the other hand, the lower the consistency between the shape of the feature region and the shape of the image fragment group in the portion corresponding to the feature region, the greater the possibility that the feature region is erroneously extracted. This is due to the fact that image fragments are extracted for each object or part thereof because adjacent pixels whose image fragments are similar to each other are grouped. In a state where the image fragment does not match the feature region, an image fragment that intersects the outline of the feature region is generated. Here, among the image fragments, those that straddle the inside and outside of the feature region and are uncertainly attributed to the feature region are referred to as uncertain fragments. For example, in a state where the background is included in the feature area due to erroneous extraction, image fragments belonging to the background straddle the inside and outside of the feature area. As the background included in the feature region increases, the degree to which individual uncertain fragments bite into the feature region increases or / and the uncertain fragments increase.

  The presence / absence determining means 43 calculates the region evaluation value by collecting the attribution information of the image fragments over the entire circumference of the feature region, and determines whether or not the feature region is correctly extracted based on the region evaluation value.

  For example, the presence / absence determination unit 43 detects an image fragment that is an indeterminate fragment, and calculates a region evaluation value according to the degree of overlap between the feature region and the uncertain fragment. The area evaluation value calculated here increases in accordance with the number of pixels inside the characteristic area of the uncertain fragment, and this will be referred to as an inner evaluation value. The presence / absence determination unit 43 determines whether or not the detection target exists in the feature region using the inner evaluation value as an extraction error of the feature region. Specifically, it can be determined that the feature region whose inner evaluation value is less than or equal to the preset reference value is correctly extracted, while the feature region whose inner evaluation value exceeds the reference value is erroneously extracted. . Accordingly, the presence / absence determination means 43 determines that there is an object in the feature region whose inner evaluation value is equal to or less than the reference value in response to the determination of whether the feature region is extracted correctly. It is determined that there is no object in the feature region that exceeds the set reference value.

  FIG. 3 is a schematic diagram for explaining the extraction correctness determination process of the presence / absence determination means 43. The processing will be specifically described with reference to FIG. FIG. 3 shows image fragments 81 to 84 having different positional relationships with respect to the feature region 80 as examples of image fragments extracted from the monitoring image. Since the image fragments 81 and 82 include the outline of the feature region 80 on the inside, they are detected as uncertain fragments. On the other hand, since the image fragments 83 and 84 do not include the outline of the feature region 80 on the inside, they are not detected as uncertain fragments. In FIG. 3, the hatched portion with the horizontal line is an overlapping portion with the feature region 80 in the uncertain fragment, while the shaded portion with the vertical line is a portion protruding from the feature region 80 in the uncertain fragment. .

As described above, the uncertain fragment may be a background image fragment, and in this case, the overlapping portion of the uncertain fragment and the feature region is a portion in which the background is erroneously extracted as a part of the feature region. That is, it can be said that the larger the overlapping portion, the higher the possibility of erroneous extraction, and the smaller the overlapping portion, the lower the possibility. Therefore, the presence / absence determination unit 43 calculates the area (number of pixels) of the overlapping portion as the inner evaluation value for each indeterminate fragment, and obtains the total value as the region evaluation value. When the identification symbol of the uncertain fragment is represented by _s and the inner evaluation value of the uncertain fragment is represented by g _s , the calculation formula of the region evaluation value J is expressed by the following equation (3).
J = Σg _s (3)

  In the equation (3), Σ represents the total sum for the uncertain fragment s of the feature region.

  The presence / absence determination means 43 compares the region evaluation value J obtained by the equation (3) with the reference value T, and determines that there is an object in the feature region below the reference value T as described above. It is determined that there is no object in the feature region that is not.

  The inner evaluation value may be weighted according to the distance from the outline of the feature region as follows instead of a simple area, and this configuration will be described below. Regardless of what extraction method is used in the feature region extraction means 40, the region of the object is rarely extracted completely. When viewed in units of pixels, there is high uncertainty in the attribution of pixels in the vicinity of the outline of the feature region (whether the pixel belongs to the object or the background). Conversely, as the distance from the contour increases, the uncertainty of attribution decreases, and the closer to the center of the feature region, the higher the probability of belonging to the object. Therefore, the presence / absence determining unit 43 can calculate a weight corresponding to the distance from the contour of the feature region to the pixel for each pixel in the overlapping portion, and obtain the sum of the weights as the inner evaluation value.

Specifically, when the overlapping pixel is e1 and the distance from the feature region outline to the pixel e1 is d _e1 , the inner evaluation value g _s is calculated by the following equation (4).
g _s = ΣF (d _e1 ) (4)

In the equation (4), Σ represents the total sum for the pixel e1. F (x) is a predetermined function that defines a weight according to the distance _de1 input as the argument x, and has a range of (0, 1) for x> 0 and an increasing function. It is. FIG. 4 shows an example of the function F, where the horizontal axis is the argument x and the vertical axis is the function value F (x).

Region evaluation value J can be obtained by (4) using the inner evaluation value g _s for each uncertain fragment given by formula (3) below.

  Further, as described below, the region evaluation value J can be determined in consideration of the protruding portion of the uncertain fragment. In this case, the presence / absence determination unit 43 obtains an evaluation value that increases in accordance with the number of pixels of the protruding portion for each indeterminate fragment. The evaluation value is referred to as an outer evaluation value.

  The uncertain fragment includes a portion where the protruding portion is larger than the overlapping portion like the uncertain fragment 81 and a portion where the protruding portion is smaller than the overlapping portion like the uncertain fragment 82. Indeterminate fragments with larger overhangs are more likely to be image fragments that originally belong to the background, while uncertain fragments with larger overlap portions are more likely to be image fragments that originally belong to the object. . Therefore, the region evaluation value J is characterized by an uncertain fragment that is highly likely to belong to the background by calculating the region evaluation value J by summing the inner evaluation values only for the uncertain fragments whose overlapping degree is smaller than the protruding degree. The influence on the region is preferably expressed, the reliability of the region evaluation value J as an extraction error is improved, and the accuracy of the presence / absence determination is improved.

Specifically, as shown in Equation (5), the presence / absence determination unit 43 compares the inner evaluation value and the outer evaluation value for each indeterminate fragment, and determines the uncertain fragment whose inner evaluation value is smaller than the outer evaluation value. The total value of the inner evaluation values is defined as a region evaluation value J.
J = Σg _s {s; g _s <n _s } (5)

In (5), an inner evaluation value _{g s} uncertain fragments s, a outer evaluation value is _{n s,} sigma represents the sum of _g s _{<n s} becomes uncertain fragment s.

The outer evaluation value n _s is preferably defined so as to be comparable with the inner evaluation value g _s . That is, when the inner evaluation value g _s is defined by the area of the overlapping portion, the outer evaluation value n _s is defined by the area of the protruding portion. On the other hand, when the inner evaluation value g _{s is obtained} by the equation (4), the outer evaluation value n _s is also obtained by the following equation using the function F similar to the equation (4).
n _s = ΣF (d _e2 ) (6)

In equation (6), Σ is the sum of the protruding portion pixel e2, and d _e2 is the distance from the contour of the feature region to the pixel e2.

(6) on the outside evaluation value n _s of formula, the pixels of the contour near the feature region high uncertainty attribution, accuracy attributable to the background as the pixels away from the feature region conversely to reflect the nature of high The FIG. 5 schematically shows a change in the value of the function F in accordance with the distance from the contour of the feature region 80. The feature region 80 is shown on the upper side of the figure, and the function along a straight line 86 across the feature region 80 is shown. The graph of F is shown on the lower side of the figure.

Note that the degree of overlap and the degree of protrusion are compared by comparing the area of the overlapping part and the area of the protruding part, and the inner evaluation value g _s used for calculating the region evaluation value J uses the result of equation (4). Presence / absence determination means 43 may be configured.

  The region evaluation value J described so far is an extraction error evaluated from the viewpoint that the feature region extraction means 40 erroneously extracted the background as a part of the feature region. Now, another aspect of erroneous extraction by the feature region extraction means 40 is a phenomenon that fails to extract a part of the object. If this extraction failure factor is taken into consideration, the reliability of the region evaluation value J is further improved. Therefore, the presence / absence determination means 43 may be configured to add an outer evaluation value for an indeterminate fragment whose degree of protrusion is equal to or less than the overlapping degree to the area evaluation value J of equation (5).

In this case, the formula for calculating the region evaluation value J is given by the following formula (7).
_{J = Σ {min (g s} , n s)} ··· (7)

  In the equation (7), Σ represents the total sum for the uncertain fragment s of the feature region. As a result, an area evaluation value J which is an extraction error evaluated from both the viewpoint that the feature area includes the background and the viewpoint that the feature area fails to extract the object is obtained. Reliability is improved, and as a result, the accuracy of existence determination is improved.

(7) as the inner evaluation value g _s and outer evaluation value n _s of formula, taking into account the weights according to the distance from the contour of the feature region (4) may be a value of (6). As described above, since the function F for weighting the pixels according to the distance represents the accuracy of the attribution of the pixels, the reliability of the region evaluation value J can be obtained by using the values of the expressions (4) and (6). This improves the accuracy of the presence / absence determination. Note that the inner evaluation value g _s and the outer evaluation value n _s in the expression (7) can also be the areas of the overlapping portion and the protruding portion, respectively.

Furthermore, in order to increase the accuracy of presence / absence determination, the fragment articulation obtained by the fragment articulation calculation means 42 can be taken into consideration. This configuration will be described below. As described in the description of the fragment articulation calculation means 42, only a part of the image of the target component or the background component may be inappropriately extracted as an image fragment due to the influence of a shadow or the like. In such inappropriate image fragments, the magnitude relationship between the degree of overlap and the degree of protrusion may be reversed from the original relationship. FIG. 6 is a schematic diagram for explaining a phenomenon that occurs when the fragment articulation is low. FIG. 6A shows a case where the fragment intelligibility of the indeterminate fragment 85-1 for the feature region 80 is high. As described above, it is highly likely that the uncertain fragment 85-1 having a high fragment clarity as described above suitably represents the image of the target component or the background component. Since the indeterminate fragment 85-1 shown as an example shown in FIG. 6A has a smaller inner evaluation value (the size of the shaded portion of the horizontal line) than the outer evaluation value, the region evaluation value J according to equation (7) is The inner evaluation value g _s is added in the calculated summation.

On the other hand, FIG. 6B shows a case where the fragment articulation of the indeterminate fragment 85-2 for the feature region 80 is low. The indeterminate fragment 85-2 shows a case where only a part of the target component or background component whose original image has a shape like the image fragment 85-1 is extracted. It is recognized that the part of 1 that belongs to the outside of the feature region 80 (the part indicated by the dotted line) constitutes another image fragment. In this case, would towards the outer evaluation value (the magnitude of the vertical line shaded area) becomes smaller than the inner evaluation value at summation to obtain the area evaluation value J by equation (7), the outer evaluation value n _s Is added. That is, the uncertain fragment 85-2 that should be evaluated as an image fragment that originally belongs to the background is erroneously evaluated as an image fragment that belongs to the object. In order to reduce the influence of such errors on the region evaluation value J, when calculating the region evaluation value J, the presence / absence determination means 43 uses the inner evaluation value and the outer evaluation value of the uncertain fragment s as the fragment of the uncertain fragment. Weighted with clarity L _s .

Specifically, the region evaluation value J corresponding to the equation (7) is calculated by the following equation in the configuration considering the fragment articulation.
_{J = Σ {L s · min} (g s, n s)} ··· (8)

  As a result, since the region evaluation value J can be calculated by applying a high weight to an uncertain fragment that has high fragment articulation and can be trusted, the reliability of the region evaluation value J is improved, and the accuracy of correctness determination is improved.

  The weighting based on the fragment articulation can also be introduced in the calculation of the region evaluation value J in the expressions (3) and (5).

  The existence determination unit 43 has been described above. The determination result of the presence / absence determination unit 43 is output to the abnormality detection unit 44.

  The abnormality detection unit 44 detects an abnormality based on the information on the feature region determined as correct extraction by the presence / absence determination unit 43. Specifically, the abnormality detection unit 44 analyzes the movement of the object by tracking the feature region determined to be correctly extracted over a plurality of times, and, for example, a movement of a predetermined distance or more is confirmed in the monitoring space. Detecting an intrusion abnormality when Further, in another embodiment for detecting an abnormality in which a suspicious person exists, similarly, an abnormality is detected when the movement is analyzed and a stay for a predetermined time or more is confirmed, or an abnormality is detected when the movement pattern matches a predetermined suspicious pattern. Is detected.

  Since the abnormality detection unit 44 detects an abnormality based on the information of the feature region from which the erroneous determination is excluded by the presence / absence determination unit 43, it is possible to reliably reduce the false alarm. When the abnormality is determined, the control unit 4 controls the output unit 5 to output an abnormality signal.

  The output unit 5 is an interface circuit that is connected to an external device and outputs an abnormal signal to the external device. The external device is an alarm display means such as a speaker, a buzzer or a lamp for alarming the presence of an intruder, and / or a remote center device connected via a communication network.

  Next, the operation of the image monitoring apparatus 1 will be described with reference to FIG. When an administrator who confirms that the monitoring space is unmanned turns on the apparatus, each unit and each means are initialized and start operating (S1). After the initialization, every time a new monitoring image is input from the image input unit 2 to the control unit 4, the processes in steps S2 to S10 are repeated as a loop process.

  The image monitoring apparatus 1 updates the background image used in the background difference process of the feature region extraction unit 40 every time the loop process ends (S2). The control unit 4 combines the image of the portion of the monitoring image in the previous loop process from which the feature region has not been extracted with the current background image, and updates the background image (S2). Note that the background image is not yet stored immediately after the power is turned on, that is, at the start of the first steps S3 to S10. In this case, the control unit 4 initializes the background image with the monitoring image obtained from the image input unit 2 as background image update processing (S2).

  When a new monitoring image is input (S3), the feature region extraction unit 40 performs background difference processing using the updated background image, and extracts a feature region from the monitoring image (S4).

  Incidentally, in the device configuration in which the feature region is extracted by the discriminator, the update process in step S2 is not necessary.

  If no feature region has been extracted (NO in S5), the process returns to step S2. On the other hand, when one or more feature regions are extracted (YES in S5), control unit 4 determines whether or not an object exists in each feature region (S6).

Details of the presence / absence determination process will be described with reference to FIG. First, the image dividing unit 41 of the control unit 4 performs segmentation processing on the monitoring image to extract a plurality of image fragments from the monitoring image (S600), and the fragment articulation calculation unit 42 of the control unit 4 extracts each extracted image. By applying the pixel values of the fragment and the surrounding image to the equation (2), the fragment clarity L _s of each image fragment is calculated (S601).

  Next, the presence / absence determining unit 43 of the control unit 4 sequentially sets the feature region extracted in step S6 in FIG. 7 as the feature region of interest and executes a loop process of the feature region (S602 to S612). In the feature region loop processing, the presence / absence determining means 43 further sets the image fragments extracted in step S600 as the target fragments in sequence, and executes the image fragment loop processing (S603 to S607).

In the loop processing of the image fragment, the presence / absence determination unit 43 determines the positional relationship between the target fragment and the target feature region (S604), and if the target fragment is determined to be an indeterminate fragment straddling the outline of the target feature region (S605). Te YES), (4) to calculate the inner evaluation value _{g s} of the image fragments by equation (6) by calculating the outer evaluation value _{n s} of the image fragment (S606).

When the loop processing of the image fragment is completed (YES in S607), the presence / absence determination unit 43 determines the fragment clarity L _s calculated in Step S601, the overlap degree g _s calculated in Step S606, and the protrusion degree n _s. Is applied to the equation (8) to calculate the region evaluation value of the feature region of interest (S608).

  After calculating the region evaluation value, the presence / absence determining unit 43 compares the region evaluation value of the feature region of interest with a preset reference value (S609). If the area evaluation value is equal to or less than the reference value (YES in S609), presence / absence determination means 43 determines that the target object exists in the feature area of interest (S610). On the other hand, if the region evaluation value exceeds the reference value (NO in S609), it is determined that no object exists in the target feature region (S611). Information of the feature area determined that the object does not exist is deleted from the temporarily stored information.

  Note that when comparing the region evaluation value and the reference value, it is preferable to scale either the region evaluation value or the reference value according to an index Ω representing the size of the feature region. For example, when the reference value is constant without depending on Ω, scaling is performed by dividing the region evaluation value J obtained by Equation (8) by Ω. For example, as Ω, an area obtained by summing the number of pixels in the feature region, a value obtained by summing the above-described weighting function F for all the pixels in the feature region, or the like can be used. On the other hand, when the size of the image of the detection target in the monitoring image can be assumed in advance, the scaling can be omitted. In addition, the reference value in the case of performing scaling can be set to a size of about 20 to 40% of the size Ω of each feature region, for example.

  If the presence or absence of an object is determined for all feature regions by the above processing (YES in S612), the processing proceeds to step S7 in FIG.

  The continuation of the image monitoring process will be described with reference to FIG. 7 again. When there is a feature region in which the presence of the target object is determined by the presence determination unit 43 (YES in S7), the abnormality detection unit 44 of the control unit 4 determines the presence / absence of an abnormality based on the feature region (S8). That is, the abnormality detection unit 44 performs tracking of the feature region, and when a movement of a distance exceeding a preset distance threshold is detected as a result of the tracking, the abnormality detection unit 44 detects an abnormality as the feature region is caused by an intruder. . For example, the tracking is performed by analyzing the color histogram of the partial image corresponding to the feature region in the monitoring image at the current time, and comparing the color histogram with a color histogram that has been analyzed and stored in the past in the same manner. This is performed by associating the feature area of the time to be performed. In preparation for tracking the next time, the abnormality detection unit 44 additionally stores in the storage unit 3 a color histogram analyzed from the monitoring image at the current time.

  When an abnormality is detected by the abnormality detection means 44 (YES in S9), the output unit 5 generates an abnormality signal and outputs the abnormality signal to, for example, an alarm display means and a communication network (S10). At this time, the output unit 5 includes the monitoring image data at the current time in the abnormality signal output to the communication network. The alarm display means to which the abnormal signal is input displays an alarm by sounding a speaker or lighting a lamp to threaten an intruder and alert the user. Also, the abnormal signal is transmitted to the center device of the security center where the monitoring staff is stationed by the communication network, and the center device that has received the abnormal signal sounds an alarm sound or displays a monitoring image. Then, the monitoring person who confirms the situation by the monitoring image takes measures according to the situation.

  On the other hand, if there is no feature region for which the presence / absence determination unit 43 has determined the presence of the object (NO in S7), the processing after step S8 is skipped. In addition, if there is still no abnormality determined even if there is a feature region in which the presence of the target object is determined (NO in S9), the process in step S10 is skipped.

  When the above processing ends, the processing returns to step S2 again, and processing for the monitoring image at the next time is performed.

  FIG. 9 is a schematic image example for explaining the processing result. FIG. 9A shows a monitoring image 90 from which feature areas 91 and 92 are extracted. The feature area 91 is obtained by erroneously extracting a swaying tree. On the other hand, intruders are extracted in the feature region 92. FIG. 9B shows an image 93 generated by performing a segmentation process on the monitoring image 90. In the image 93, each solid line closed region is an image fragment. In FIG. 9B, the outlines of the feature regions 91 and 92 are shown by dotted lines for assisting understanding.

  FIG. 10 is an enlarged schematic view of the feature area 91 and its vicinity area, while FIG. 11 is an enlarged schematic view of the feature area 92 and its vicinity area. In FIG. 10, two image fragments having a portion overlapping the feature region 91 indicated by the dotted line are both uncertain fragments. Both of these two uncertain fragments protrude greatly outside the feature region 91, and the protruding portion is larger than the overlapping portion. Therefore, the region evaluation value is given by the sum of the inner evaluation values of the uncertain fragments. In addition, since the overlapping portion with the uncertain fragment occupies the entire feature region, the region evaluation value exceeds the reference value in the comparison process S609. As a result, it is determined that no object exists in the feature area 91.

  On the other hand, in FIG. 11, there are three image fragments having a portion overlapping with the feature region 92 indicated by the dotted line. Two of them are uncertain fragments that protrude greatly outside the feature region 92, and they have a feature region 92 and overlapping portions 96 and 97. The remaining one uncertain fragment is an uncertain fragment whose most part exists inside the feature region 92, and the uncertain fragment has a protruding portion 98 from the feature region 92. In this case, the region evaluation value is given by the sum of the inner evaluation value calculated from the overlapping portions 96 and 97 and the outer evaluation value calculated from the protruding portion 98, and the region evaluation value is equal to or less than the reference value. As a result, it is determined that the person who is the object exists in the feature region 92.

  In the above-described embodiment, the configuration in which the image input unit 2 includes the monitoring camera and detects the object based on the monitoring image obtained from the monitoring space in real time has been described. In another embodiment, the image input unit 2 may be an image recording device such as an HDD recorder, and the recorded monitoring image may be input to the control unit 4. In this configuration, the image monitoring device 1 detects an object from monitoring images captured in the past.

  In the above-described embodiment, false alarms are reduced by performing an extraction correctness determination on all extracted feature areas prior to the abnormality detection process. On the other hand, when an abnormality detection process is performed first, an extraction correct / incorrect determination is performed on a feature area in which an abnormality is detected, and it is determined that a feature area in which an abnormality has been detected is correctly extracted, an output unit 5 can also be configured to output an abnormal signal.

In the above-described embodiment, the region evaluation value J represents a high possibility of extraction error, that is, an extraction error. Conversely, the region evaluation value J has a low possibility of extraction error. In other words, it can be defined as an index representing the accuracy of extraction. For example, it is possible to determine the sum of the reciprocals of the inner evaluation value g _s, or the sum of the inverse of and outside evaluation value n _s of the inner evaluation value g _s, a regional assessment value J indicating the accuracy of the extraction. In this case, the definition formula of the region evaluation value J corresponding to (3) is the following formula (9), and the definition formula corresponding to the formula (8) is the following formula (10).
J = Σ (1 / g _s ) (9)
_{J = Σ {L s / min} (g s, n s)} ··· (10)

  Note that when the region evaluation value J of the equations (9) and (10) is used, the magnitude relationship between the region evaluation value J and the reference value T is reversed, and the feature region in which the region evaluation value J is greater than or equal to the reference value T. It is determined that there is no target object, and it is determined that there is no target object in the feature region that is not.

  DESCRIPTION OF SYMBOLS 1 Image monitoring apparatus, 2 Image input part, 3 Memory | storage part, 4 Control part, 5 Output part, 30 Information for extraction, 40 Feature area extraction means, 41 Image division means, 42 Fragment articulation calculation means, 43 Existence determination means, 44 anomaly detection means, 80 feature areas, 81, 82 indeterminate fragments.

Claims (5)

In the object detection device that detects the detection object based on the determination target image,
Feature region extraction means for extracting a feature region having the characteristics of the detection object from the determination target image;
Image dividing means for dividing the determination target image into a plurality of image fragments each having a pixel value having a predetermined similarity;
An indeterminate fragment straddling the inside and outside of the feature region is detected from among the image fragments, an inner evaluation value that increases according to the number of pixels inside the feature region is determined from the indeterminate fragment, and the inner evaluation value is and determining existence judgment unit and the sense target to the feature region equal to or less than a predetermined reference value exists,
The object detection apparatus characterized by having. In the object detection device that detects the detection object based on the determination target image,
Feature region extraction means for extracting a feature region having the characteristics of the detection object from the determination target image;
Image dividing means for dividing the determination target image into a plurality of image fragments each having a pixel value having a predetermined similarity;
An indeterminate fragment straddling the inside and outside of the feature region is detected in the image fragment, and an inner evaluation value that increases according to the number of pixels inside the feature region is determined from the indeterminate fragment, and the inner evaluation value is Presence / absence determination means for determining whether or not the detection object exists in the feature region,
Have
The presence / absence determining means obtains the inner evaluation value and an outer evaluation value that increases according to the number of pixels outside the feature region for each indeterminate fragment, and the outer evaluation value is larger than the inner evaluation value. An object detection apparatus comprising: determining using the inner evaluation value for an uncertain fragment. The object detection device according to claim 2,
The presence / absence determining means includes the inner evaluation value for the uncertain fragment whose outer evaluation value is greater than the inner evaluation value, and the outer evaluation for the uncertain fragment whose outer evaluation value is equal to or less than the inner evaluation value. An object detection device characterized by determining using a total value with a value. In the object detection device according to claim 2 or claim 3,
The presence / absence determining means calculates the inner evaluation value or the outer evaluation value by summing weights that increase in accordance with a distance between a pixel constituting the uncertain fragment and a contour of the feature region. An object detection device. In the target object detection device according to any one of claims 1 to 4,
Fragment intelligibility calculation means for obtaining a fragment intelligibility according to a predetermined difference between the pixel values inside and outside the indeterminate fragment,
In the determination, the presence / absence determination unit weights the inner evaluation value of the uncertain fragment with the fragment intelligibility of the uncertain fragment.

Images