JP2018055621A - Computer program, moving body detection method and moving body detection device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a computer program, a moving body detection method, and a moving body detection device capable of suppressing erroneous detection attributable to local brightness variation.SOLUTION: In a computer program causing a computer to detect a moving body imaged on an image, a computer storing a processing target image serving as a target of moving body detection processing is caused to execute processing including: a brightness correction step of performing brightness correction for adjusting bright distribution in the processing target image with brightness distribution in a specific reference image; and a step of performing moving body detection on the processing target image already subjected to the brightness correction.SELECTED DRAWING: Figure 4

Description

  The present invention relates to a computer program, a moving object detection method, and a moving object detection device for detecting a moving object from an image.

  In images such as moving images taken with a camera or images taken repeatedly, a background consisting of a stationary object such as a tree or a building and a moving object such as a person or a car moving relative to the background are shown. Conventionally, there is a need to detect moving objects from captured images. For example, recording of a moving image is started when a moving object is detected in the moving image captured by the surveillance camera. In general, a method called background difference is used to detect a moving object from an image. Background difference is a method of comparing a previously captured image with a newly captured image and extracting a portion not included in the previously captured image as a moving object from the newly captured image.

  The background difference has a problem that a portion that is originally a background may be erroneously detected as a moving object. For example, a local brightness change due to a change in the lighting environment, or a local background change such as a tree swing or a water surface change may be erroneously detected as a moving object. Various methods have been developed to solve such problems. Patent Document 1 discloses an example of background difference.

  As a method used for background difference, there is a method using a background model. There are mainly two methods for generating a background model: a method for performing background modeling using statistical information and a method for performing background modeling using spatial information. In the method of performing background modeling using statistical information, the temporal change of the pixel value included in the image is statistically modeled. In the method of performing background modeling using spatial information, the pixel of interest in the image and its surrounding area are used. The distribution pattern of the pixel values is modeled. In any of the methods, a portion outside the model can be detected as a moving object.

  In the method using a background model using statistical information, a certain degree of robustness can be secured against local background changes. On the other hand, there is a problem that local brightness fluctuations are erroneously detected as moving objects. In particular, erroneous detection increases when the camera automatically corrects brightness by adjusting exposure or white balance. In addition, the method using a background model that uses spatial information can ensure a certain degree of robustness against local brightness fluctuations in addition to local background changes, and uses statistical information. There is a tendency that the false detection rate is lower than the method using the background model. On the other hand, since the accuracy and speed of processing vary according to the setting of the peripheral region, there is a problem that it is difficult to adjust parameters.

Japanese Patent Laid-Open No. 2006-76015

  The problem that local brightness fluctuations are erroneously detected as moving objects by using a background model that uses statistical information is avoided for a short period of time by not correcting the brightness of the camera. be able to. However, when moving objects are detected for a long period of time, such as when the camera is operated throughout the day and night, the brightness of the image changes greatly when the brightness is not corrected, and a normal image is acquired. Can not be. As a result, there is a problem that it is difficult to detect moving objects.

  The present invention has been made in view of such circumstances, and an object of the present invention is to provide a computer program, a moving object detection method, and a moving object detection capable of suppressing erroneous detection due to local brightness fluctuations. To provide an apparatus.

  The computer program according to the present invention is a computer program for causing a computer to detect a moving object in an image. In a computer program for storing a processing target image to be processed for detecting a moving object, the brightness in the processing target image is stored. Including a brightness correction step for performing brightness correction to match the distribution of brightness with the brightness distribution in a specific reference image, and a step for performing moving object detection on the processing target image after performing the brightness correction. Is executed.

  In the computer program according to the present invention, in the brightness correction step, the brightness of each pixel in the processing target image is determined based on a difference in brightness of each pixel included in the reference image and the processing target image. A correction function generating step for generating, for each pixel, a correction function for correcting a plurality of numerical values representing the color of each pixel so as to convert to the brightness of each pixel in the reference image; and each of the processing target images And calculating a plurality of numerical values representing colors whose brightness has been corrected by the correction function generated for each pixel from a plurality of numerical values representing the color of the pixel.

  In the computer program according to the present invention, the correction function generating step calculates a smoothed grayscale value obtained by performing edge preserving smoothing on a grayscale value of each pixel included in the processing target image; and the reference image And a ratio of smoothed grayscale values of each pixel included in the processing target image by dividing the smoothed grayscale value in the reference image by the smoothing grayscale value in the processing target image And for each pixel whose ratio is 1 or less, each numerical value representing the color in the processing target image is inputmg (x, y), each numerical value representing the color whose brightness is corrected is outputlmg (x, y), When the ratio is ratio (x, y), output mg (x, y) = input mg (x, y) * ratio (x, y) A step of generating the correction function represented, and a smoothed grayscale value in the reference image as a reference grayscale value in the reference image and a smoothed grayscale in the processing target image with respect to the pixels whose ratio exceeds 1 The value is inputMap (x, y), and the maximum value that each numerical value representing the color can take is max. Outputmg (x, y) = inputlmg (x, y) * rate + offset, rate = (max−referenceMap (x, y) ) / (Max-inputMap (x, y)) and generating the correction function represented by offset = referenceMap (x, y) -rate * inputMap (x, y). .

  The computer program according to the present invention is characterized in that the reference image is specified by a background difference using a background model using statistical information.

  A moving object detection method according to the present invention is a moving object detection method in which a computer including a calculation unit and a storage unit detects a moving object copied in an image. In the moving object detection method, a processing target image to be processed is detected in the storage unit. A step of storing, a step of performing a brightness correction to match a brightness distribution in the processing target image with a brightness distribution in a specific reference image, and a processing target image after performing the brightness correction And moving object detection.

  The moving object detection device according to the present invention is a moving object detection device that detects a moving object that is captured in an image, and a storage unit that stores a processing target image that is a target of processing for detecting the moving object, and brightness within the processing target image. A brightness correction unit that performs brightness correction that matches the distribution of brightness with a brightness distribution in a specific reference image, and a moving object detection unit that performs moving object detection on the processing target image after performing the brightness correction. It is characterized by providing.

  In the present invention, the moving object detection device performs brightness correction on the processing target image to match the brightness distribution with the brightness distribution in a specific reference image, and then performs moving object detection. By performing the moving object detection from the processing target image in which the local brightness change is reduced, it is possible to suppress the erroneous detection of the local brightness fluctuation as the moving object.

  Further, in the present invention, when performing the brightness correction, the moving object detection device calculates a plurality of numerical values representing the color of each pixel based on the difference in brightness of each pixel included in the reference image and the processing target image. A correction function for correction is generated for each pixel, and a plurality of numerical values representing the color of each pixel are corrected using the correction function. For each pixel, the brightness distribution in the processing target image is corrected so that the brightness distribution matches the reference image by correcting the color of the pixel based on the difference in brightness between the reference image and the processing target image. The

  Further, in the present invention, when generating the correction function, the moving object detection device calculates a smoothed grayscale value obtained by performing edge preserving smoothing on the grayscale value of each pixel, and calculates the reference image and the processing target image. A ratio of the smoothed gray scale values is calculated, and a correction function is generated for each pixel depending on whether the calculated ratio is 1 or less or exceeds 1. Noise is removed by edge preserving smoothing. Even when the ratio of the smoothed gray scale value exceeds 1, the correction function is generated separately according to the value of the smoothed gray scale value ratio so that the correction function is linear. The occurrence of flying is prevented.

  Further, in the present invention, the moving object detection device includes a reference image that does not include an initial moving object based on a difference value based on a background difference using a background model that uses statistical information from among images included in a moving image. Can be identified.

  In the present invention, it is possible to reliably reduce the false detection of moving objects, and to provide excellent effects such as enabling the detection of moving objects even when moving objects are detected over a long period of time. Play.

It is a schematic diagram which shows the structural example of the camera system which image | photographs an image and detects a moving body from a picked-up image. It is a functional block diagram which shows the function structure of a moving body detection apparatus. It is a block diagram which shows the structural example of a moving body detection apparatus. It is a flowchart which shows the procedure of the process which a moving body detection apparatus performs. It is a flowchart which shows the detailed procedure of the brightness correction process of S3. It is a schematic diagram which shows the example of a reference image. It is a schematic diagram which shows the example of an input image. It is a characteristic view which shows the example of a correction function. It is a characteristic view which shows the example of a correction function. It is a flowchart which shows the detailed procedure of the process of a moving body detection of S4. It is a schematic diagram which shows the example of the input image which copied the moving body. It is a schematic diagram which shows the example of a detection of a moving body. It is a schematic diagram which shows the example of a detection of a moving body.

Hereinafter, the present invention will be specifically described with reference to the drawings showing embodiments thereof.
FIG. 1 is a schematic diagram illustrating a configuration example of a camera system that captures an image and detects a moving object from the captured image. The camera system includes a camera 2 that performs photographing and a moving object detection device 1 that performs moving object detection. The camera 2 is connected to the moving object detection apparatus 1. For example, the camera 2 is a surveillance camera. The camera 2 creates an image in the form of a moving image obtained by continuously photographing the outside of the camera 2 or a plurality of still images obtained by repeatedly photographing the outside of the camera 2, and inputs the created images to the moving object detection device 1.

  FIG. 2 is a functional block diagram showing a functional configuration of the moving object detection apparatus 1. The moving object detection apparatus 1 includes a processing unit 11 that performs information processing, an interface unit 14, and a storage unit 15. The interface unit 14 is connected to the camera 2. An image input from the camera 2 is received by the interface unit 14 and stored in the storage unit 15. The image is created by the camera 2 as electronic data of a predetermined format such as bitmap data and stored in the storage unit 15. The image is composed of a plurality of pixels, and the electronic data representing the image includes pixel values that are a plurality of numerical values representing the color of each pixel. For example, the color of each pixel is represented by a combination of lightness values of R (red), G (green), and B (blue). The storage unit 15 stores at least a reference image 151 referred to for moving object detection and an input image 152 input from the camera 2. The input image 152 corresponds to the processing target image.

  The processing unit 11 includes a brightness correction unit 12 that performs image brightness correction, and a moving object detection unit 13 that detects a moving object from the brightness-corrected image. The brightness correction unit 12 includes a grayscale value calculation unit 121 that calculates a grayscale value of an image, a smoothing unit 122 that performs smoothing of the grayscale value, and a grayscale between the reference image 151 and the input image 152. A comparison unit 123 that compares values, a correction function generation unit 124 that generates a correction function for correcting brightness, and a correction calculation unit 125 that calculates brightness correction are included. The moving object detection unit 13 includes a background model generation unit 131, a moving object candidate detection unit 132 that detects a moving object candidate from the input image 152, and a moving object specification unit 133 that specifies a moving object from the moving object candidates. .

  The moving object detection device 1 is configured by a computer such as a server device or a personal computer. FIG. 3 is a block diagram illustrating a configuration example of the moving object detection apparatus 1. The moving object detection apparatus 1 includes a CPU (Central Processing Unit) 101 that performs calculation, a RAM (Random Access Memory) 102, a drive unit 103 that reads information from a recording medium 3 such as an optical disk, a nonvolatile storage unit 104, an interface Part 14. The CPU 101 corresponds to a calculation unit. For example, the non-volatile storage unit 104 is a hard disk or a non-volatile semiconductor memory. The CPU 101 causes the drive unit 103 to read the computer program 31 recorded on the recording medium 3 and stores the read computer program 31 in the nonvolatile storage unit 104. The computer program 31 may be downloaded from the outside of the moving object detection device 1 or may be stored in advance in the nonvolatile storage unit 104. The CPU 101 loads the computer program 31 from the nonvolatile storage unit 104 to the RAM 102 as necessary, and executes various processes according to the loaded computer program 31. The processing unit 11 is realized when the CPU 101 executes processing according to the computer program 31. The storage unit 15 includes a RAM 102 and a nonvolatile storage unit 104. In the figure, an example in which the reference image 151 and the input image 152 are stored in the nonvolatile storage unit 104 is shown, but the reference image 151 and the input image 152 may be stored in the RAM 102. Note that part or all of the processing unit 11 may not be realized by processing using software, but may be configured by hardware.

  Below, operation | movement of the moving body detection apparatus 1 is demonstrated. FIG. 4 is a flowchart illustrating a procedure of processing executed by the moving object detection apparatus 1. The CPU 101 executes the following processing based on the computer program 31. An image is input from the camera 2 to the moving object detection apparatus 1 (S1), and the moving object detection apparatus 1 stores the reference image 151 and the input image 152 input from the camera 2 in the storage unit 15 (S2). The image input from the camera 2 is the latest part of the moving image captured by the camera 2 or the latest still image captured by the camera 2. The reference image 151 is an image that can be expected to include only the background. For example, the reference image 151 is an image of the first frame of a moving image shot by the camera 2 or an image created in advance. The input image 152 is, for example, one frame image included in the input moving image or an input still image. In S <b> 1, the CPU 101 stores the input image 152 in the nonvolatile storage unit 104 and the RAM 102, reads the reference image 151 stored in advance in the nonvolatile storage unit 104, and stores the reference image 151 in the RAM 102. The CPU 101 corrects the brightness of the input image 152 so that the brightness distribution in the input image 152 matches the brightness distribution in the reference image 151 (S3), and then the brightness correction is performed. A moving object is detected from the input image 152 (S4). The brightness correction process in S3 corresponds to the brightness correction unit 12, and the moving object detection process in S4 corresponds to the moving object detection unit 13.

  FIG. 5 is a flowchart showing a detailed procedure of the brightness correction process in S3. FIG. 6 is a schematic diagram illustrating an example of the reference image 151, and FIG. 7 is a schematic diagram illustrating an example of the input image 152. The reference image 151 shown in FIG. 6 and the input image 152 shown in FIG. 7 are images taken of the same room. Compared to the reference image 151 shown in FIG. 6, the brightness locally fluctuates in the input image 152 shown in FIG. 7. The CPU 101 calculates the gray scale value of each pixel included in the input image 152 (S31). The gray scale value is a value representing the brightness of each pixel. When the gray scale value is represented by 8 bits, the gray scale value is any value from 0 to 255. The gray scale value is calculated by a predetermined method from a plurality of pixel values representing the color of each pixel. For example, the CPU 101 calculates a gray scale value using a predetermined calculation formula from the RGB brightness values recorded for each pixel.

  Next, the CPU 101 performs edge preserving smoothing on the gray scale value of each pixel of the input image 152 (S32). A smoothed gray scale value obtained by performing edge preserving smoothing on the gray scale value of each pixel of the input image 152 is calculated. The CPU 101 performs edge preserving smoothing calculation using a predetermined filter. The edge preserving smoothing smoothes the brightness distribution and removes noise while preserving the edges in the input image 152. The smoothed gray scale value of each pixel of the reference image 151 is calculated in advance and stored in the nonvolatile storage unit 104. The CPU 101 reads the smoothed gray scale value of each pixel of the reference image 151 from the nonvolatile storage unit 104 and stores it in the RAM 102. Note that the CPU 101 may also calculate the gray scale value of each pixel of the reference image 151 in S31, and may also calculate the smoothed gray scale value of each pixel of the reference image 151 in S32.

  Next, the CPU 101 calculates the ratio of the smoothed gray scale values of the pixels included in the reference image 151 and the input image 152 (S33). Specifically, a value obtained by dividing the smoothed grayscale value of the reference image 151 by the smoothed grayscale value of the input image 152 is calculated. The pixel position in the image is represented by (x, y), the smoothed grayscale value of the reference image 151 is referred to as referenceMap (x, y), and the smoothed grayscale value of the input image 152 is inputMap (x, y). And The calculated ratio value ratio (x, y) is ratio (x, y) = referenceMap (x, y) / inputMap (x, y).

Next, the CPU 101 selects one pixel from a plurality of pixels included in the input image 152 (S34). Next, the CPU 101 determines whether or not the ratio ratio (x, y) of the smoothed grayscale values calculated in S33 for the selected pixel is 1 or less (S35). When the ratio of the smoothed gray scale value is 1 or less (S35: YES), the CPU 101 calculates a correction function for correcting the color of the pixel according to the ratio of the smoothed gray scale value being 1 or less. Generate (S36). The pixel value in the input image 152 is set as inputmg (x, y), and the corrected pixel value is set as outputmg (x, y). The correction function when the ratio of the smoothed gray scale values is 1 or less is a point (0, 0) on the two-dimensional coordinate system where input mg (x, y) = 0 and output mg (x, y) = 0. , Inputmg (x, y) = inputMap (x, y), outputmg (x, y) = referenceMap (x, y) and a line passing through a point (inputMap (x, y), referenceMap (x, y)) Let it be a function. In S36, the CPU 101 generates a correction function represented by the following equation (1).
outputlmg (x, y) = inputlmg (x, y) * ratio (x, y) (1)

When the ratio ratio (x, y) of the smoothed gray scale value exceeds 1 (S35: NO), the CPU 101 generates a correction function corresponding to the ratio of the smoothed gray scale value exceeding 1 ( S37). The maximum value that the pixel value can take is assumed to be max. The correction function when the ratio of the smoothed gray scale values exceeds 1 is a point (inputMap (x, y), referenceMap (x, y)), inputlmg (x, y) = max, outputlmg (x, y ) = A linear function passing through a point (max, max). When the pixel value is represented by 8 bits, max = 255. In S37, the CPU 101 generates a correction function represented by the following equations (2) to (4).
outputmg (x, y) = inputmg (x, y) * rate + offset (2)
rate = (max-referenceMap (x, y)) / (max-inputMap (x, y)) (3)
offset = referenceMap (x, y) -rate * inputMap (x, y) (4)

  8 and 9 are characteristic diagrams showing examples of the correction function. FIG. 8 shows a correction function when the ratio ratio (x, y) of the smoothed gray scale values is 1 or less, and FIG. 9 shows a correction function when the ratio of the smoothed gray scale values exceeds 1. . In FIGS. 8 and 9, the correction function is indicated by a solid line, and a straight line having a slope 1 connecting points (0, 0) and (max, max) on the two-dimensional coordinate system is indicated by a broken line. Whether the ratio of the smoothed gray scale values is 1 or less or exceeds 1, the correction function is a linear function and passes through (inputMap (x, y), referenceMap (x, y)). Further, the inclination of the correction function is 1 or less. When the ratio of the smoothed gray scale values is 1, the slope of the correction function is 1. Each of the correction functions is a function for correcting the pixel value so that the brightness of the pixel in the input image 152 matches the brightness in the reference image 151.

  When the ratio of the smoothed gray scale values exceeds 1, if the correction function is a linear function passing through the point (0, 0) as shown in the equation (1), the value inputmg (x, y) before correction is max. The corrected value outputlmg (x, y) reaches max before the value reaches the maximum value. That is, the correction function becomes non-linear, and there is a possibility that whiteout occurs due to brightness correction. As shown in FIG. 9, by generating the correction function separately according to the value of the smoothed grayscale value ratio so that the correction function becomes linear even when the ratio of the smoothed grayscale value exceeds 1. The occurrence of overexposure due to brightness correction is prevented.

  After S <b> 36 or S <b> 37 ends, the CPU 101 calculates a pixel value representing a color whose brightness has been corrected from the pixel value in the input image 152 using the generated correction function (S <b> 38). In S38, the CPU 101 substitutes the pixel value for inputmg (x, y) in the correction function, and calculates the corrected pixel value outputlmg (x, y). The corrected pixel value outputlmg (x, y) is calculated for each of a plurality of pixel values representing the color of the pixel using the same correction function. For example, for each of the RGB brightness values, the brightness value with the brightness corrected is calculated by the same correction function. If the ranges that can be taken by a plurality of pixel values representing the color of the pixel are different, the value of max in the equation (3) is set to a value corresponding to the range that the pixel value can take, and the rest is the same correction function. The corrected pixel value is calculated. Through S38, each pixel value representing the color of the pixel is corrected so that the brightness of the pixel in the input image 152 matches the brightness in the reference image 151. The CPU 101 stores the calculated pixel values after correction in the storage unit 15.

  Next, the CPU 101 determines whether or not there is a pixel that has not been processed in S34 to S38 (S39). If there is an unprocessed pixel (S39: YES), the CPU 101 returns the process to S34 and selects one pixel from the unprocessed pixels. Brightness correction is performed so that the brightness distribution in the input image 152 matches the brightness distribution in the reference image 151 by calculating pixel values representing corrected colors for all pixels in the input image 152. Is done. By storing the corrected pixel values for all the pixels in the input image 152, the brightness-corrected input image 152 is stored in the storage unit 15. If there is no unprocessed pixel (S39: NO), the CPU 101 ends the brightness correction process in S3. Note that the CPU 101 may generate a correction function for each pixel and then perform a process of calculating a corrected pixel value for each pixel. The processing of S31 corresponds to the gray scale value calculation unit 121, the processing of S32 corresponds to the smoothing unit 122, the processing of S33 corresponds to the comparison unit, and the processing of S34 to S37 corresponds to the correction function generation unit 124. , S38 corresponds to the correction calculation unit 125.

  FIG. 10 is a flowchart showing a detailed procedure of the moving object detection process of S4. FIG. 11 is a schematic diagram illustrating an example of the input image 152 in which a moving object is copied. The input image 152 shown in FIG. 11 shows a moving object that is a person in addition to the background shown in the reference image 151 shown in FIG. The CPU 101 generates a background model to be used for the background difference by background modeling using statistical information based on the image subjected to brightness correction (S41). The non-volatile storage unit 104 stores a plurality of images in which the background is copied and the brightness is corrected. The CPU 101 reads out the plurality of images, and based on the read out plurality of images, the statistical information is stored. Perform background modeling using information. The plurality of images that are the basis of the background model include a plurality of input images 152 that have been input in the past and subjected to brightness correction. The reference image 151 may be included in the plurality of images that are the basis of the background model. For example, the CPU 101 generates a background model by statistically processing changes in pixel values of a portion corresponding to the background in the past input image 152 that has been subjected to brightness correction. In S41, the CPU 101 may read a background model generated and stored in advance, or may update a background model generated in advance.

  Next, the CPU 101 detects a moving object candidate from the input image 152 on which the brightness correction has been performed based on the background difference (S42). In S42, the CPU 101 compares the background model with the input image 152 subjected to brightness correction, extracts pixels whose pixel values are different from those of the background model, and detects a region including the extracted pixels as a moving object candidate. To do. For example, the CPU 101 determines that the pixel value is different from the background model when the pixel value is not included in the allowable range defined for each pixel using the average and standard deviation of the pixel value in the background model. In S <b> 42, the CPU 101 encloses an area including pixels whose pixel values are different from the background model with a rectangular frame, and labels the area enclosed by the rectangular frame in the input image 152 as a moving object candidate. In S42, moving object candidates may not be detected, and a plurality of moving object candidates may be detected.

  Next, the CPU 101 determines whether there is a moving object candidate (S43). When there is a moving object candidate (S43: YES), the CPU 101 selects one moving object candidate (S44). Next, the CPU 101 determines whether or not the size of the rectangular frame of the selected moving object candidate in the input image 152 is within a predetermined range (S45). When the upper limit value and the lower limit value of the size of the rectangular frame are determined in advance, and the size of the rectangular frame is not less than the lower limit value and not more than the upper limit value, the CPU 101 determines that the size of the rectangular frame of the moving object candidate is within a predetermined range. Is determined. The upper limit value and the lower limit value of the size of the rectangular frame are determined in advance according to an appropriate size of the moving object that can appear in the input image 152. The upper limit value and the lower limit value are stored in advance in the nonvolatile storage unit 104 or are recorded in the computer program 31.

  When the size of the rectangular frame of the moving object candidate is within the predetermined range (S45: YES), the CPU 101 identifies an area surrounded by the rectangular frame in the input image 152 as a moving object (S46). The CPU 101 stores information indicating the area identified as a moving object in the nonvolatile storage unit 104 in association with the input image 152. When the size of the rectangular frame of the moving object candidate is out of the predetermined range (S45: NO), or after S46 ends, the CPU 101 returns the process to S43 and determines whether there is an unprocessed moving object candidate. To do.

  If there is no moving object candidate in S43 (S43: NO), the moving object detection process in S4 is terminated. Finally, a moving object may not be detected at all, and a plurality of moving objects may be detected. The process of S41 corresponds to the background model generation unit 131, the process of S42 corresponds to the moving object candidate detection unit 132, and the processes of S43 to S46 correspond to the moving object specifying unit 133. Note that the CPU 101 may update the background model and store it in the nonvolatile storage unit 104 after the process of S43. After the moving object detection process of S4 ends, the moving object detection apparatus 1 ends the process.

  12 and 13 are schematic diagrams showing examples of moving object detection. FIG. 12 illustrates an example in which a moving object is detected from the input image 152 illustrated in FIG. 11 without performing brightness correction. FIG. 13 shows an example in which a moving object is detected after performing brightness correction from the input image 152 shown in FIG. 11 according to the present embodiment. In the example shown in FIG. 12, a portion whose brightness has fluctuated locally is erroneously detected as a moving object. In the example shown in FIG. 13, a portion whose brightness has fluctuated locally is not erroneously detected, and the moving object shown in the image is specified by being surrounded by a rectangular frame.

  Each time an image is input from the camera 2, the moving object detection device 1 executes the processes of S <b> 1 to S <b> 4. For example, each time a frame of a moving image shot by the camera 2 is input, the moving object detection apparatus 1 executes the processes of S1 to S4. The reference image 151 used in the processing may be the same, and the reference image 151 may be changed as time passes. Note that the moving object detection apparatus 1 may execute the process at other timings instead of executing the process every time an image is input. For example, the moving object detection device 1 stores the input image in the storage unit 15 and is stored when a predetermined amount of image is stored, when an instruction to start processing is received, or periodically. You may perform the process of S3 and S4 about each of an image.

  The detection result of the moving object in the moving object detection device 1 is used for processing in another device. For example, when a moving object is detected, a process of starting recording a moving image taken by the camera 2 is performed. In addition, for example, when a moving object is detected, a process for issuing an alarm is performed. Further, for example, when a moving object is detected, a process of starting the customer service by the customer service robot is performed.

  As described above in detail, in the present embodiment, the moving object detection apparatus 1 performs brightness correction on the input image 152 and then performs moving object detection based on background differences. With background difference using a background model using statistical information, it is possible to detect a moving object that is robust against local background changes. That is, it is possible to prevent a local background change such as a tree swing from being erroneously detected as a moving object. Further, by performing brightness correction so that the brightness distribution in the input image 152 matches the brightness distribution in the reference image 151, the local brightness in the input image 152 compared to the reference image 151 is obtained. The change in is reduced. By performing moving object detection from the input image 152 in which the change in local brightness is reduced, it is possible to perform moving object detection that is robust against local brightness fluctuations. That is, erroneous detection of local brightness fluctuations as moving objects is suppressed. Therefore, as shown in FIG. 13, erroneous detection of moving objects is reliably reduced.

  Since erroneous detection due to local brightness fluctuations is suppressed, erroneous detection is suppressed even when the camera 2 corrects the brightness by adjusting exposure or white balance. For this reason, even when a moving object is detected over a long period of time, a normal image can be acquired, and the moving object detection can be continued.

  In the present embodiment, a mode in which a moving object is detected from an image created by the camera 2 is shown. However, the moving object detection device 1 is a mode in which a moving object is detected from an image created by a method other than shooting by the camera. May be. In addition to setting the first frame image included in the moving image as the reference image 151, the moving object detection device 1 uses background difference using a background model using statistical information among all the frames included in the moving image. It is also possible to specify an image having the smallest difference value as a reference image 151 that is a base image that does not include an initial moving object. Moreover, although the form which detects a moving body from a color image was shown in this embodiment, the form which detects a moving body from a monochromatic image may be sufficient as the moving body detection apparatus 1. FIG. In the present embodiment, the moving object detection device 1 and the camera 2 are separated from each other. However, the moving object detection device 1 and the camera 2 may be integrated. For example, you may comprise the moving body detection apparatus 1 and the camera 2 which were united with the smart phone.

1 Moving Object Detection Device 101 CPU
102 RAM
DESCRIPTION OF SYMBOLS 104 Nonvolatile memory | storage part 11 Processing part 12 Brightness correction | amendment part 13 Moving body detection part 15 Memory | storage part 151 Reference image 152 Input image 2 Camera 3 Recording medium 31 Computer program

Claims (6)

In a computer program that causes a computer to detect a moving object in an image,
In a computer that stores a processing target image that is a target of processing for detecting a moving object,
A brightness correction step for performing brightness correction to match the brightness distribution in the processing target image with the brightness distribution in a specific reference image;
A computer program for executing a process including a step of performing moving object detection on a processing target image after performing brightness correction. The brightness correction step includes
Based on the difference in brightness of each pixel included in the reference image and the processing target image, so as to convert the brightness of each pixel in the processing target image to the brightness of each pixel in the reference image, A correction function generation step for generating, for each pixel, a correction function for correcting a plurality of numerical values representing the color of each pixel;
Calculating a plurality of numerical values representing colors whose brightness has been corrected by the correction function generated for each pixel from a plurality of numerical values representing the color of each pixel included in the processing target image. The computer program according to claim 1. The correction function generation step includes:
Calculating a smoothed grayscale value obtained by performing edge preservation smoothing on the grayscale value of each pixel included in the processing target image;
By dividing the ratio of the smoothed grayscale value of each pixel included in the reference image and the processing target image by dividing the smoothing grayscale value in the reference image by the smoothing grayscale value in the processing target image A calculating step;
For pixels with a ratio of 1 or less, each numerical value representing a color in the processing target image is inputmg (x, y), each numerical value representing a color whose brightness is corrected is outputlmg (x, y), and the ratio A ratio (x, y), and generating the correction function represented by the following equation (1):
outputlmg (x, y) = inputlmg (x, y) * ratio (x, y) (1)
For pixels with the ratio exceeding 1, the smoothed grayscale value in the reference image is represented by referenceMap (x, y), the smoothed grayscale value in the processing target image is represented by inputMap (x, y), and the color is represented. A step of generating the correction function represented by the following formulas (2) to (4), where max is a maximum value that each numerical value can take;
outputmg (x, y) = inputmg (x, y) * rate + offset (2)
rate = (max-referenceMap (x, y)) / (max-inputMap (x, y)) (3)
offset = referenceMap (x, y) -rate * inputMap (x, y) (4)
The computer program according to claim 2, comprising:   The computer program according to claim 1, wherein the reference image is specified by a background difference using a background model using statistical information. In a moving object detection method for detecting a moving object copied in an image by a computer including a calculation unit and a storage unit,
Storing a processing target image to be a target of processing for detecting a moving object in the storage unit;
Performing brightness correction to match the brightness distribution in the processing target image with the brightness distribution in a specific reference image;
A moving object detection method comprising: performing a moving object detection on a processing target image after performing brightness correction. In a moving object detection apparatus for detecting a moving object shown in an image,
A storage unit that stores a processing target image that is a target of processing for detecting a moving object;
A brightness correction unit that performs brightness correction to match the brightness distribution in the processing target image with the brightness distribution in a specific reference image;
A moving object detection device comprising: a moving object detection unit configured to detect a moving object with respect to a processing target image after performing brightness correction.

Images