JP5215952B2 - Object detection device - Google Patents

Description

  The present invention relates to an object detection device that detects an object from an image obtained by imaging a monitoring space, and more particularly to an object detection device that detects an object based on edge information extracted from the image.

  In order to detect an object from an image, a discriminator (discriminator) that learns the object and features other than the object in advance is used. At this time, edge information representing the outline of the object may be used as a feature for identifying the object. In order to obtain sufficient edge information for identification, sufficient contrast is required between the object and the background.

  Japanese Patent Application Laid-Open No. H10-228561 describes that a correction process for normalizing a luminance value is performed on a partial image in a detection target image in order to improve a problem of failing to distinguish an object due to a change in contrast. If there is a person wearing dark clothes in the shade, or there is a person wearing light clothes in the sun, an improvement in identification accuracy can be expected when the contrast is insufficient. .

JP 2007-25766 Gazette

  However, if there is a person wearing light-colored clothes in the shaded area or a person wearing dark-colored clothes in the sunlit part, sufficient contrast is already between the object and the background. If correction processing is performed in the case where there is, an edge will be extracted from a slight shading in the background and the object. As described above, when extra edges other than the outline of the object are extracted, there is a problem that the possibility that these edges become disturbances and fail to detect the object increases.

  Further, if correction processing is performed when the target object is not included in the partial image, extra edges other than the contour of the target object are extracted excessively from slight shading in the background. As a result, there is a problem that these extra edges become disturbances and there is a high possibility that a part of the background is erroneously detected as an object.

  In the prior art, whether the partial image is a shaded part where a person wearing dark clothes is present, a shaded part where no person is present, or a sunny part where a person wearing dark clothes is present A distinction cannot be made. For this reason, the conventional technique has a problem in that the correction processing may cause excessive edges to lower the detection accuracy of the object.

  The present invention has been made to solve the above-described problem, and an object detection device capable of detecting an object with high accuracy by appropriately emphasizing the edge of the object's outline without causing excessive edges. The purpose is to provide.

  An object detection device according to the present invention is an object detection device that detects an object from an image in which a monitoring space is imaged, and is imaged when there is no object in advance and an imaging unit that images the monitoring space. A storage unit that stores the background image, and a luminance distribution analysis unit that analyzes the input image input from the imaging unit to calculate the input side luminance distribution and analyzes the background image to calculate the background side luminance distribution; Determining whether or not the input-side luminance distribution and the background-side luminance distribution are similar, and if similarities are determined, image correction means for generating a corrected image by performing contrast correction on the input image, and generating a corrected image Edge information is extracted from the corrected image when the correction image is generated, and edge information is extracted from the input image when the correction image is not generated, and the edge information extracted by the edge extraction unit is And object detecting means for detecting an object Zui, characterized by comprising a.

According to such a configuration, when there is a light-colored object in the shade and the like and a sufficient contrast has already been obtained between the background and the object, the contrast is corrected because similarity is not determined. If the edge information is extracted from the input image that is not present, and there is a dark object in the shade, etc., and the contrast between the background and the object is insufficient, the similarity is determined and contrast correction is performed. An edge image is extracted from the corrected image.
Therefore, if the contrast is already sufficient, correction is not performed to prevent excessive edges, and if the contrast is insufficient, correction can be performed to extract sufficient edge information to detect the object. Will improve.

In a preferred aspect of the present invention, the object detection device further includes candidate area setting means for setting a candidate area having a size set in advance according to the object at a predetermined position in the input image, and the luminance The distribution analysis means calculates an input side luminance distribution from the input image in the candidate area, and calculates a background side luminance distribution from the background image in the candidate area.
According to such a configuration, the input-side luminance distribution and the background-side luminance distribution are calculated in the minimum size area necessary for detecting the object. Therefore, it is possible to accurately determine similarity without being affected by the extra background portion.

In a preferred aspect of the present invention, the candidate area setting means compares the input image with the background image and sets a candidate area at a position where a difference is detected.
According to such a configuration, since only a region having a high probability that the object exists is set as a candidate region, it is possible to prevent contrast correction from being performed only in the background region, and it is possible to prevent the occurrence of excessive edges.

In a preferred aspect of the present invention, the candidate area setting means detects a human head area having an elliptical shape from the input image, and sets the candidate area at a position where the human head area is detected.
According to such a configuration, since only a region having a high probability that a human body as a target exists is set as a candidate region, it is possible to prevent contrast correction from being performed only in the background region, and it is possible to prevent the occurrence of excessive edges.

Further, in a preferred aspect of the present invention, the unevenness determination means for determining whether or not the background side luminance distribution is unevenly distributed in two different luminance ranges, and when the uneven distribution is determined, the background luminance distribution A region dividing means for detecting a first image region composed of pixels distributed in one of the luminance ranges and a second image region composed of pixels distributed in the other of the luminance ranges in the background luminance distribution; The luminance distribution analysis unit, the image correction unit, and the edge extraction unit individually process the first image region and the second image region.
When an object exists across the shade and the sun, a situation arises in which a sufficient contrast is obtained in one of the shade and the sun, but the contrast is insufficient on the other.
According to such a configuration, processing can be performed separately in the shaded area and the sunny area, so that in one area where sufficient contrast is obtained, excessive edge is prevented without performing contrast correction, and contrast is insufficient. In the other area, contrast correction can be performed to extract sufficient edge information.
Therefore, even if a region with sufficient contrast and a region with insufficient contrast are mixed, the detection accuracy of the object can be improved.

  According to the present invention, since the contrast correction is performed only when necessary based on the similarity determination between the input-side luminance distribution and the background-side luminance distribution, the edge between the background and the object can be emphasized without causing excessive edges, The detection accuracy of the object is improved.

1 is an overall configuration diagram of an image monitoring apparatus according to the present embodiment. It is a figure which shows an example of the process by the luminance distribution analysis means 42 and the image correction means 43. It is a figure which shows the process by the image correction means 43 in another example. It is a figure which shows the process by the luminance distribution analysis means 42 and the image correction means 43 in another example. It is a figure which shows the flowchart of an image monitoring process. It is a figure which shows the one part flowchart of a passer-by detection process. It is a figure which shows the flowchart of the remaining part of a passer-by detection process.

  As an example of a preferred embodiment of the present invention, an image monitoring apparatus will be described in which an entrance of a building is used as a monitoring space, a passerby passing through the monitoring space is detected as an object, and unauthorized traffic is monitored.

[Configuration of Image Monitoring Apparatus 1]
A functional block diagram of the image monitoring apparatus 1 is shown in FIG.
The image monitoring apparatus 1 includes an imaging unit 2, a storage unit 3, and an output unit 5 connected to a signal processing unit 4.

  The imaging unit 2 is a so-called surveillance camera. The imaging unit 2 sequentially outputs images obtained by imaging the monitoring space at predetermined time intervals to the signal processing unit 4. Hereinafter, the unit of time that is recorded at the above time interval is referred to as time. In the present embodiment, the imaging unit 2 captures and outputs a color image in which each RGB color has 256 gradations.

  The imaging unit 2 is installed on the ceiling of the entrance of a building with the optical axis directed vertically downward. Direct sunlight from the outside enters the entrance, and the shadows of the building structures are reflected. Therefore, the image picked up by the image pickup unit 2 includes a high-luminance sun part and a low-brightness shade part.

  The storage unit 3 is a memory device such as a ROM or a RAM. The storage unit 3 stores various programs and various data, and inputs and outputs such information to and from the signal processing unit 4. Various data includes a background image 30 and object identification information 31.

  The background image 30 is an image that includes only the background image of the monitoring space and does not include the passerby image. The background image 30 is generated and stored prior to the passer-by detection process.

  The object identification information 31 is reference information for identifying whether or not a passerby exists in the image. Prior to the passer-by detection process, the object identification information 31 is generated and stored based on edge information (luminance gradient information) extracted from at least a sample image in which the passer-by is captured.

As the edge information, a feature amount called HOG (Histograms of Oriented Gradients) can be adopted. The HOG feature amount is a parameter series including an edge angle distribution and an edge strength calculated in each of a plurality of small local areas set in the image. In the HOG feature amount, the edge strength is used as a weighting coefficient of the edge angle distribution. As a reference for HOG features, N.
Dalal, B. Triggs: Histograms
of oriented gradients for human detection: Proc. of IEEE Conference on Computer
Vision and Pattern Recognition (CVPR): 886-893 (2005).

  The object identification information 31 is a parameter representing an identification surface that discriminates passers-by from others in a feature space spanned by a multi-dimensional HOG feature. This parameter is used for supporting vector machines and boosting methods for HOG feature values extracted from a large number of sample images taken by passersby and HOG feature values extracted from a number of sample images not taken by passersby. It is learned by applying.

  In another embodiment, the object identification information 31 is generated as a passer's edge pattern that is obtained by averaging the edge intensities extracted from a plurality of sample images in which the passer is imaged. In yet another embodiment, the edge angle extracted from a plurality of sample images in which a passerby is imaged is generated as an edge pattern averaged for each pixel.

  The object identification information 31 thus generated mainly represents the characteristics of the passer's edge common to a plurality of sample images, that is, the outline (outer shape) characteristics of the passer.

  The signal processing unit 4 is an arithmetic device such as a DSP (Digital Signal Processor) or MCU (Micro Control Unit). The signal processing unit 4 describes the operations of the background image generation unit 40, the candidate area setting unit 41, the luminance distribution analysis unit 42, the image correction unit 43, the edge extraction unit 44, the object detection unit 45, the abnormality detection unit 46, and the like. It functions as each means by reading the program from the storage unit 3 and executing it. The signal processing unit 4 appropriately corrects an image input from the imaging unit 2 (hereinafter referred to as an input image) to detect a passer-by, and when detecting unauthorized traffic by tracking the detected passer-by, the signal processing unit 4 outputs to the output unit 5. An abnormal signal is output.

  The background image generation unit 40 stores the entire input image captured by the imaging unit 2 in the storage unit 3 as the background image 30 when there is no passerby in the monitoring space. In addition, the background image generation unit 40 updates the background image 30 by synthesizing the background image 30 with an image of a portion of the input image where no passer-by is detected, in order to adapt to illumination fluctuations.

  The candidate area setting unit 41 sets a partial area having the size of the passer's image in the input image as a candidate area, and sets information on the candidate area as a luminance distribution analysis unit 42, an image correction unit 43, an edge extraction unit 44, Output to the object detection means 45.

  The size of the passerby image changes depending on the positional relationship between the passerby and the imaging unit 2 or the like. Therefore, the candidate area setting means 41 sets the size of the passer's image using the position on the image as a variable. Specifically, a camera parameter such as the installation height of the image pickup unit 2 and the angle of view, and a three-dimensional model that approximates a standing human figure with a spheroid is applied to the pinhole camera model for monitoring. An image in which the three-dimensional model virtually arranged at each position in the space is projected on the image is calculated, and a rectangular frame surrounding the projected image is associated with the position of the projected image on the image and stored in the storage unit 3 in advance. Keep it. The candidate area setting means 41 reads out and sets a rectangular frame corresponding to the position where the candidate area is to be set as a candidate area.

  Here, the candidate area setting means 41 can also set candidate areas sequentially over the entire area of the image, but it is better to set candidate areas by narrowing down to positions where there is a high possibility that a passerby exists. The probability of false detection is reduced, and the processing cost is also reduced.

  Therefore, the candidate area setting unit 41 extracts a change area having a difference from the background image 30 in the input image, sets the candidate area at a position based on the change area, and an elliptical image is captured in the input image. The human head region is extracted and a candidate region is set at a position based on the human head region. In particular, by limiting the extraction range of the human head region to the change region, the accuracy of narrowing down can be increased, and the processing cost can be reduced.

  The change area can be extracted by a difference process between the input image and the background image 30 or a correlation calculation between the input image and the background image 30. The head region can be extracted by generating an edge image from the input image and performing a Hough transform on the generated edge image, or by matching the generated edge image with a preset elliptic pattern. The edge image can be generated by applying filtering such as a Sobel filter or a Canny filter to the input image.

  When a predetermined image region is designated by the image correction unit 43, the luminance distribution analyzing unit 42 calculates the characteristic amount of the luminance distribution of the input image and the background image 30 in the region, and calculates the calculated characteristic amount of the luminance distribution. It outputs to the image correction means 43. As a feature value of the luminance distribution, the luminance distribution analyzing means 42 calculates a luminance histogram of each image, detects a peak from the calculated luminance histogram, and calculates a luminance range of the peak.

  The luminance histogram is the frequency of pixels for each luminance. The luminance histogram can be calculated by counting the pixels constituting each image for each luminance. In this embodiment, each frequency is divided by the total number of pixels and normalized to a range of 0-1. Note that the input image and the background image 30 are RGB color images, but the luminance distribution analysis means 42 converts these into 256-tone monochrome images, and then calculates the characteristic amount of the luminance distribution.

  The mountain is a part where high-frequency luminance is dense. The luminance distribution analyzing means 42 applies the two peak detection thresholds Tg1, Tg2 (Tg1> Tg2) to the luminance histogram to detect the peak luminance distribution. That is, the luminance distribution analyzing means 42 identifies the luminance whose frequency exceeds Tg1, and detects the range where the frequency continuously exceeds Tg2 from the specified luminance to the low luminance side and the high luminance side as the luminance range of the mountain portion. To do. If there are a plurality of peaks, the same number of luminance ranges are detected. Each of Tg1 and Tg2 may be set to a value corresponding to about 20% or about 5% of the total number of pixels in the designated image area, for example.

  Hereinafter, the feature amount of the luminance distribution calculated for the input image is also referred to as an input-side luminance distribution, and the feature amount of the luminance distribution calculated for the background image 30 is also referred to as a background-side luminance distribution.

Here, in order to sufficiently obtain the effect of the contrast correction process, it is preferable that the distribution range of the peaks is sufficiently narrow. That is, it is better that the mountain portion is unevenly distributed.
Therefore, the luminance distribution analysis means 42 determines the uneven distribution of the peaks and detects only the uneven peaks. Specifically, the luminance distribution analyzing means 42 compares the width of the peak distribution range (luminance width) with the uneven distribution determination threshold value Tw, and detects only the peak portions whose distribution range is equal to or smaller than Tw. Output. Specifically, if the peak satisfies (Hb−Lb) ≧ Tw, the peak is detected. If (Hb−Lb) <Tw, the peak is not detected. For example, Tw may be set to a value of about one third of the gradation. In this example, 256 ÷ 3≈85 is set.

  The image correcting unit 43 performs a contrast correction process on the input image in the candidate area to generate a corrected image, and outputs the generated corrected image to the object detecting unit 45. The contrast correction process is a correction process for expanding a distribution biased in a narrow luminance range.

The image correction means 43 performs contrast correction processing by performing luminance conversion called histogram expansion (Histogram Spreading or Histogram Stretching). Specifically, the image correcting unit 43 converts the luminance I of the pixels of the input image whose luminance is not less than Li and not more than Hi into {(I−Li) × 255 ÷ (Hi−Li)}. However, the gradation of the input image is 0 to 255, and the luminance range of the peak portion of the input image is Li to Hi. In the corrected image, the luminance difference between the pixels at the peak and the surrounding pixels is increased, and the edge is emphasized.
At this time, the image correction unit 43 converts the luminance of the pixels of the input image whose luminance is less than Li to 0, and converts the luminance of the pixels of the input image whose luminance is higher than Hi to 255.

  In another embodiment, the image correction means 43 performs contrast correction processing by performing luminance conversion called histogram equalization. Specifically, the image correcting unit 43 converts the luminance I of the pixels of the input image whose luminance is not less than Li and not more than Hi into {255 ÷ N × ΣP (I)}. However, N represents the total number of pixels from Li to Hi, and ΣP (I) represents the sum of frequencies from luminance Li to I. This process also increases the brightness difference between the pixels in the crest and the surrounding pixels in the corrected image, and the edges are emphasized. Note that the conversion for pixels with luminance less than Li and for pixels with luminance higher than Hi is the same as described above.

  Here, in detecting a passerby based on edge information of the input image, there are the following trade-offs.

  First, if the contrast between the background and the passer is weak, edge information on the outline of the passer cannot be obtained sufficiently, and it becomes easy to fail to detect the passer. In this case, by performing the contrast correction process, the edge information of the contour of the passerby is emphasized, and it is possible to make it difficult to detect the passerby.

  On the other hand, however, if the contrast correction process is performed when the contrast between the background and the passer is sufficiently strong, even edge information other than the outline of the passer is emphasized, and extra edge information is extracted. The extra edge information is edge information that is extracted because a weak luminance difference such as a shadow or a pattern in the background area or the passer-by area is emphasized. Excessive edge information is a disturbance for the passer-by detection process, and causes a decrease in detection accuracy.

  Therefore, the image correction unit 43 determines whether or not to perform the contrast correction process, and performs the contrast correction process only when it is determined that the process should be performed. Therefore, the image correction unit 43 determines whether or not the input-side luminance distribution and the background-side luminance distribution are similar, and determines that the contrast correction processing should be performed if they are similar. Determines that the contrast correction processing should not be performed.

  Specifically, the image correcting unit 43 obtains the absolute value of the difference in frequency at each luminance between the luminance histogram calculated from the input image and the luminance histogram calculated from the background image 30, and sums them up to 2 minutes. The distance between the luminance histograms is calculated by multiplying by 1. If the distance is equal to or smaller than a preset similarity determination threshold Ts, it is determined to be similar, and if the distance is greater than Ts, it is determined to be different. For example, the similarity determination threshold value Ts may be set to a value corresponding to about 40% of the total number of pixels.

As a result, the contrast between the background and the object is weak, such as an input image obtained by capturing a dark-colored passerby in the shade, or an input image obtained by capturing a bright-colored passerby in the sun. Since the input image has a luminance distribution similar to that of the background image 30, it is determined that contrast correction processing should be performed.
And there is already enough contrast between the background and the object, such as an input image taken of a light-colored passerby in the shade or an input image taken of a dark-colored passerby in the sun. An input image that is strong against the background image 30 is determined not to be subjected to contrast correction processing because the luminance distribution is not similar to the background image 30.
Therefore, the edge information of the contour of the target object is emphasized for the input image with a low contrast between the background and the target object, and the extraction of the extra edge information is prevented for the input image that already has a sufficient contrast between the background and the target object. Therefore, the detection accuracy of the object is improved.

  Also, if a background-only area where no passerby exists is processed, similarity is determined and extra edge information is extracted. Therefore, the image correcting unit 43 processes the candidate region set by the candidate region setting unit 41, that is, the candidate region set in the position where the change area and the head region are extracted and the passerby exists with high probability. By doing so, it is avoided that contrast correction is performed on only the background region, and the occurrence of excessive edges is prevented.

Further, in order to perform an appropriate similarity determination without being affected by an extra background portion, it is desirable that a region consisting only of a passerby and a background in the vicinity thereof be a processing target. For this reason, the luminance distribution analyzing unit 42 and the image correcting unit 43 use the candidate area set by the candidate area setting unit 41, that is, the candidate area set in accordance with the size of the passer's image as a processing target. Appropriate luminance distribution analysis and similarity determination based on necessary and minimum image information consisting only of a person's image and a background image in the vicinity thereof are performed.
Furthermore, since contrast correction is performed based on the luminance distribution obtained in this way, it is possible to enhance the edge information of the outline of the passerby to the maximum extent.

  FIG. 2 schematically shows an example of processing by the luminance distribution analysis means 42 and the image correction means 43. In this example, a person in dark clothes is passing through the shade.

  An image 100 represents a background image in the candidate area. Due to the shade, the background image 100 is uniformly low in brightness. The luminance distribution analyzing means 42 analyzes the background image 100 and calculates a luminance histogram 101. In the luminance histogram 101, a unimodal distribution biased toward the low luminance side is formed. The luminance distribution analysis unit 42 applies the peak detection thresholds Tg1 and Tg2 to the luminance histogram 101, and calculates the minimum luminance value Lb and the maximum luminance value Hb (distribution range) of the peaks.

  The image 110 represents an input image in the same candidate area as the background image 100, and a passerby is imaged in the center. The luminance distribution analysis unit 42 analyzes the input image 110 and calculates a luminance histogram 111. Since the passerby is also dark, a unimodal distribution is formed in the luminance histogram 111 that is biased toward the low luminance side. The luminance distribution analyzing means 42 applies the peak detection thresholds Tg1 and Tg2 to the luminance histogram 111, and calculates the minimum luminance value Li and the maximum luminance value Hi (distribution range) of the peaks.

The image correcting unit 43 calculates the distance between the luminance histogram 101 and the luminance histogram 111 and compares it with the similarity determination threshold value Ts. In this example, since the distance is equal to or less than Ts, the image correction unit 43 determines similarity and performs contrast correction processing on the input image 110.
In this way, a corrected image 120 with improved contrast between the object and the background is generated. The converted luminance histogram is as shown in a graph 121.

  FIG. 3 illustrates another three cases. Processing of the image correction unit 43 in each case will be described.

  In a candidate area set for a passerby who is lightly dressed in the sun, images such as the background image 130 and the input image 131 are obtained. In this case, since the background side luminance distribution and the input side luminance distribution are similar, it is determined that contrast correction should be performed, and a corrected image is generated from the input image 131.

  In a candidate area set for a passerby who is lightly dressed in the shade, images such as the background image 140 and the input image 141 are obtained. In this case, since the background-side luminance distribution and the input-side luminance distribution are different, it is determined that contrast correction should not be performed, and a corrected image is not generated.

  In a candidate area set for a passerby in dark clothes who passes through the sun, images such as the background image 150 and the input image 151 are obtained. In this case, since the background-side luminance distribution and the input-side luminance distribution are different, it is determined that contrast correction should not be performed, and a corrected image is not generated.

  The case where the input image in the candidate area is uniformly dark or uniformly bright has been described above. However, the input image in the candidate area may include a boundary between a bright area (such as the sun) and a dark area (such as the shade). . In such a case, there is a mixed situation where the contrast is weak in the bright region but the contrast is sufficiently strong in the dark region, or the contrast is sufficiently strong in the bright region but the contrast is weak in the dark region. If this is applied, extra edge information is partially extracted, which causes a decrease in detection accuracy of the object.

  For this purpose, the image correction unit 43 includes a bimodal determination unit 431 and a light / dark region detection unit 432.

  The bimodal determination unit 431 refers to the background luminance distribution to determine whether the background luminance distribution is a bimodal distribution that is unevenly distributed in each of the high luminance region and the low luminance region. The bimodal determination unit 431 determines that the bimodal distribution is obtained when there are two pieces of information on the peaks included in the background side luminance distribution.

  The bright / dark area detecting means 432 detects a dark area composed of bright areas composed of pixels distributed in the high luminance area and dark areas composed of pixels distributed in the low luminance area in the background side luminance distribution for which the bimodal distribution is determined. The light / dark area detecting means 432 sets a division threshold value Td between two peaks, detects a pixel group composed of pixels having a luminance higher than Td as a bright area, and has a luminance of Td or less in the background side luminance distribution. A pixel group consisting of pixels is detected as a dark region.

  The image correcting unit 43 causes the luminance distribution analyzing unit 42 to analyze the input side luminance distribution and the background side luminance distribution of the bright region and the dark region, and determines whether it is necessary for each of the dark region and the bright region and contrasts as necessary. Perform correction processing.

  As a result, even in an input image in which a bright area and a dark area are mixed, a portion where the contrast between the background and the object is already sufficient while emphasizing the edge information of the area where the contrast between the background and the object is weak Since it is possible to prevent extraction of extra edge information from the object, the detection accuracy of the object is improved.

  FIG. 4 schematically shows an example of processing performed by the luminance distribution analysis unit 42 and the image correction unit 43 when a candidate area is set at the position of a dark-dressed passer-by image passing through the boundary between the sun and the shade. It is a thing.

  An image 200 represents a background image in the candidate area, a graph 201 represents a luminance histogram calculated from the background image 200, and an image 210 represents an input image in the candidate area.

  In the luminance histogram 201, a bimodal distribution biased toward the low luminance side and the high luminance side is formed. The luminance distribution analysis means 42 applies the peak detection thresholds Tg1 and Tg2 to the luminance histogram 101, and sets the minimum luminance value and the maximum luminance value (distribution range) of the two peaks as Lb1, Hb1, and Lb2, respectively. Hb2 is calculated.

  Since two peaks are detected, the image correcting unit 43 calculates the division threshold Td as {(Lb2-Hb1) / 2}, applies Td to the background image 200, and sets the candidate area as the bright area 202 and the dark area 202. Divide into regions 203.

  The luminance distribution analyzing unit 42 calculates a luminance histogram 221 of the background image 220 in the bright area 202 and a luminance histogram 241 of the input image 240 in the bright area 202. Since only the high-luminance background is captured in the background image 220, a single-peak distribution biased toward high luminance is formed in the luminance histogram 221. On the other hand, since the input image 240 captures a high-luminance background and a low-luminance passerby, the luminance histogram 241 forms a bimodal distribution that is biased toward high and low luminance. The image correcting unit 43 compares the luminance histogram 221 and the luminance histogram 241, but the similarity is not obtained and the contrast correction is not performed on the bright region 202.

  The luminance distribution analyzing unit 42 calculates the luminance histogram 231 of the background image 230 in the dark region 203, the luminance histogram 251 of the input image 250 in the dark region 203, and the peak distribution range. The image correcting unit 43 calculates the luminance histogram. 231 and the luminance histogram 251 are compared. In the dark region 203, similarity is obtained because the shaded background and the passersby in dark clothes have similar brightness distributions. The image correction unit 43 performs contrast correction processing on the input image 250 in the dark region 203 for which the similarity is determined.

The edge extracting unit 44 extracts edge information from the corrected image when the corrected image is generated by the image correcting unit 43, and extracts and extracts edge information from the input image when the corrected image is not generated. The edge information is output to the object detection means 45.
The edge information extracted by the object detection means 45 is the same type of edge information as the object identification information 31, and the above-described HOG feature value or edge pattern can be adopted.

  Here, when the image correction means 43 determines the bimodality and the bright area and the dark area are processed separately, the edge extraction means 44 extracts the edge information from each of the bright area and the dark area, and the bright area The edge information of the dark region and the edge information of the dark region are combined to combine the edge information of the entire candidate region, and the combined edge information is output to the object detection means 45.

  The object detection unit 45 detects the object by comparing the edge information input from the edge extraction unit 44 with the object identification information 31, and information on the input image in the candidate area where the object is detected, that is, the traffic Information about the human image is output to the abnormality detection means 46.

When the parameter of the identification plane in the HOG feature space is learned as the object identification information 31, the object detection unit 45 reads the parameter, sets it in a preset classifier, and uses the extracted HOG feature value. The comparison is performed by inputting to the discriminator. The discriminator discriminates whether or not the input HOG feature amount is in the space on the object side bounded by the discrimination plane in the feature space, and outputs the discrimination result.
When an edge pattern is created as the object identification information 31, the object detection means 45 performs the above comparison by reading the edge pattern and performing pattern matching with the extracted edge information.

  When the information on the passerby image is input from the object detection unit 45, the abnormality detection unit 46 extracts a feature amount such as a color histogram, a position, and a shape from the passerby image and stores it in the storage unit 3 and moves back and forth. By identifying the feature value extracted at the time, the passerby is tracked to detect unauthorized traffic. That is, the location of the building entrance and the location of the authentication device are set in the storage unit 3 in advance, the movement trajectory obtained as a result of tracking is compared with these positions, and the movement trajectory reaching the location of the building entrance is identified by the authentication device. If it does not go through the position of, it is determined that the traffic is illegal. When unauthorized traffic is detected, the abnormality detection means 46 outputs an abnormality signal to the output unit 5.

  When an abnormality signal is input from the abnormality detection means 46 of the signal processing unit 4, the output unit 5 is a security company through a sound device such as a buzzer or a speaker that performs sound output for warning of unauthorized traffic, and a communication line. Including a communication device that transmits an abnormal signal to a host computer installed in a security center in the building or a security room in the building, and outputs to the outside that unauthorized traffic has occurred.

[Operation of the image monitoring apparatus 1]
Hereinafter, the operation of the image monitoring apparatus 1 will be described with reference to FIG.

  When an administrator who confirms that the monitoring space is unmanned turns on the apparatus, each unit and each means are initialized and start operating (S1). The background image generation means 40 stores the input image being initialized in the storage unit 3 as the background image 30.

  After initialization, every time a new image is input from the imaging unit 2 to the signal processing unit 4, the processing of S2 to S8 is repeated.

  When a new image is input (S2), processing for detecting passers-by from the input image is performed (S3). Details of the passer-by detection process will be described below with reference to the flowchart of FIG.

  In steps S30 to S32, candidate areas are set by the candidate area setting means 41 of the signal processing unit 4.

  First, the candidate area setting means 41 performs a difference process between the background image 30 and the input image, and extracts a group of pixels having a difference equal to or greater than a preset difference threshold value as a change area (S30). As an extraction result, a mask image in which the pixel value in the change area is set to 1 and the pixel value outside the change area is set to 0 is generated.

  Next, the candidate area setting unit 41 generates an edge image from the input image, masks the edge image with the input mask image, removes the edge outside the change area, and performs the Hough transform on the masked edge image. Then, an ellipse is detected, and an area where the ellipse is detected is extracted as a head area (S31). As a result, a region where an elliptical image exists within the change region is extracted. The candidate area setting means 41 calculates the barycentric position of the head area, and reads out from the storage unit 3 a rectangular frame stored in association with the position as a candidate area (S32).

  If no change area is extracted or if a head area is not extracted, no candidate area is set. If the candidate area is not set, the signal processing unit 4 ends the passer detection process on the assumption that no passer is detected (NO in S33 to S4).

  On the other hand, when one or more candidate areas are set (YES in S33), signal processing unit 4 sequentially sets individual candidate areas as processing targets, and executes the loop process of steps S34 to S57.

  In the loop processing, first, the image correction unit 43 of the signal processing unit 4 designates a candidate region to be processed and causes the luminance distribution analysis unit 42 of the signal processing unit 4 to calculate the background side luminance distribution (S35). The luminance distribution analyzing means 42 outputs the luminance histogram of the background image 30 in the candidate region to be processed and the information on the peak portion determined to be unevenly distributed as the background side luminance distribution. The image correction unit 43 branches the process into three with reference to the number of information on the peaks.

  First, when there are two pieces of peak information detected in step S35, the background side luminance distribution is determined to be a bimodal distribution by the bimodal determination unit 431 of the image correction unit 43 (in S36). YES), the process proceeds to steps S43 to S53 in FIG. This process will be described later.

  Second, if there is only one peak information detected in step S35, the background side luminance distribution is assumed to be a unimodal distribution (YES in S37), and the processes in steps S38 to S42 are performed.

  The image correcting unit 43 designates a candidate area to be processed and causes the luminance distribution analyzing unit 42 to analyze the input side luminance distribution. The calculated input side luminance distribution and the background side luminance distribution calculated in step S35 are similar. It is determined whether or not to perform (S38, S39).

  When the similarity is determined, the image correction unit 43 performs a correction process for expanding the contrast on the input image in the candidate area to generate a correction image, and the edge extraction unit 44 extracts the HOG feature amount from the generated correction image. (YES in S39 → S40 → S41).

  On the other hand, when the difference is determined, the image correction unit 43 does not perform the correction process, and the object detection unit 45 extracts the HOG feature amount from the input image of the candidate area (NO in S39 → S42).

  Third, if the background-side luminance distribution calculated in step S35 is neither a bimodal distribution nor a unimodal distribution, the image correction unit 43 does not perform the correction process, and the edge extraction unit 44 from the input image of the candidate region. The HOG feature value is extracted (NO in S36 → NO in S37 → S42).

  With reference to FIG. 7, the process when the background side luminance distribution calculated in step S35 is a bimodal distribution will be described.

  In this case, the image correcting unit 43 detects the bright region and the dark region from the candidate regions by the light / dark region detecting unit 432 (S44), and sequentially sets each of the bright region and the dark region as a processing target, and loops of steps S43 to S53. Execute the process.

  In the loop processing, the image correcting unit 43 designates a bright region or dark region to be processed, causes the luminance distribution analyzing unit 42 to calculate the background side luminance distribution and the input side luminance distribution (S45, S46), and calculates the calculated background. It is determined whether or not the side luminance distribution and the input side luminance distribution are similar (S47).

  When the similarity is determined, the image correcting unit 43 performs a correction process for expanding the contrast on the input image in the bright region or dark region to be processed to generate a corrected image, and the edge extracting unit 44 uses the HOG feature from the corrected image. The amount is extracted (YES in S47 → S48 → S49).

  On the other hand, when the difference is determined, the image correction unit 43 does not perform the correction process, and the edge extraction unit 44 extracts the HOG feature amount from the input image in the bright region or dark region to be processed (NO in S47 → S51). ).

  When the HOG feature amount is extracted from each of the bright region and the dark region in this way, the object detection unit 45 combines the pieces of edge information into a HOG feature amount for the entire candidate region (YES in S52 → S53).

  When the HOG feature value is extracted as described above, the object detection unit 45 compares the extracted HOG feature value with the object identification information 31 in the storage unit 3 and a passerby exists in the candidate area. And the determination result is temporarily stored in the storage unit 3 in association with the information of the candidate area (S54).

  When the processing is completed for all candidate areas (YES in S55), the object detection unit 45 passes the input image in the candidate area, and the position and size of the candidate area where the passer-by is detected. Output as information.

  Returning to FIG. 6 again, the continuation of the image monitoring process will be described.

  When a passer-by is detected (YES in S4), the abnormality detection means 46 of the signal processing unit 4 associates the detected passer-by information with the passer-by information detected before the current time. By detecting the movement trajectory of each passerby, the presence of an abnormal passerby who has not stopped at the authentication device is determined based on the movement trajectory (S5).

  When an abnormality is determined (YES in S5), the abnormality detection means 46 outputs an abnormality signal to the output unit 5 (S7), and the output unit 5 to which the abnormality signal is input performs an output operation such as an alarm or a report. . If no passerby is detected (NO in S4), steps S5 to S7 are skipped, and if no abnormality is determined (NO in S6), step S7 is skipped.

  When the processing up to step S7 is completed, the background image generation means 40 of the signal processing unit 4 updates the background image 30 by synthesizing the background image 30 with an image of an area in which no passerby is detected in the input image ( S8).

[Modification]
In the above embodiment, the luminance distribution analyzing unit 42 calculates a luminance histogram, and the image correcting unit 43 determines the similarity of the luminance histogram.
In another embodiment, the luminance distribution analyzing means 42 counts the number of pixels N1 having a luminance equal to or higher than the upper limit luminance threshold TH1 and the number N2 of pixels having a luminance value equal to or lower than the lower limit luminance threshold TH2, and the image correcting means 43. The similarity is determined when the difference between N1 of the background side luminance distribution and N1 of the input side luminance distribution and the difference between N2 of the background side luminance distribution and N2 of the input side luminance distribution are both equal to or less than a predetermined value. The image correcting unit 43 performs luminance conversion similar to the expansion of the histogram on the luminance range of 0 to TH1 when N1 ≧ TN1 and N2 <TN1, and when N1 <TN1 and N2 ≧ TN1, Brightness conversion similar to the expansion of the histogram is performed on the 255 luminance range, and when N1 ≧ TN2 and N2 ≧ TN2, it is assumed that the distribution is a bimodal distribution, and the division threshold Td = (TH2-TH1) / 2・ Process by dividing the dark area.
TH1 can be set to about 85, TH2 can be set to about 170, TN1 can be set to a value corresponding to 50% or more of the total number of pixels, and TN2 can be set to a value corresponding to 30% or more of the total number of pixels.

  In still another embodiment, the luminance distribution analyzing unit 42 calculates the average A and the variance σ of the luminance, and the image correcting unit 43 calculates the difference between the background side luminance distribution A and the input side luminance distribution A and the background side luminance distribution. Similarity is determined when the difference between σ and σ of the input side luminance distribution is both equal to or smaller than a predetermined value. Then, the image correction means 43 performs luminance conversion similar to the expansion of the histogram on the luminance range of A ± α × σ when A ≦ TH1 and σ <Tσ1, or A ≧ TH2 and σ <Tσ1, and TH1 When <A <TH2 and σ ≧ Tσ2, it is assumed that the distribution is a bimodal distribution, and the processing is performed by dividing the bright / dark region by the division threshold Td = A.

In the embodiment described above, the luminance distribution analysis unit 42 determines the uneven distribution. In another embodiment, the unevenness determination is also performed when the image correction unit 43 makes a similar determination. In this case, the image correction unit 43 performs the contrast correction process only when similarity and uneven distribution are determined.

DESCRIPTION OF SYMBOLS 1 ... Image monitoring apparatus 2 ... Imaging part 3 ... Memory | storage part 4 ... Signal processing part 5 ... Output part 30 ... Background image 31 ... Object identification information 40 ... Background image generation means 41 ... Candidate area setting means 42 ... Luminance distribution analysis means 43 ... Image correction means 44 ... Edge extraction means 45 ... Object detection means 46 ... Anomaly detection means 431 ... Bimodal determination means 432 ... Bright / dark region detection means

Claims (5)

An object detection device that detects an object from an image obtained by imaging a monitoring space,
An imaging unit for imaging the monitoring space;
A storage unit that stores a background image captured when the object does not exist in advance;
A luminance distribution analyzing means for analyzing the input image input from the imaging unit to calculate an input side luminance distribution and analyzing the background image to calculate a background side luminance distribution;
Determining whether or not the input-side luminance distribution and the background-side luminance distribution are similar, and image correction means for generating a corrected image by performing contrast correction on the input image when similarity is determined;
Edge extraction means for extracting edge information from the corrected image when the corrected image is generated, and extracting edge information from the input image when the corrected image is not generated;
Object detection means for detecting the object based on edge information extracted by the edge extraction means;
An object detection apparatus comprising: Candidate area setting means for setting a candidate area of a size set in advance according to the object at a predetermined position in the input image;
The object according to claim 1, wherein the luminance distribution analysis unit calculates the input-side luminance distribution from the input image in the candidate area, and calculates the background-side luminance distribution from the background image in the candidate area. Detection device.   The object detection apparatus according to claim 2, wherein the candidate area setting unit sets the candidate area at a position where a difference is detected by comparing the input image with the background image.   The target object according to claim 2 or 3, wherein the candidate area setting unit detects a human head area having an elliptical shape from the input image, and sets the candidate area at a position where the human head area is detected. Detection device. Bimodal determination means for determining whether the background side luminance distribution is a bimodal distribution unevenly distributed in a high luminance region and a low luminance region;
When the bimodal distribution is determined, a bright region composed of pixels distributed in the high luminance region in the background luminance distribution and a dark region composed of pixels distributed in the low luminance region in the background luminance distribution. A light / dark area detecting means for detecting,
5. The object detection device according to claim 1, wherein the luminance distribution analysis unit, the image correction unit, and the edge extraction unit individually process the bright region and the dark region.