JP2012128508A - Invasion monitoring device and invasion monitoring method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an invasion monitoring device capable of correctly detecting camera obstruction while avoiding erroneous detection of congestion of passersby and vehicles taken by a camera as an event of monitoring obstruction to the camera.SOLUTION: An invasion monitoring device includes a moving object area estimation part 23 that measures the number of detection per unit time for each small area to estimate density of passersby based on the detection result made by a moving object detection part 22, and a feature amount effective area selection part 24 that selects an area where many robust feature points over a brightness variation are found based on a reference image generated by a reference image generation part 21. The invasion monitoring device is configured to dynamically select an obstruction detection determination area to an imaging part 1. The invasion monitoring device correctly detects camera obstruction while avoiding erroneous detection of congestion of passersby and vehicles taken by a camera as an event of monitoring obstruction to the camera.

Description

  The present invention relates to an intrusion monitoring apparatus and an intrusion monitoring method for detecting a moving object from a photographed image and at the same time detecting interference with a camera image.

  As a conventional proposal for intrusion monitoring, a “camera disturbance detection method” described in Patent Document 1 is known. The camera tampering detection method described in Patent Document 1 performs luminance difference processing between an image obtained from a captured image and a reference image that is generated in advance and does not show a moving object, and performs a difference value in pixel units. Measures the number of pixels greater than or equal to the first threshold, and detects that the number of pixels is greater than or equal to the second threshold as a disturbance to the camera.

Japanese Patent Laid-Open No. 2008-077517

  However, in the above-described conventional technology, camera jamming is detected by using only the amount of pixels having a certain difference between the input image and the reference image, so that congestion of people, camera aperture control, etc. Therefore, even when a luminance change occurs in most of the camera video, there is a problem that the threshold value for determining that the camera is disturbed is exceeded and erroneous determination is made. For example, FIG. 4 is a diagram illustrating a problem of the related art. As shown in the figure, since a large number of persons 101 and vehicles 102 exist in the current image 100 that is an input image, a luminance change occurs in most of the camera video. For this reason, in the comparison with the reference image 110, the threshold value for determining that the camera is disturbed may be exceeded, resulting in an erroneous determination.

  The present invention has been made in view of such circumstances, and intrusion monitoring capable of accurately detecting camera disturbance without erroneously detecting congestion of a person or vehicle reflected in the camera as an event such as monitoring disturbance to the camera. An object is to provide a device and an intrusion monitoring method.

  An intrusion monitoring apparatus according to the present invention includes an imaging unit that captures an area to be monitored, a reference image generation unit that generates a reference image in which no moving object is reflected from image data obtained from the imaging unit, and the reference image generation unit A moving object detection unit that detects a region where an object exists by comparing the reference image generated in step 1 and the image obtained from the imaging unit, and a unit for each small region based on the detection result of the moving object detection unit. There are many feature points that are robust to fluctuations in brightness based on the reference image generated by the reference image generation unit and the moving object existence region estimation unit that measures the number of detections per hour and estimates the degree of traffic. A priority is given to the determination area when performing camera tampering detection based on the estimation result of the feature quantity effective area selection section for selecting the area, the estimation result of the moving object existence area estimation section, and the selection result of the feature quantity effective area selection section. Judgment area priority to select A matching evaluation value is obtained by matching the image obtained from the imaging unit with the reference image generated by the reference image generation unit in the interference detection determination region selected by the setting unit and the determination region priority setting unit. And a camera tampering detection unit that determines whether or not the camera tampers with the camera.

  According to the above configuration, it is possible to estimate a region where a moving object can exist, and to select an effective region having a feature amount suitable for matching processing between an image obtained by the imaging unit and a reference image. Furthermore, by using both types of information, it is possible to select the priority of the region suitable for matching, so it is possible to avoid the effects of busy alleys and disturbances (regions where the leaves are constantly shaking). It can be reduced, and the camera tampering detection can be performed with high accuracy.

  In the above configuration, the moving object existence region estimation unit divides the image into blocks, and calculates and updates the existence probability for the block to which the moving locus coordinates of the moving object that are detected periodically belong.

  According to the above configuration, it is possible to further reduce the influence of misinformation and sunshine fluctuations in a downtown area where the amount of traffic changes every moment.

  In the above-described configuration, the moving object existence region estimation unit calculates, records, and updates the existence probability for each time zone in units of blocks.

  According to the above configuration, it is possible to further reduce the influence of misinformation or sunshine fluctuations in a busy street where the traffic is intense depending on the time of day.

  In the above configuration, when the calculated existence probability is uniform within the target area, the moving object existence area estimation unit randomly or fixedly selects a block in which the moving object is not detected within the target area.

  According to the above configuration, it is possible to reduce the rate of erroneous detection that camera shake has occurred as the moving object moves.

  In the above configuration, the feature quantity effective region selection unit calculates a feature quantity in the reference image for each block, and measures the number and evaluation value.

  According to the said structure, the misdetection resulting from the fluctuation | variation of a luminosity when it irradiates to a sunlight fluctuation | variation and the headlight of a vehicle can be suppressed.

  In the above configuration, the feature quantity effective region selection unit excludes a region located in the sky on the image from being calculated.

  According to the above configuration, it is possible to reduce an excessive calculation processing load and to improve matching accuracy later.

  In the above configuration, the determination area priority setting unit determines the priority of the area for performing the matching process in consideration of the probability that the moving object does not exist and the effective feature amount area with respect to the area where the moving object does not exist. To do.

  According to the above configuration, a region that is easily influenced by the movement of a moving object or a region that is not suitable for matching can be excluded from the matching target, and the accuracy of camera tampering detection can be improved.

  In the above configuration, the camera tampering detection unit performs a matching process using the feature amount used by the feature amount effective region selection unit in the determination region having a high priority selected by the determination region priority setting unit.

  According to the above configuration, even when there are few feature values that are robust to changes in brightness, matching processing can be executed, and detection omission can be suppressed.

  In the above-described configuration, the camera tampering detection unit dynamically changes the block to be matched according to the priority of changing every frame in the matching process.

  According to the above configuration, the camera tampering detection process can be accurately performed according to the camera installation environment that changes every moment.

  The intrusion monitoring method of the present invention includes a step of generating a reference image in which a moving object is not shown from image data obtained from an imaging unit that images a monitoring target region, the reference image, and an image obtained from the imaging unit. Detecting a region where an object is present by the comparison process, measuring the number of detections per unit time for each small region based on the moving object detection result, estimating the degree of traffic, and the reference image Priority is given to the judgment area when performing camera tampering detection based on the step of selecting areas where there are many feature points that are robust to brightness fluctuations based on the results of estimating the existence area of the moving object and selecting the effective area of the feature amount In the interference detection determination area in which the priority of the determination area is selected, the image obtained from the imaging unit is matched with the reference image, and the matching evaluation value is determined. And determining whether the camera disturbance Te equipped with.

  According to the above method, it is possible to estimate a region where a moving object can exist, and to select an effective region having a feature amount suitable for matching processing between an image obtained by the imaging unit and a reference image. Furthermore, by using both types of information, it is possible to select the priority of the region suitable for matching, so it is possible to avoid the effects of busy alleys and disturbances (regions where the leaves are constantly shaking). It can be reduced, and the camera tampering detection can be performed with high accuracy.

  According to the present invention, it is possible to accurately detect camera interference without erroneously detecting congestion of a person or vehicle reflected in the camera as an event such as monitoring interference with the camera.

The block diagram which shows schematic structure of the intrusion monitoring apparatus which concerns on one embodiment of this invention The figure for demonstrating operation | movement of the intrusion monitoring apparatus of FIG. The flowchart for demonstrating operation | movement of the intrusion monitoring apparatus of FIG. Diagram for explaining conventional intrusion monitoring technology

  DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments for carrying out the invention will be described in detail with reference to the drawings.

  FIG. 1 is a block diagram showing a schematic configuration of an intrusion monitoring apparatus according to an embodiment of the present invention. In the figure, the intrusion monitoring apparatus 2 of the present embodiment includes a reference image generation unit 21, a moving object detection unit 22, a moving object existence region estimation unit 23, a feature quantity effective region selection unit 24, and a determination region priority. A degree setting unit 25 and a camera disturbance detection unit 26 are provided.

  The reference image generation unit 21 generates a reference image that does not show a moving object from image data obtained from an imaging unit (camera) 1 that captures a monitoring target region. The moving object detection unit 22 detects a region where an object exists by comparing the reference image generated by the reference image generation unit 21 and the image obtained from the imaging unit 1. The moving object existence region estimation unit 23 measures the number of detections per unit time for each small region based on the detection result of the moving object detection unit 22, and estimates the amount of traffic. That is, the moving object existence region estimation unit 23 divides the image into blocks of about 8 × 8 pixels and periodically (for example, every frame) based on the position data where the object is output from the moving object detection unit 22. The existence probability is calculated and updated for the block to which the moving locus coordinates of the moving object (for example, the center lower end point of the detection region) belong. By doing so, it is possible to further reduce the influence of misinformation and sunshine fluctuations in a busy street where the amount of traffic changes every moment. Further, the existence probability is learned (calculated, recorded, and updated) for each time slot in units of blocks. By doing so, it is possible to further reduce the influence of misinformation or sunshine fluctuations in a busy street where the traffic is intense depending on the time of day. Furthermore, when the existence probability is uniform within the target region, the block is selected at random (or fixed) from the blocks in which the moving object is not detected within the target region. By doing so, it is possible to reduce the rate of erroneous detection that camera shake has occurred as the moving object moves.

  The feature quantity effective region selection unit 24 selects a region where there are many feature points that are robust against brightness fluctuations based on the reference image generated by the reference image generation unit 21. That is, in the reference image generated by the reference image generation unit 21, the feature amount is calculated for each block of about 8 × 8 pixels, the number and evaluation value thereof are measured, the number of feature amounts is large, and the evaluation value is high. The region is selected as a feature amount effective region. Here, feature values are robust to color information and brightness changes (SIFT: Scale Invariant Feature Transform, HOG: Histograms of Oriented Gradients, RRC: Radial Reach Correlation, etc.) It is possible to suppress erroneous detection due to luminance fluctuation when the headlight is irradiated. In addition, a region positioned in the sky on the image is excluded from the calculation target, and an unnecessary calculation processing load is reduced, which leads to an improvement in the matching accuracy later.

  The determination region priority setting unit 25 gives priority to the determination region when performing camera tampering detection based on the estimation result of the moving object existence region estimation unit 23 and the selection result of the feature amount effective region selection unit 24. To do. The determination region priority setting unit 25 determines the priority of the region to be subjected to the matching process in consideration of the probability that the moving object does not exist and the effective feature amount region with respect to the region where the moving object does not exist. The priority is set high in areas where people do not pass (low existence probability) and features are high (areas where the effectiveness of feature quantities is high), and there are many people (high existence probability) and few features. Set low in the region (region where the effectiveness of the feature amount is low). In this way, by setting the determination region priority, it is possible to exclude regions that are easily influenced by the movement of the moving object or regions that are not suitable for matching from the matching target, and improve the accuracy of camera tampering detection. it can.

  The camera tampering detection unit 26 performs matching between the image obtained from the imaging unit 1 and the reference image generated by the reference image generation unit 21 in the tampering detection determination region selected by the determination region priority setting unit 25. It is determined whether or not the camera is obstructed according to the evaluation value. In this case, the matching process using the feature amount used by the feature amount effective region selecting unit 24 is performed in the high priority block selected by the determination region priority setting unit 25. However, if a block with a small amount of features is selected with high priority, block matching using luminance information and color information is performed, and matching processing is executed even when there are few features that are robust to changes in brightness. And detection omission can be suppressed. In addition, the blocks to be matched can be detected with high accuracy according to the camera installation environment that changes every moment by dynamically changing the blocks according to the priority that changes every frame. The camera disturbance detection unit 26 outputs a camera disturbance detection signal when it is determined that the camera disturbance has occurred.

  Next, the operation of the intrusion monitoring apparatus 2 of the present embodiment will be described with reference to the operation explanatory diagram shown in FIG. 2 and the flowchart shown in FIG.

  First, the reference image generation unit 21 inputs the image data obtained by the imaging unit 1 (step S1), and then learns / updates the input image data by brightness value, color smoothing, etc. An image (that is, the reference image 110) in which the person 101 and the vehicle 102 are not reflected is generated (step S2). Then, the reference image data is input to the moving object detection unit 22 together with the image data obtained by the imaging unit 1 (that is, the current image 100).

  Next, the moving object detection unit 22 detects the moving object existing in the current image 100 by comparing the current image 100 and the reference image 110 in units of pixels by luminance difference or the like, and determines the position of the moving object on the image. Output (step S3). Next, the moving object existence region estimation unit 23 divides the image into blocks of about 8 × 8 pixels, and moves trajectory coordinates (based on the position data of the moving object output from the moving object detection unit 22 for each frame). The existence probability is calculated and updated for the block to which the lower center point of the detection area belongs, and the existence probability of the moving object is calculated for each block. Further, the existence probability is learned (calculated, recorded, and updated) for each time zone in units of blocks. Further, if the existence probability is uniform within the target region, the block is selected at random (or fixed) from blocks where no moving object is detected within the target region (step S4). In this way, an area with a lot of traffic is learned ((A) in FIG. 2).

  Next, the feature amount effective region selection unit 24 calculates the feature amount for each block of about 8 × 8 pixels in the reference image 110 generated by the reference image generation unit 21, measures the number and evaluation value, A region having a large number of feature amounts and a high evaluation value is selected as a feature amount effective region (step S5). In this manner, a region having a large feature amount is extracted from the reference image 110 ((B) in FIG. 2).

  Next, the determination region priority setting unit 25 targets a region other than the region where the moving object detected by the moving object detection unit 22 exists, and the probability that there is no moving object estimated by the moving object existence region estimation unit 23 The priority of the region to be matched is set in consideration of the feature amount effective region extracted by the feature amount effective region selection unit 24 (step S6).

  Next, the camera tampering detection unit 26 uses the priority set by the determination region priority setting unit 25 to match the current image 100 obtained by the imaging unit 1 with the reference image 110 generated by the reference image generation unit 21. It is performed in a high-frequency area (step S7), and the presence or absence of camera interference is determined (step S8). However, when a block with a small feature amount is selected with high priority, block matching using luminance information and color information is performed. In addition, the blocks to be matched are dynamically changed according to the priority that changes every frame, so that the detection process can be performed with high accuracy according to the camera installation environment that changes every moment.

  When the camera interference detection unit 26 detects camera interference, it outputs a camera interference detection signal (step S9). When the camera disturbance detection signal is output, the notification unit 31 notifies the camera disturbance. The display unit 32 displays a message for notifying that camera interference has been detected.

  As described above, according to the intrusion monitoring apparatus 2 of the present embodiment, the moving object is estimated by measuring the number of detections per unit time for each small region based on the detection result of the moving object detection unit 22. An existence region estimation unit 23, and a feature amount effective region selection unit 24 that selects a region where there are many feature points that are robust to brightness fluctuations based on the reference image generated by the reference image generation unit 21; Since the disturbance detection judgment area to 1 is dynamically selected, the camera disturbance can be detected accurately without erroneously detecting the congestion of people and vehicles reflected in the camera as events such as monitoring disturbance to the camera. be able to.

  The present invention has the effect of being able to detect camera interference with high accuracy without erroneously detecting congestion of a person or vehicle reflected in the camera as an event such as monitoring interference to the camera. Useful for action detection.

DESCRIPTION OF SYMBOLS 1 Imaging part 2 Intrusion monitoring apparatus 21 Reference | standard image generation part 22 Moving object detection part 23 Moving object presence area estimation part 24 Feature-value effective area | region selection part 25 Judgment area priority setting part 26 Camera disturbance detection part 31 Notification part 32 Display part 100 Current image 101 People 102 Vehicle 110 Reference image

Claims (10)

An imaging unit for imaging the monitoring target area;
A reference image generation unit that generates a reference image that does not show a moving object from the image data obtained from the imaging unit;
A moving object detection unit that detects a region where an object exists by a comparison process between the reference image generated by the reference image generation unit and the image obtained from the imaging unit;
Measure the number of detection per unit time for each small area based on the detection result of the moving object detection unit, and the moving object existence region estimation unit for estimating the amount of traffic,
A feature amount effective region selection unit that selects a region where there are many feature points that are robust to variations in brightness based on the reference image generated by the reference image generation unit;
A determination region priority setting unit that selects and determines a determination region when performing camera tampering detection based on the estimation result of the moving object existence region estimation unit and the selection result of the feature amount effective region selection unit;
The image obtained from the imaging unit is matched with the reference image generated by the reference image generation unit in the disturbance detection determination region selected by the determination region priority setting unit, and the camera is selected according to the matching evaluation value A camera tampering detection unit that determines whether or not the
Intrusion monitoring device with   The intrusion monitoring apparatus according to claim 1, wherein the moving object existence region estimation unit divides the image into blocks, and calculates and updates the existence probability for the block to which the moving locus coordinates of the moving object detected periodically belong.   The intrusion monitoring apparatus according to claim 2, wherein the moving object existence region estimation unit calculates, records, and updates the existence probability for each time zone in a block unit.   The said moving object presence area estimation part selects the block from which the moving object is not detected within the object area at random or fixed when the calculated existence probability is uniform within the object area. 3. The intrusion monitoring apparatus according to 3.   The intrusion monitoring apparatus according to any one of claims 1 to 4, wherein the feature quantity effective region selection unit calculates a feature quantity in a reference image for each block and measures the number and evaluation value thereof.   The intrusion monitoring apparatus according to any one of claims 1 to 5, wherein the feature amount effective region selection unit excludes a region located in the sky on the image from being calculated.   2. The determination area priority setting unit determines a priority of an area to be subjected to matching processing in consideration of a probability that no moving object exists and an effective feature amount area for an area where no moving object exists. The intrusion monitoring apparatus according to any one of claims 6 to 6.   6. The camera tampering detection unit performs matching processing using a feature amount used by the feature amount effective region selection unit in a determination region having a high priority selected by the determination region priority setting unit. Item 8. The intrusion monitoring device according to any one of items 7.   The intrusion monitoring apparatus according to any one of claims 1 to 8, wherein the camera tampering detection unit dynamically changes a matching target block according to a priority of changing every frame in a matching process. Generating a reference image in which the moving object is not reflected from the image data obtained from the imaging unit that images the monitoring target region;
Detecting a region where an object exists by comparing the reference image and an image obtained from the imaging unit;
Measuring the number of detections per unit time for each small area based on the detection results of moving objects, estimating the amount of traffic,
Selecting an area where there are many feature points that are robust to brightness fluctuations based on the reference image;
Prioritizing and selecting a judgment area when performing camera tampering detection based on a moving object existence area estimation result and a feature amount effective area selection result;
Matching between the image obtained from the imaging unit and the reference image in the interference detection determination region in which the priority of the determination region is selected, and determining whether or not it is a camera interference according to a matching evaluation value; ,
Intrusion monitoring method with

Images