JP2011123562A - Image processing apparatus, image processing method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processing apparatus, an image processing method and a program for surely detecting a fluctuating area from an overall picture, and for highly precisely detecting an object to be monitored. <P>SOLUTION: The image processing apparatus is provided with: a difference processing part for calculating a difference between the image data of one image and the image data of the past image photographed in the past from one image; a fluctuation detection part for calculating the time change of the difference, and for extracting a fluctuation area in a picture equivalent to a section where the fluctuation of nature has been photographed and a non-fluctuation area which is not the fluctuation area based on the periodicity of the time change; and an object detection part for detecting an object moving in the picture by changing the detecting precision according to the fluctuation area and the non-fluctuation area. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an image processing apparatus, an image processing method, and a program.

  When a suspicious person or the like is monitored using an imaging device such as a monitoring camera, it always places a great burden on the supervisor to look at the suspicious person by looking at the monitor. Therefore, in monitoring using a monitoring camera or the like, a moving object detection technique for detecting a moving object (hereinafter also referred to as “moving object”) in an image, which is one of image processing techniques, is used.

  The moving object detection uses the luminance value of the moving object region in the image that changes when the moving object appears in the video. That is, a moving object is detected by detecting an area where a difference in luminance value occurs in the video as a difference area. There are (1) a difference detection method using a difference between frames and (2) a background difference method using a background image as a technique for detecting a difference area of an image for moving object detection.

  In the difference detection method using the inter-frame difference, the difference is detected by calculating the difference in luminance value between the image taken at the current time and the image taken before the unit time. And it determines with a moving body existing in the area | region where the brightness value produced the difference. On the other hand, in the background subtraction method using a background image, an image in which no moving object is reflected is recorded in advance as a background image. And a moving body area | region is detected by comparing the image image | photographed at the present time with a background image.

JP 2006-107457 A JP 2000-322581 A Japanese Patent Laid-Open No. 9-50585

  By the way, as shown in FIG. 13, in the conventional motion detection, in the video 50, not only a person 54 or a vehicle that is actually desired to be monitored but also disturbances existing in the natural world such as a tree fluctuation 52 and a wave fluctuation. (Fluctuation) is also detected as a moving object. The reason why the fluctuation is detected as a moving body is because it is difficult to distinguish the difference in fluctuation from the person or vehicle to be monitored by the luminance change of one frame. FIG. 13 is an explanatory diagram illustrating an example of a video on which a monitoring target, a background, and the like are displayed.

  As shown in FIG. 14A, the fluctuation is an object in which the luminance value fluctuates in a short time and the luminance value moves periodically. On the other hand, as shown in FIG. 14B, the background does not move the luminance value. In addition, as shown in FIG. 14C, the brightness value of a detection target such as a person or a vehicle moves in a certain direction. FIG. 14 is a graph showing temporal changes in luminance values of pixels or blocks in an image.

  In Patent Document 1, in order to determine fluctuations and moving objects, a plurality of past images and a current image are compared, and a luminance change of the moving objects in a plurality of frames is calculated. However, in the above method, the detection accuracy of the entire screen is uniform. For this reason, if the detection accuracy is lowered in order to suppress erroneous detection of fluctuation as a moving object, there is a problem that a moving object cannot be detected in a dark region or the like, resulting in detection failure.

  In Patent Document 2, a moving image area is detected in order to distinguish fluctuation from a moving object, and the detected image area is tracked. Then, using the fact that the fluctuation is periodically moved, if the detected image area is reciprocating, it is determined that the fluctuation is not a moving object. However, depending on the target, it is difficult to keep tracking accurately, and periodicity may not be determined.

  Furthermore, in Patent Document 3, fluctuations and moving objects are discriminated based on the shape and speed of the detected object. This is to determine a horizontally long moving body or a moving body faster than the set speed as a fluctuation. However, there are fluctuations that are vertically long and those that move slowly, so some fluctuations cannot be excluded.

  Therefore, the present invention has been made in view of the above problems, and an object of the present invention is to reliably detect a fluctuation region from the entire screen and to accurately detect an object to be monitored. It is another object of the present invention to provide a new and improved image processing apparatus, image processing method and program.

  In order to solve the above-described problem, according to an aspect of the present invention, a difference processing unit that calculates a difference between image data of one image and image data of a past image captured in the past from the one image, and a difference And a fluctuation detection unit that extracts a fluctuation area in the screen corresponding to the part where the fluctuation of the natural world is photographed, a non-fluctuation area that is not a fluctuation area, and a fluctuation area based on the periodicity of the time change And an object detection unit that detects an object that moves in the screen by changing detection accuracy according to the non-fluctuation region.

  The difference processing unit calculates a pixel-by-pixel difference, and the fluctuation detection unit calculates a periodicity of the pixel-by-pixel difference, and a fluctuation pixel determination unit that detects a fluctuation pixel corresponding to a portion in which a fluctuation in the natural world is photographed. A fluctuation weighting unit that performs weighting processing on the fluctuation pixels and a fluctuation area determination unit that extracts a fluctuation area and a non-fluctuation area in the screen according to the weighting of the fluctuation pixels may be included.

  The fluctuation pixel determination unit includes the number of times that the difference between pixel values in pixel units is equal to or greater than a first threshold per unit time, and the second threshold that is lower than the first threshold per unit time in pixel value differences. The fluctuation pixel may be detected based on a certain number of times.

  The object detection unit may be set so that an object moving in the screen is more easily detected in the non-fluctuation region than in the fluctuation region.

  In order to solve the above problem, according to another aspect of the present invention, the difference processing unit calculates a difference between the image data of one image and the image data of the past image captured in the past from the one image. The step of calculating and the fluctuation detection unit calculate the temporal change of the difference, and based on the periodicity of the temporal change, the fluctuation area in the screen corresponding to the portion where the fluctuation of the natural world is photographed and the non-fluctuation that is not the fluctuation area An image processing method is provided that includes a step of extracting a region and a step of detecting an object that moves in a screen by changing detection accuracy according to a fluctuation region and a non-fluctuation region.

  In order to solve the above problem, according to another aspect of the present invention, a means for calculating a difference between image data of one image and image data of a past image taken in the past from the one image, the difference Based on the periodicity of the time change, a means to extract the fluctuation area in the screen corresponding to the part where the fluctuation of the natural world was photographed and the non-fluctuation area that is not the fluctuation area, fluctuation area and non-fluctuation There is provided a program for causing a computer to function as means for detecting an object moving in a screen by changing detection accuracy according to a region.

  As described above, according to the present invention, a fluctuation region can be reliably detected from the entire screen, and an object to be monitored can be detected with high accuracy.

1 is a block diagram illustrating an imaging apparatus 100 according to an embodiment of the present invention. FIG. 2 is a block diagram illustrating an image processing unit 110 according to the same embodiment. It is a block diagram which shows the fluctuation area detection part 115 which concerns on the same embodiment. It is a block diagram which shows the fluctuation pixel determination part 120 which concerns on the same embodiment. It is explanatory drawing which shows the method of determining which pixel is which of the said three candidates. It is explanatory drawing which shows the method of integrating the determination result by two threshold values. It is a graph which shows the time change of the difference value of a certain pixel. It is a graph which shows the time change of the difference value of a certain pixel. It is explanatory drawing which shows the fluctuation weight value and the possibility of fluctuation. FIG. 10A is an explanatory diagram illustrating an example of a video 150 in which a fluctuation region is clearly shown. FIG. 10B is an explanatory diagram illustrating an example of a video on which a monitoring target, a background, and the like are displayed. FIG. 11A is a graph showing temporal changes in luminance values of a certain pixel. FIGS. 11B to 11D are graphs showing changes in the difference value over time. FIG. 12A is a graph showing temporal changes in the luminance value of a certain pixel. FIGS. 12B to 12D are graphs showing temporal changes in the difference values. It is explanatory drawing which shows an example of the image | video on which the monitoring object, the background, etc. were displayed. It is a graph which shows the time change of the luminance value of the pixel or block in an image.

  Exemplary embodiments of the present invention will be described below in detail with reference to the accompanying drawings. In addition, in this specification and drawing, about the component which has the substantially same function structure, duplication description is abbreviate | omitted by attaching | subjecting the same code | symbol.

The description will be made in the following order.
1. Configuration of one embodiment Operation of one embodiment

<1. Configuration of one embodiment>
[Imaging device 100]
First, with reference to FIG. 1, a configuration of an imaging apparatus 100 according to an embodiment of the present invention will be described. FIG. 1 is a block diagram illustrating an imaging apparatus 100 according to the present embodiment.
The imaging device 100 is, for example, a monitoring camera or the like, and is fixed at a certain location and images a monitoring target such as a person or a vehicle that passes through the imaging range. The imaging apparatus 100 transmits image data obtained by photographing to a monitoring board having a monitor. According to the present embodiment, an object to be monitored can be detected with high accuracy without the monitor constantly gazing at the monitor.

  The imaging apparatus 100 includes an imaging unit 102, an imaging signal processing unit 104, an imaging data processing unit 106, a transmission unit 108, an image processing unit 110, and the like.

  The imaging unit 102 includes, for example, an optical system such as a lens and an imaging element such as a CCD image sensor or a CMOS image sensor. The imaging unit 102 receives light from the subject and converts the light into an electrical signal. The imaging unit 102 sends the generated electrical signal to the imaging signal processing unit 104.

  The imaging signal processing unit 104 changes the signal received from the imaging unit 102 to a video signal such as a color image or performs noise removal processing on the video signal. The imaging signal processing unit 104 sends the processed video signal to the imaging data processing unit 106 and the image processing unit 110.

  The imaging data processing unit 106 performs a compression process or the like on the video signal processed by the imaging signal processing unit 104 and changes the format so that the video signal can be transmitted from the transmission unit 108.

  The transmission unit 108 associates the video signal received from the imaging data processing unit 106 and the metadata received from the image processing unit 110, and sends them to the outside. The output destination of the video signal and metadata from the transmission unit 108 is, for example, a network, and a monitoring board connected to the network.

  The image processing unit 110 receives the video signal processed by the imaging signal processing unit 104 and performs image processing on the video signal to generate metadata. In the present embodiment, metadata is information that an object to be monitored has passed through the imaging range, and is, for example, an alarm signal.

  In the present embodiment, the transmission unit 108 transmits both the video signal and the metadata to the outside, but the present invention is not limited to this example. For example, the transmission unit 108 may output only metadata such as an alarm without outputting a video signal.

  In the above description, the imaging device 100 such as a monitoring camera having the imaging unit 102 has been described. However, the present invention may be a signal conversion device that does not have the imaging unit 102. The signal conversion apparatus includes an input unit that can be connected to a camera that outputs an analog video signal instead of the imaging unit 102. The signal processing device generates metadata based on the analog video signal and outputs the metadata to the outside.

  Furthermore, in the above description, the imaging apparatus 100 including the imaging data processing unit 106 has been described. However, the present invention may be an image processing dedicated apparatus. The image processing dedicated device is, for example, a personal computer or a server. The dedicated image processing apparatus includes an image processing unit 110 that receives a video signal and generates metadata. The generated metadata may be output to the outside, or may be output to a monitor or a warning light connected to the image processing dedicated device.

[Image processing unit 110]
Next, the image processing unit 110 will be described. FIG. 2 is a block diagram illustrating the image processing unit 110 according to the present embodiment.
The image processing unit 110 includes an image storage unit 111, a difference processing unit 112, a labeling processing unit 113, a tracking processing unit 114, a fluctuation region detection unit 115, a threshold image generation unit 116, and the like.

  The image storage unit 111 stores the current image captured by the imaging unit 102 and processed by the imaging signal processing unit 104, and outputs a past image.

  The difference processing unit 112 receives the current image from the imaging signal processing unit 104 and simultaneously receives the past image from the image storage unit 111. Then, the difference processing unit 112 performs difference processing on the current image and a plurality of past images. As a result, the difference processing unit 112 can perform moving object detection in units of pixels, and outputs a moving object image when the moving object is included in the image.

  The labeling processing unit 113 receives the moving body image from the difference processing unit 112 and connects the moving body information detected for each pixel in the difference processing unit 112 to generate a moving body region. Then, the labeling processing unit 113 assigns a label number to the generated moving object region, and outputs a label image to which the label number is assigned.

  The tracking processing unit 114 associates the labeled moving object region with the moving object information of the previous frame. The tracking processing unit 114 updates / inherits moving object information when the current frame and the previous frame can be associated with each other. Then, the tracking processing unit 114 outputs the updated / inherited moving object region.

  Here, the combination of the labeling processing unit 113 and the tracking processing unit 114 is an example of an object detection unit, and detects an object moving in the screen.

  The fluctuation region detection unit 115 receives the current image from the imaging signal processing unit 104 and simultaneously receives the past image from the image storage unit 111. Then, the fluctuation region detection unit 115 detects a fluctuation region in the screen corresponding to a portion where a disturbance (fluctuation) existing in the natural world such as trees and waves is captured using an image, and a non-fluctuation region that is not a fluctuation region. . Then, the fluctuation region detection unit 115 outputs the fluctuation region to the threshold image generation unit 116.

  The threshold image generation unit 116 assigns different values for each pixel to the fluctuation region obtained by the fluctuation region detection unit 115 and the non-fluctuation region that is not the fluctuation region. Then, the threshold image generation unit 116 generates a threshold image that can be subjected to threshold processing from the fluctuation region. The threshold image generation unit 116 outputs the generated threshold image to the difference processing unit 112.

  As described above, according to the present embodiment, the fluctuation region detection unit 115 and the threshold image generation unit 116 can distinguish the fluctuation region and the non-fluctuation region. In the labeling processing unit 113 and the tracking processing unit 114, different detection levels can be set for the fluctuation region and the non-fluctuation region. For example, the tracking processing unit 114 sets the object that moves in the screen more easily in the non-fluctuation region than in the fluctuation region. As a result, a moving object can be detected relatively insensitively in the fluctuation region, and a moving object can be detected relatively sensitively in the non-fluctuation region. Therefore, since this embodiment can change detection accuracy for each region, it is possible to prevent erroneous detection of fluctuations and omission of detection of moving objects.

[Fluctuation area detection unit 115]
Next, the fluctuation region detection unit 115 will be described. FIG. 3 is a block diagram illustrating the fluctuation region detection unit 115 according to the present embodiment.
The fluctuation region detection unit 115 includes a smoothing processing unit 117, a smoothed image storage unit 118, a smoothing difference processing unit 119, a fluctuation pixel determination unit 120, a fluctuation weight image storage unit 121, a fluctuation weighting unit 122, The block processing unit 123 and the like are included.

  The smoothing processing unit 117 receives the current image and a plurality of past images, and averages the current image and the plurality of past images to remove noise in the image. The smoothing processing unit 117 outputs the current smoothed image from which noise has been removed to the smoothed image storage unit 118 and the smoothing difference processing unit 119.

  The smooth image storage unit 118 receives the current smooth image from the smoothing processing unit 117 and stores the current smooth image. The smoothed image storage unit 118 outputs the stored past smoothed image to the smoothed difference processing unit 119.

  The smoothing difference processing unit 119 receives the current smoothed image from the smoothing processing unit 117 and simultaneously receives the past smoothed image from the smoothed image storage unit 118, calculates a difference value in pixel units of the past smoothed image with respect to the current smoothed image, A difference image composed of the difference values is generated. The smoothing difference processing unit 119 outputs the generated difference image to the fluctuation pixel determination unit 120.

  The fluctuation pixel determination unit 120 is an example of a fluctuation detection unit, receives the difference image from the smoothing difference processing unit 119, and calculates the periodicity of the time change of the difference from a plurality of difference images. Then, the fluctuation pixel determination unit 120 detects a fluctuation pixel in the image corresponding to the portion where the fluctuation in the natural world is photographed based on the periodicity of the temporal change of the difference. Then, the fluctuation pixel determination unit 120 outputs the fluctuation image to the fluctuation weighting unit 122. The fluctuation image is an image including positional information of fluctuation pixels.

  The fluctuation weight image storage unit 121 receives the fluctuation weight image from the fluctuation weighting unit 122 and stores the fluctuation weight image. The fluctuation weight image indicates the possibility of fluctuation for each pixel. Further, the fluctuation weight image storage unit 121 outputs a past fluctuation weight image to the fluctuation weighting unit 122.

  The fluctuation weighting unit 122 receives a fluctuation image from the fluctuation pixel determination unit 120, information related to the moving body region of the previous frame from the tracking processing unit 114, and a fluctuation weight image from the fluctuation weight image storage unit 121. The fluctuation weighting unit 122 calculates the weight value by adding to the pixels of the fluctuation weight image excluding the moving body area of the previous frame if it is a fluctuation pixel and subtracting if it is a non-fluctuation pixel. Then, the fluctuation weighting unit 122 updates the fluctuation weight image composed of the weight values. The fluctuation weighting unit 122 sends the updated fluctuation weight image to the fluctuation weight image storage unit 121 and the blocking processing unit 123.

  The blocking processing unit 123 receives the fluctuation weight image from the fluctuation weighting unit 122 and divides the fluctuation weight image for each block. A block is a set of pixels. The blocking processing unit 123 identifies whether the weight value of the pixel in the block is positive or negative. Then, the blocking processing unit 123 determines that the entire block is a fluctuation region when pixels of a certain ratio or more are positive values, specifies coordinates and a range, and outputs the fluctuation region to the threshold image generation unit 116. .

[Fluctuation pixel determination unit 120]
Next, the fluctuation pixel determination unit 120 will be described. FIG. 4 is a block diagram illustrating the fluctuation pixel determination unit 120 according to the present embodiment.
The fluctuation pixel determination unit 120 includes a high threshold determination unit 124, a low threshold determination unit 125, a result integration unit 126, and the like.

  The high threshold determination unit 124 receives the difference image from the smoothed difference processing unit 119 and performs threshold processing on the difference image in units of pixels. Then, the high threshold determination unit 124 counts the number of times that the pixel difference is equal to or greater than the first threshold, and outputs the number of times obtained by counting to the result integration unit 126 as a high threshold determination result.

  The low threshold determination unit 125 receives the difference image from the smoothing difference processing unit 119 and performs threshold processing on the difference image in units of pixels. Then, the low threshold determination unit 125 counts the number of times that the pixel difference is equal to or larger than the second threshold, and outputs the number of times obtained by counting to the result integration unit 126 as a low threshold determination result.

  The result integration unit 126 receives the high threshold determination result from the high threshold determination unit 124 and the low threshold determination result from the low threshold determination unit 125, and integrates the high threshold determination result and the low threshold determination result. Then, the result integration unit 126 generates a fluctuation image including the position information of the pixel determined to be a fluctuation candidate, and outputs the generated fluctuation image to the fluctuation weighting unit 122.

<2. Operation of one embodiment>
Next, the fluctuation area detection process of the image processing unit 110 according to the present embodiment will be described.

  In the fluctuation area detection process, in a scene with a lot of noise, the noise may be erroneously determined as fluctuation. Therefore, noise is first removed from the image. The average of the current image and the image for the past K frames is calculated for each pixel. Thereby, a smooth image from which noise is removed is generated.

  Next, a difference image between the generated current smooth image of the current frame and the past N number of past smooth images is generated. Then, based on the generated difference image, a pixel that may be a moving object (moving object candidate pixel), a pixel that may be a fluctuation (fluctuation candidate), and a background may be present based on the difference between the current and the past A certain pixel (background candidate pixel) is obtained.

  FIG. 5 is an explanatory diagram showing a method for determining which of the above three candidates a pixel is. For example, among the past N smooth images, if the number of times that the current and past difference values are equal to or greater than the threshold Th is 0, the target pixel is determined as the background candidate pixel. When the number of times that the current and past difference values are equal to or greater than the threshold Th is N, the target pixel is determined to be a moving object candidate pixel. On the other hand, when the number of times that the current and past difference values are equal to or greater than the threshold Th is 1 to N−1, the target pixel is determined to be a fluctuation candidate pixel. At this time, as the past smooth image used for generating the difference image, the old one of the stored smooth images is preferentially used. As a result, the difference in the moving object region increases with time, so that it becomes easier to distinguish the moving object candidate from the fluctuation candidate.

  The above processing is performed on two threshold values, the high threshold value Th1 and the low threshold value Th2, and the results are integrated using the integration method described below. Then, a fluctuation image including the position information of the pixel determined to be a fluctuation candidate is generated. Here, the values of the high threshold Th1 and the low threshold Th2 are, for example, the maximum value and the minimum value of the luminance change that may occur in the fluctuation to be detected.

  FIG. 6 is an explanatory diagram illustrating a method of integrating the determination results based on two threshold values. 7 and 8 are graphs showing temporal changes in the difference value of a certain pixel. First, as shown in B, D, and E of FIG. 6, a pixel determined as a fluctuation candidate with at least one of the threshold values is determined as a fluctuation pixel (shaded portion in FIG. 6). And since it is a fluctuation pixel, the fluctuation weighting part 122 adds with respect to the pixel of a fluctuation weight image.

  For example, in FIG. 6B, as shown in FIG. 7B, the number of times that is equal to or greater than the high threshold Th1 is 0, and the number of times that is equal to or greater than the low threshold Th2 is 1 to N-1. Therefore, for the pixel, the high threshold determination result is a background candidate, and the low threshold determination result is a fluctuation candidate. Similarly, as shown in FIG. 8A, D in FIG. 6 indicates that the number of times that is equal to or higher than the high threshold Th1 is 1 to N-1 times, and that that is equal to or higher than the low threshold Th2 is also 1 to N-1 times. It is. Therefore, both the high threshold determination result and the low threshold determination result of the pixel are fluctuation candidates. In FIG. 6E, as shown in FIG. 8B, the number of times that is equal to or higher than the high threshold Th1 is 1 to N−1 times, and the number of times that is equal to or higher than the low threshold Th2 is N times. Therefore, for the pixel, the high threshold determination result is a fluctuation candidate, and the low threshold determination result is a moving object candidate.

  Also, as shown in FIG. 6C, when the target pixel is a background candidate if the high threshold determination result is a background candidate, and the low threshold determination result is a moving object candidate, it is determined as a fluctuation pixel. And since it is a fluctuation pixel, the fluctuation weighting part 122 adds with respect to the pixel of a fluctuation weight image. In C of FIG. 6, as shown in FIG. 7C, the number of times that is equal to or higher than the high threshold Th1 is 0, and the number of times that is equal to or higher than the low threshold Th2 is N. Note that a case like C in FIG. 6 may be determined as a moving pixel instead of a fluctuation pixel.

  Then, as shown in FIG. 6F, a pixel determined as a moving object candidate pixel by both of the two threshold values is determined as a moving object pixel. And since it is a moving body pixel, the fluctuation weighting part 122 does not add or subtract with respect to the pixel of a fluctuation weight image. In FIG. 6F, as shown in FIG. 8C, the number of times that is equal to or higher than the high threshold Th1 is N times, and the number of times that is equal to or higher than the low threshold Th2 is also N times.

  Further, as shown in FIG. 6A, a pixel determined as a background candidate pixel by both of the two threshold values is determined as a background pixel. And since it is a background pixel, the fluctuation weighting part 122 subtracts with respect to the pixel of a fluctuation weight image. In FIG. 6A, as shown in FIG. 7A, the number of times that is equal to or higher than the high threshold Th1 is zero, and the number of times that is equal to or higher than the low threshold Th2 is also zero.

  In this way, the fluctuation weight image is updated. The fluctuation weight image is generated by adding a value when the target pixel is a fluctuation pixel and subtracting when the target pixel is a background pixel. The fluctuation weight image indicates whether or not the target pixel is likely to fluctuate in time series. Therefore, as shown in FIG. 9, the fluctuation weight image has a higher possibility of being fluctuated as the value is larger in the positive direction, and is more likely to be the background as the value is larger in the negative direction. FIG. 9 is an explanatory diagram showing the fluctuation weight value and the possibility of fluctuation. Note that, when the pixel is determined as a moving object pixel, since the weighting value is not changed, it is possible to distinguish whether the fluctuation weight image is the background or the fluctuation.

  Next, the fluctuation weight image is divided into blocks, and pixels that may be fluctuations detected in units of pixels are connected as fluctuation areas. First, an image is divided by an arbitrary block size. Then, when the pixels of a certain ratio or more out of the pixels of each block have a positive value, the entire block is determined as a fluctuation region. The above processing is performed on all blocks to obtain the fluctuation region of the entire image.

  As described above, according to the present embodiment, the luminance change of each pixel is calculated from the difference between the current image and a plurality of past images, and the pixel that fluctuates in the video 150 is detected as shown in FIG. . FIG. 10A is an explanatory diagram illustrating an example of a video 150 in which a fluctuation region is clearly shown. Furthermore, by evaluating the frequency of occurrence of fluctuation, a fluctuation area 152 and a background area that is not the fluctuation area 152 are determined. Then, as shown in FIG. 10B, a monitoring target such as a person or a vehicle that has entered the video 150 is detected, and a detection frame 152 is displayed on the video 150. At this time, of the regions obtained by the determination, the moving object is detected insensitive to the fluctuation region and sensitive to the background region. FIG. 10B is an explanatory diagram illustrating an example of a video on which a monitoring target, a background, and the like are displayed. In this way, by giving different processing to each region, it is possible to prevent erroneous detection of fluctuations in which fluctuation is erroneously detected as a moving object, and it is possible to detect moving objects that could not be detected up to now and prevent detection omissions. .

  The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to such examples. It is obvious that a person having ordinary knowledge in the technical field to which the present invention pertains can come up with various changes or modifications within the scope of the technical idea described in the claims. Of course, it is understood that these also belong to the technical scope of the present invention.

  For example, in the above embodiment, the fluctuation detection method sets two threshold values, and determines whether or not a pixel is a fluctuation candidate based on the number of times the pixel difference is equal to or greater than the threshold value. It is not limited to examples. For example, the fluctuation may be detected by comparing the luminance value of a pixel at a certain time with the waveform of the luminance value before that time.

  For example, FIG. 11A is a graph showing a temporal change in the luminance value of a certain pixel, and shows a change in the luminance value when a moving object passes. The difference between the luminance value at time A and the waveform of the luminance value before time A is as shown in FIG. Similarly, time B and time C are as shown in FIGS. FIGS. 11B to 11D are graphs showing temporal changes in difference values. On the other hand, FIG. 12A is also a graph showing a temporal change in the luminance value of a certain pixel, and shows a change in the luminance value when fluctuation occurs. The difference between the luminance value at time A and the waveform of the luminance value before time A is as shown in FIG. Similarly, time B and time C are as shown in FIGS. FIGS. 12B to 12D are graphs showing temporal changes in the difference values.

  As can be seen by comparing FIGS. 11 (B) to (D) and FIGS. 12 (B) to (D), the temporal change in the difference value occurs only in a negative slope in the case of a moving pixel, whereas In the case of a fluctuation pixel, a positive inclination also occurs. This is because the fluctuations shake back in a short time. Therefore, if the waveform representing the temporal change of the difference value includes a positive slope, the pixel is determined as a fluctuation candidate. According to such a determination method, the number of cases in which a moving object is erroneously determined to be fluctuating is reduced. The subsequent stage is the same as in the above-described embodiment, and the fluctuation weight image is updated by adding a value when the target pixel is a fluctuation candidate pixel and subtracting a value when the target pixel is a background pixel.

DESCRIPTION OF SYMBOLS 100 Imaging device 102 Imaging part 104 Imaging signal processing part 106 Imaging data processing part 108 Transmission part 110 Image processing part 111 Image preservation | save part 112 Difference processing part 113 Labeling processing part 114 Tracking processing part 115 Fluctuation area | region detection part 116 Threshold image generation part 117 Smoothing processing unit 118 Smooth image storage unit 119 Smooth difference processing unit 120 Fluctuation pixel determination unit 121 Fluctuation weight image storage unit 122 Fluctuation weighting unit 123 Blocking processing unit 124 High threshold determination unit 125 Low threshold determination unit 126 Result integration unit

Claims (6)

A difference processing unit that calculates a difference between image data of one image and image data of a past image taken in the past from the one image;
A fluctuation detection unit that calculates a temporal change of the difference and extracts a fluctuation area in a screen corresponding to a portion where the fluctuation of the natural world is photographed and a non-fluctuation area that is not the fluctuation area based on the periodicity of the time change. When,
An image processing apparatus comprising: an object detection unit configured to detect an object moving within the screen by changing detection accuracy according to the fluctuation region and the non-fluctuation region. The difference processing unit calculates the difference in pixel units,
The fluctuation detection unit
A fluctuation pixel determining unit that calculates the periodicity of the difference in units of pixels and detects a fluctuation pixel corresponding to a part in which fluctuations in the natural world are photographed;
A fluctuation weighting unit that performs weighting processing on the fluctuation pixels;
The image processing apparatus according to claim 1, further comprising: a fluctuation area determination unit that extracts the fluctuation area and the non-fluctuation area in the screen according to the weighting of the fluctuation pixels. The fluctuation pixel determination unit
The number of times that the pixel value difference of the pixel unit is equal to or greater than the first threshold per unit time, and the number of times that the difference of the pixel value of the pixel unit is equal to or greater than the second threshold value that is lower than the first threshold per unit time. The image processing apparatus according to claim 2, wherein the fluctuation pixel is detected based on the fluctuation pixel.   The image processing apparatus according to claim 1, wherein the object detection unit is set to detect an object moving in the screen in the non-fluctuation area more easily than the fluctuation area. A step of calculating a difference between the image data of one image and the image data of a past image captured in the past from the one image;
A fluctuation detecting unit calculates a temporal change of the difference, and based on the periodicity of the temporal change, a fluctuation area in the screen corresponding to a portion where the fluctuation of the natural world is photographed, and a non-fluctuation area that is not the fluctuation area. Extracting, and
An object detection unit comprising: an object detection unit that detects an object moving within the screen by changing detection accuracy according to the fluctuation region and the non-fluctuation region. Means for calculating a difference between image data of one image and image data of a past image taken in the past from the one image;
Means for calculating a time variation of the difference, and extracting a fluctuation region in the screen corresponding to a portion where the fluctuation of the natural world is photographed based on the periodicity of the time variation, and a non-fluctuation region that is not the fluctuation region;
A program for causing a computer to function as means for detecting an object moving within the screen by changing detection accuracy according to the fluctuation region and the non-fluctuation region.

Images