JP5012703B2 - Target detection system - Google Patents

Description

  The present invention relates to a technique for detecting a target pattern from an image, and more particularly to a pattern detection technique suitable for a panchromatic satellite image.

  Land observation and earth observation satellites such as QuickBird (DigitalGlobe, USA), ALOS “Daichi” (Japan AXA), and TERRA / ASTER (Japan ERSDAC) orbit around the Earth in a polar orbit at an altitude of about 450 km. It is equipped with sensors and synthetic aperture radar (SAR), continuously scanning the ground, and transmitting image data to the earth station on a regular basis. Three types of images are mainly used, and the optical sensor's multi-spectral mode (color photo) can be photographed at night in a 4-band configuration that adds the near-infrared to the RGB band of visible light. It is effective for collecting disasters and other situations. The SAR image is a radar image generated by radio waves and transmits even when it is cloudy. Therefore, the image can be collected even when the weather is bad, and a ship in operation can be measured using the Doppler effect (see Patent Document 1). . Among them, the panchromatic mode (monochrome photo) of the optical sensor has a resolution of less than 1 m (the world's highest resolution QuickBird 0.6 m) that can identify passenger cars on the ground, and is useful for monitoring traffic conditions. With regard to traffic monitoring, aerial photography with a surveillance camera, helicopter, or aircraft installed on the ground provides clearer images, but it is difficult to observe a wide area at the same time, and it is hidden behind buildings. In recent years, satellite images have attracted attention due to the problem that a blind spot always occurs.

  In addition, a target detection method using template matching that specializes in aircraft and ships (see Patent Document 2), a template is prepared in advance, and pixel blocks of the same size as the template are extracted while sequentially scanning from the first pixel of the image data Also, a method of matching with a template (see Patent Document 3) has been proposed.

However, the methods of Patent Document 2 and Patent Document 3 search the entire image, and thus have a problem that the search range is wide and the processing load is high.
Japanese Patent No. 3863014 Japanese Patent No. 3978979 JP 2001-67470 A

  However, the methods disclosed in Patent Document 2 and Patent Document 3 have a problem that the search range is wide and the processing load is high because the entire satellite image is searched. Usually, the search processing using template matching itself requires a considerable processing load, so it is desirable to narrow down the search range in advance if possible.

  Therefore, an object of the present invention is to provide a target detection system capable of accurately setting a search range and reducing a search processing load when a target made of a metal material is targeted.

  In order to solve the above problems, the present invention is a system for detecting a target from a search target image that is a search target using a template image including the target, and at least for a pixel block of a predetermined size. Pixels calculated by a predetermined pixel evaluation calculation formula for each pixel group or between different pixel groups, a pixel determination reference value corresponding to the pixel evaluation calculation formula, and a pixel evaluation calculation formula for two or more pixel groups Using the pixel evaluation rule in which a setting condition based on the relationship between the evaluation value and the pixel determination reference value is defined, the predetermined number of pixel blocks are sequentially extracted from the search target image, and the pixels in the pixel block are used to A pixel evaluation value is calculated by a pixel evaluation calculation formula, the pixel evaluation value is compared with the pixel determination reference value, and the extraction is performed according to whether a comparison result satisfies the setting condition. Search area setting means for setting a pixel in the pixel block as either a search target or a non-search target, and sequentially moving the extraction position from the start pixel position to the final pixel position with respect to the set search target area Then, a pixel block having the same size as each template image is extracted from the search target image, a similarity evaluation value between the template image and the pixel block is calculated, and the similarity evaluation value and the set determination criterion Similarity determination means for determining similarity by comparing with a value, and when the similarity determination means determines that there is similarity, the ID of the template image at that time, the pixel block on the search target image Provided is a target detection system having a search result output means for outputting a pixel position for specifying a position and a rotation angle of the template image. That.

  According to the present invention, in detecting a target from a search target image by comparing a template image including the target and a pixel block in the search target image, at least two pixel blocks of a predetermined size are detected in advance. Using the pixel evaluation rule in which a predetermined pixel evaluation calculation formula and a pixel determination reference value corresponding to the pixel evaluation calculation formula are defined for each pixel group or between different pixel groups, by defining the above pixel groups, In the search target image, an area where the target is highly likely to be detected is set as the search target area, and only the search target area is compared with the template image. It is possible to reduce the search processing load in the case of, and to prevent a false search from an area where there is no possibility that the target exists, and to improve the search accuracy.

  According to the present invention, the search range is accurately set, the processing load of the search when the target is made of a metal material is reduced, and the erroneous search of the target from outside the search range is suppressed and the target search is performed. The accuracy can be improved.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.
(1. System configuration)
First, the configuration of the target object detection system according to the present invention will be described. FIG. 1 is a configuration diagram of a target object detection system according to the present invention. In FIG. 1, 10 is a search target image storage unit, 20 is a template preparation unit, 30 is a template storage unit, 40 is a target search unit, 50 is a search result storage unit, and 60 is a search region setting unit.

  The search target image storage unit 10 stores a search target image that is an image to be searched and a sample image that stores a sample of the target. If there are many targets to be searched on the search target image, one of them can be used as a sample. In this case, the search target image also serves as a sample image. In the present embodiment, a satellite image that is an image taken by an artificial satellite is stored as a search target image. The template preparation unit 20 interactively sets a sample image to prepare a template image, and includes a template creation unit 21, a determination reference value setting unit 22, and a template update unit 23. The template storage unit 30 stores the template image prepared by the template preparation unit 20. The target searching unit 40 searches for a target by comparing a predetermined region in the search target image with the template image, and includes an angle-independent similarity determining unit 41, an angle-dependent similarity determining unit 42, A search result output unit 43, a detected target erasing unit 44, and a second template updating unit 45 are provided. The search result storage unit 50 stores the results searched by the target object search unit 40. The search area setting unit 60 sets a search target area that is a range to be actually searched in the search target image using a pixel evaluation rule applied to a pixel block of a predetermined size. Means 61 and search area setting means 62 are provided. The target object detection system shown in FIG. 1 is actually realized by incorporating a dedicated program into a general-purpose computer. Each storage unit is realized by a storage device such as a hard disk built in or connected to the computer.

(2. Setting search area)
Next, the processing operation of the target object detection system shown in FIG. 1 will be described. In the target object detection system, the search area setting unit 62 reads a search target image from the search target image storage unit 10, and an area where the target is likely to be detected on the search target image is set as a search target area. Set. In this system, as will be described later, the target is detected by comparing the pixel block on the template image and the pixel block on the search target image. In order to improve the processing efficiency, it is desirable to reduce the actual search range on the search target image as much as possible. Therefore, in the present system, before entering the actual search process, a process for setting an area where the target is likely to be detected as a search target area is performed.

The search target area is set by determining whether each pixel on the search target image is a search target area or a search asymmetric area. In the present invention, pixels to be applied to a pixel block of a predetermined size are used. Use evaluation rules. The shape and size of the pixel block can be determined as appropriate, but a square shape of M × M pixels is desirable. FIG. 2A is a diagram illustrating a pixel block to which the pixel evaluation rule in the present embodiment is applied. As shown in FIG. 2A, in this embodiment, 4 × 4 (M = 4) square pixel blocks are extracted from the search target image, and pixel evaluation rules are applied to the extracted pixel blocks. To do. The pixel evaluation rule is evaluated by using the pixel value (V _{i1 to} V _i4 ) of the center 2 × 2 pixels and the pixel values of the surrounding 12 pixels (V _{O1 to} V _O12 ) among the extracted pixel blocks. It is a rule that determines whether or not. Specifically, first, according to the following [Formula 1], the peripheral minimum value V _omin , the peripheral maximum value V _omax , the central minimum value V _imin , the central maximum value V _imax , the peripheral average value V _oave , and the central average value V _iave The peripheral standard deviation V _dir is obtained.

[Formula 1]
V _omin = MIN [V _ok ]
V _omax = MAX [V _ok ]
V _imin = MIN [V _ik ]
V _imax = MAX [V _ik ]
V _oave = (Σ _{k = 1,12} V _ok ) / 12
V _iave = (Σ _{k = 1,4} V _ik ) / 4
_{V dir = {Σ k = 1,12} (V oave -V ok) 2/12} 1/2

Further, using the seven threshold values S _aoi , S _doi , S _imin , S _imax , S _omin , S _omax , S _dir which are pixel determination reference values, a determination is made according to the following [Equation 2].

[Formula 2]
1) Average level difference around the center | V _oave −V _iave |> S _aoi
2) Contrast around the center When V _omax −V _imin > V _imax −V _omin , V _omax −V _imin > S _doi
When V _omax −V _imin <V _imax −V _omin , V _imax −V _omin > S _doi
3) Surrounding average level S _omin <V _oave <S _omax
4) If the median average level is V _oave > V _iave , then V _iave <S _imin
In the case of V _oave <V _iave , V _iave > S _imax
5) Peripheral standard deviation V _dir > S _dir

  The five conditions in the above [Equation 2] capture the characteristics of an object when a metal material is photographed. In other words, when a metal material is photographed, 1) the difference between the average level of the center and the periphery is greater than a predetermined value, 2) the difference between the maximum value of the periphery and the minimum value of the center, or the maximum value of the center and the minimum value of the periphery The difference becomes greater than or equal to a predetermined value. 3) The average level of the surrounding area falls within a predetermined range. 4) The central average level is less than a predetermined value when it is smaller than the peripheral average level. It becomes. The threshold value used for the central average level differs depending on the magnitude relationship with the peripheral average level. 5) The peripheral standard deviation is a predetermined value or more. In the case of a metal material, all of these conditions are considered to be satisfied.

  The seven pixel determination reference values are set as appropriate according to the specific content of the target. In the case of the central average level, the threshold value to be used differs depending on the magnitude relationship with the peripheral average level. The formula shown in [Formula 1], the pixel determination reference value, and the condition shown in [Formula 2] are set in advance as pixel evaluation rules.

The seven pixel determination reference values S _aoi , S _doi , S _imin , S _imax , S _omin , S _omax , S _dir can be set directly, but in this embodiment, a sample image of the target object Is used to set a value such that the sample is detected as a target. As the sample image, an image obtained by cutting out only a sample of the target may be used, or a search target image in which the target is shown may be used. Here, a case where the search target image is used as a sample image will be described as an example.

  In this case, when the system is activated, the pixel determination reference value setting unit 61 displays a search target image as a sample image on the screen and prompts the user to specify a target sample. FIG. 3 is a diagram illustrating an example of a search target image in which a target is shown. In the example of FIG. 3, one central dark part type and one central bright part type are shown. In the case of a metal material, the pixel values at the center and the periphery are different. In this case, a case where the center is darker than the periphery is referred to as a central dark portion type, and a portion where the center is brighter than the periphery is referred to as a central bright portion type. For example, a black automobile is a central dark part type, and a white automobile is a central bright part type.

A square cursor is displayed on the screen, and the user can move the screen using a mouse or the like. On the screen, a cursor is overlaid on each of the central dark part type and the central bright part type, but only one cursor is actually displayed. The size of the cursor is M × M pixels corresponding to the pixel evaluation rule. When the user moves the cursor to the central dark part type or the central bright part type and issues a confirmation instruction, the pixel determination reference value setting unit 61 extracts a pixel block of M × M pixels corresponding to the confirmed position. Then, the pixels in the extracted pixel block, by performing a process in accordance with the following [Equation 3], the pixel determination reference value _{S aoi, S doi, S imin} , S imax, S omin, S omax, Set S _dir .

[Formula 3]
S _aoi = MIN (| V _oave −V _iave |) · 0.9
When V _omax −V _imin > V _imax −V _omin , S ₁ = MIN (V _omax −V _imin )
When V _omax −V _imin <V _imax −V _omin , S ₂ = MIN (V _imax −V _omin )
S _aoi = MIN (S ₁ , S ₂ ) · 0.9
When V _oave > V _iave , S _imin = MAX (V _iave ) · 1.1
When V _oave <V _iave , S _imax = MIN (V _iave ) · 0.9
S _omin = MIN (V _oave ) · 0.9
S _omax = MAX (V _oave ) · 1.1
S _dir = MIN (V _dir ) · 0.9

A plurality of samples can be selected on the sample image. In the above [Equation 3], when a plurality of locations are selected, the minimum value and the maximum value calculated in each pixel block are MIN and MAX, respectively. When the pixel determination reference value is set according to the above [Equation 3], only one of S _imin and S _imax is set due to the magnitude relationship between the central average level V _oave and the peripheral average level V _iave. A determination reference value is set. In the above [Equation 3], in calculating any pixel determination reference value, after performing a predetermined calculation, an operation of multiplying 0.9 or 1.1 is performed. This is performed in order to make the extraction conditions slightly relaxed in order to detect the same target as the specified sample without omission.

  On the sample image, a plurality of samples can be specified at the same time. When a plurality of samples are specified at the same time, all of the target objects similar to the specified sample are detected, so that the conditions for the seven pixel determination reference values are the most lenient. Set the value.

The search area setting unit 62 sets the search area in a state where the pixel evaluation rule as described above is set. Specifically, a 4 × 4 pixel block having the pixel (0, 0) that is the origin of the search target image as a representative pixel is extracted. When the pixel (0, 0) is a representative pixel, 16 pixels from (0, 0) to (3, 3) as shown in FIG. 2B are extracted. Then, the search area setting means 62 applies the pixel evaluation rule according to the pixel arrangement of FIG. That is, the processing according to the above [Equation 1] is executed using all 16 pixels with the pixel value of the pixel (0, 0) as V _{O1 ,.} When the condition of [Expression 2] is determined and all of the conditions of [Expression 2] are satisfied, the representative pixel (0, 0) is set as a search target. If even one of the conditions of [Formula 2] is not satisfied, the representative pixel (0, 0) is set as a non-search target.

  When the setting for the pixel (0, 0) is completed, the search area setting unit 62 then converts the pixel (1, 0) as shown in FIG. 2C into a 4 × 4 pixel block. On the other hand, the pixel evaluation rule is applied to set the pixel (1, 0) as either a search target or a search non-target. Similarly, the search area setting means 62 performs a process of setting all pixels of the search target image as either a search target or a non-search target. When such processing is performed, the pixels belonging to the last three columns and three rows are not processed as representative pixels, but are only a part when viewed from the entire image, and the search region setting processing is This is not a problem because it is performed to reduce the load of the later search process.

  Information indicating it is recorded in the pixel set as the search area. There are various specific recording methods for determining whether or not each pixel of the search target image is a search region. In the present embodiment, 1 bit out of 8 bits representing the value of one pixel is used. . Therefore, the search area setting means 62 reads the original search target image expressed in 256 gradations per pixel, converts the values of 0 to 255 of each pixel into values of 0 to 127, and the 2nd to 8th bits. To store. Then, the search area setting means 62 sets a value indicating whether it is a search target area or a search non-target area regardless of the gradation in the first bit. Therefore, in this embodiment, the search target image is converted from 256 gradations to 128 gradations simultaneously with the setting of the search area.

  After setting all the pixels on the search target image as representative pixels and setting the search target or non-search target, the search area setting unit 62 performs a process of expanding the search target in units of pixel blocks. Specifically, for the representative pixel that is the search target, all 4 × 4 pixels included in the pixel block corresponding to the representative pixel are set as the search target. For example, when three pixels with dark shading as shown in FIG. 4A are to be searched, all pixels with light shading as shown in FIG. 4B are searched. Set as a target.

(3. Preparation of template image)
Next, preparation of a template image will be described. First, the search target image storage unit 10 stores a satellite image captured by an artificial satellite as a search target image, and further sets a search target region and a search non-target region as described above. In the present embodiment, the satellite image is in the Geo-TIFF format compliant with the Adobe TIFF standard, and the image data with geographic information (such as latitude / longitude information) added thereto is adopted. The satellite image shows objects that can be targets such as transportation facilities (aircraft, ships, automobiles, trains, etc.), buildings, traffic paths (roads, railways, etc.).

  In this state, the template preparation unit 20 converts the search target image stored in the search target image storage unit 10 from the source image format to the work image format. This is done to increase the search efficiency. Specifically, it is converted into a work image format using a known histogram flattening technique. In this embodiment, the work image is in the RXY format (raw file format in an extremely simple image file format compared to the TIFF format composed of only minimum header information such as the number of vertical and horizontal pixels and the number of bits of each pixel and the pixel arrangement). Also called). The template preparation unit 20 uses this search target image as a sample image.

  Subsequently, the template preparation unit 20 displays the search target image as the sample image converted into the work image format on the screen, and prompts the user to specify an image area for creating the template image. On the other hand, the user designates the contour of an image area having a polygonal shape (in this embodiment, a free rectangular shape including four vertices) in the search target image displayed on the screen. This is performed by designating four vertices of a square shape by operating the mouse. For example, when a car is an object, the user finds a car in a satellite image and designates an effective image area including the car. Specifically, the template creation means 21 displays a square of a square shape as an initial state on the screen, and the user selects one of the vertices on the screen, and drags the vertex to a desired position to make the vertex The same operation is sequentially performed on the four vertices.

  The quadrangle generated by the template creation means 21 on the screen is a double quadrangle, and the width of each side constituting the outer quadrangle and the side constituting the inner quadrangle is set to a predetermined value. Therefore, when the user moves the vertex of either the outer rectangle or the inner rectangle by dragging, the template creation means 21 moves accordingly so that the vertex of the other rectangle also keeps the distance of each piece of both rectangles. Let When the designation is made, as shown in FIG. 5 (a), the template creating means 21 circumscribes the square so as not to protrude even if the outer square is rotated 360 degrees outside the outer square (FIG. 5). (Outer square shape of (a)) is generated simultaneously. Therefore, in the initial state, a triple square is displayed on the screen, and when the user transforms the inner double square shape into a desired square, the outermost square shape maintains a square in conjunction with it. Change. Furthermore, in order to reduce the burden on the user, when the user deforms the inner quadrangular shape, the template creation means 21 interlocks with the outer quadrangular shape and the inner quadrangular shape outwards by a predetermined width. The offset shape is automatically defined. In other words, the user only needs to define only the inner rectangular shape, and the outer rectangular shape may be corrected to the automatically set shape.

  When the user completes designation of a predetermined quadrangular shape in the displayed search target image, the template creation means 21 is based on the square shape defined outside the quadrangular shape specified from within the search target image. Trim. Further, the template creation means 21 includes an outer mask area (mask gradation value 0) that is an outer portion of the outer rectangular shape designated by the user in the template image, and an inner mask that is between the outer rectangular shape and the inner rectangular shape. A mask value is set for each of the area (mask gradation value 1) and the effective image area (mask gradation value 2). At this time, the pixel gradation values in the inner mask area and the effective image area are not changed, and the pixel gradation values in the outer mask area are uniformly changed to zero. As a result, a square-sized template image is obtained, and the inside is composed of an effective image area, an inner mask area, and an outer mask area. Specifically, as shown in FIG. 5B, a mask value “2” is set for the effective image area, a mask value “1” is set for the inner mask area, and a mask value “0” is set for the outer mask area. Is done. Also, pixel values “0” to “127” are set in the effective image area and the inner mask area, but no pixel value is set in the outer mask area (all pixel values are cleared to “0”). . The reason why the pixel values of the effective image area and the inner mask area are set to “0” to “127” is to match the gradation of the search target image after the search target area is set.

  Furthermore, the template preparation unit 20 also registers in the template storage unit 30 a polygon shape and a pattern obtained by rotating only the inside of the polygon shape at a predetermined angle. In the present embodiment, the angle is rotated in units of 45 degrees, and template images according to eight different inclination angles are created. FIG. 6 shows eight template images having different angles. At this time, since the outer square shape is defined so that it does not protrude even if the polygon specified by the user is rotated 360 degrees, the polygons of the eight types of angles are represented by template images of the same size. can do. In this embodiment, patterns with different angles are registered with the same template ID as the same template image. Therefore, the template images shown in FIG. 6 all have the same template ID, and each pattern is specified by its angle. In this embodiment, eight template IDs can be registered. In this case, since eight template images are registered at eight angles, 64 patterns are registered in total.

  In the present embodiment, when matching processing between the template image and the search target image is performed, processing that does not depend on each angle pattern of the template image and processing that depends on each angle pattern are performed. Among these, the processing not depending on each angle pattern is executed on a common effective image area which is an effective image area common to all angle patterns. For example, in the case of a template image having an angle pattern as shown in FIG. 7A, the common effective image area is an area surrounded by inner squares in all angle patterns as shown in FIG. 7B. That is, the common effective image area is an area as shown in FIG. Actually, the template creation means 21 recognizes pixels belonging to the effective image area in each angle pattern (pixels having a mask value of “2”), and applies the mask value “3” to pixels belonging to the effective image area in all angle patterns. By assigning "", a common effective image area is set.

  When the template image is registered, the template preparation unit 20 prompts the user to specify an object to be matched on the search target image. When the user designates an object to be matched, the determination reference value is calculated and the template image is updated using the pixel values near the designated position. The calculation of the determination reference value and the update of the template image will be described with reference to the flowchart of FIG.

  First, the user looks at the contents of the search target image displayed on the screen, finds a target object that he wants to include in the search target, and designates a point on the target object (S101). FIG. 9A is a diagram illustrating a state of a search target image when an object is specified. When the user designates a point on the target object to be included in the search target image in the search target image, a square outer frame and a double polygonal shape of the template image are displayed as shown in FIG. Is done. For example, when it is desired to detect a car from a satellite image, a point on one car in the satellite image is designated.

  Subsequently, the template preparation unit 20 performs matching processing with the template image using the designated position as a base point (S102). The reason for this processing is that the point specified by the user on the screen is based on visual judgment as shown in FIG. 9A and is ambiguous. Therefore, the judgment reference value is calculated as it is based on this point. This is because there is a high possibility that an inappropriate search is performed. Therefore, in this process, the point designated by the user is corrected to an optimum position. Specifically, as shown in FIG. 9B, the base point (Xc, Yc) on the designated search target image and the center position of the template image are aligned, and from there, the template image is set to 1 in the x and y directions. Matching processing is performed while moving within a set range in pixel units. Details of this matching process will be described later. At the same position, matching processing is performed on eight types of template images with different angles, a template image having the highest similarity is selected, and the pixel position of the search target image is specified as the optimum position.

  Next, the determination reference value setting unit 22 sets a determination reference value so that the object on the search target image is determined to be compatible (S103). When the determination reference value is set, the template updating unit 23 combines the pixel block of the search target image at the optimum position and the template image, and updates the template image with the combined image (S104).

  Next, the details of the matching process in S102 will be described using the flowchart of FIG. Of the processes of S201 to S206 shown in FIG. 10, S201 and S202 are processes that do not depend on the angle of the polygon in the template image, and S203 to S206 are processes that depend on the angle of the polygon in the template image. . The processes of S201 and S202 that do not depend on the angle are executed for the common effective image area only once for one template image. However, the processes of S203 to S206 that depend on the angle are effective images of patterns of each angle. Since it is executed for a region, it is executed eight times for one template image. When performing the matching process, first, each variable that becomes the similarity evaluation value is initialized. Specifically, the minimum value of the difference of the luminance histogram (gradation histogram) independent of the angle, the minimum value of the difference of luminance dispersion (tone distribution) independent of the angle, and the minimum value of the difference of the luminance histogram dependent on the angle 6 variables (similarity evaluation values) are initialized: a minimum value of the difference in luminance dispersion depending on the angle, a minimum value of the gray gradation pixel difference, and a maximum value of the normalized correlation coefficient. In the present embodiment, all of these six variables take values from 0 to 1000, with respect to the normalized correlation coefficient, the initial value is the minimum value 0, and for the other five variables, the initial value is the maximum value 1000. To do.

  When the initialization of each variable is completed, the brightness histogram is checked without depending on the angle (S201). Specifically, first, the size of the template image is set to S × S, and the gradation value at the pixel (x, y) (0 ≦ x ≦ S−1, 0 ≦ y ≦ S−1) of the template image is set to Vt ( x, y) (0 to 127), the mask value is Mt (x, y) (0,1,2,3), and the position of the search target image (Xc + dx, Yc + dy) The image gradation value corresponding to the template position (x, y) is assumed to be V (x + Xc + dx, y + Yc + dy) (1-127). A 16-gradation luminance histogram H of the search target image based on the value of V (x, y) / 8 at the pixel (x, y) where Mt (x, y) = 3 within the range of S × S pixels. (v) A 16-tone luminance histogram Ht (v) (0 ≦ v ≦ 15) of the template image based on the values of (0 ≦ v ≦ 15) and Vt (x, y) / 8 is calculated. Furthermore, the process according to the following [Equation 4] is executed to calculate the luminance histogram difference, that is, the Euclidean distance between the luminance histograms.

[Formula 4]
(Luminance histogram difference) = 1000 · _{Σv = 0,15} | H (v) −Ht (v) | / _{Σv = 0,15} (H (v) + Ht (v))

  Subsequently, collation of luminance dispersion is performed (S202). Specifically, as in S201, the size of the template image is S × S, and C pixels (x, y) with Mt (x, y) = 3 are used within the range of S × S pixels. The average gradation value Va of the search target image and the average gradation value Vat of the template image are calculated by executing processing according to the following [Equation 5].

[Formula 5]
Va = Σy _{= 0, S-1} Σx _{= 0, S-1 | Mt (x, y) = 3} V (x + Xc + dx, y + Yc + dy) / C
Vat = Σy _{= 0, S-1} Σx _{= 0, S-1 | Mt (x, y) = 3} Vt (x, y) / C
C = Σy _{= 0, S-1} Σx _{= 0, S-1 | Mt (x, y) = 3} 1

  Then, using the pixel (x, y) where Mt (x, y) = 3 within the range of S × S pixels, the luminance dispersion value D of the search target image and the luminance dispersion value Dt of the template image are expressed as follows: Calculation is performed by executing processing according to [Formula 6].

[Formula 6]
D = Σy _{= 0, S-1} Σx _{= 0, S-1 | Mt (x, y) = 3} | V (x + Xc + dx, y + Yc + dy) -Va | / C
Dt = Σ _{y = 0, S-1} Σ _{x = 0, S-1 | Mt (x, y) = 3} | Vt (x, y) -Vat ｜ / C

  Then, within the range of S × S pixels, using the pixel (x, y) with Mt (x, y) = 3, the process according to the following [Formula 7], which is a well-known calculation formula, is executed. To calculate the luminance dispersion difference.

[Formula 7]
(Luminance dispersion difference) = 1000 · | D-Dt | / (D + Dt)

  As described above, S201 and S202 are processes that do not depend on the angle. Therefore, it is only necessary to perform once for each template image. The processing of S201 and S202 for the common effective image area may use pixel values of any angle pattern, but in the present embodiment, pixel values of a pattern of 0 degrees are representatively used. Since S203 to S206 from here are processes depending on the angle, the template image having the same ID is also processed for each angle. In the processing of S203 to S206, the pixels to be processed on the template image of each angle pattern are the pixels of the effective image region and the inner mask region.

  Next, the brightness histogram depending on the angle is collated (S203). Specifically, within the range of S × S pixels, the 16th floor of the search target image based on the value of V (x, y) / 8 at the pixel (x, y) where Mt (x, y)> 0. Luminance histogram H (v) (0 ≦ v ≦ 15) of tone and 16-gradation luminance histogram Ht (v) (0 ≦ v ≦ 15) of the template image based on the value of Vt (x, y) / 8 calculate. Further, the process according to the above [Equation 4] is executed to calculate the luminance histogram difference, that is, the Euclidean distance between the luminance histograms.

  Next, collation of luminance dispersion depending on the angle is performed (S204). Specifically, using C pixels (x, y) where Mt (x, y)> 0 within the range of S × S pixels, the average gradation value Va of the search target image and the template image The average gradation value Vat is calculated by executing the process according to [Formula 5]. Then, using the pixel (x, y) where Mt (x, y)> 0 within the range of S × S pixels, the process according to the above [Equation 7] is executed to calculate the luminance dispersion difference. .

  Next, a gray gradation pixel difference is calculated (S205). Specifically, in the same manner as described above, the size of the template image is S × S, and the gradation value at the pixel (x, y) (0 ≦ x ≦ S-1, 0 ≦ y ≦ S-1) of the template image Is Vt (x, y) (0 to 127), the mask value is Mt (x, y) (0,1,2,3), and the search target image position (Xc + dx, Yc + dy) is the starting point. An image gradation value corresponding to the template position (x, y) in the region is V (x + Xc + dx, y + Yc + dy) (0 to 127). Here, the following tone conversion is performed to raise the shadow side to the highlight side and drop the halftone portion to the shadow side. When Vt (x, y) ≥64, Vt '(x, y) = Vt (x, y) -64, when Vt (x, y) <64, Vt' (x, y) = 63-Vt (x, y). When V (x, y) ≥64, V '(x, y) = V (x, y) -64, when V (x, y) <64, V' (x, y) = 63-V (x, y). Further, within the range of S × S pixels, using C pixels (x, y) satisfying Mt (x, y)> 0, the processing according to the following [Equation 8] is executed, and the search target A gray gradation pixel difference that is a pixel difference between the image and the template image is calculated.

[Formula 8]
(Gray gradation pixel difference) = 1000 · Σ _{y = 0, S-1} Σx _{= 0, S-1 | Mt (x, y) ≠ 0} | V '(x + Xc + dx, y + Yc + dy ) −Vt ′ (x, y) | / (V ′ (x + Xc + dx, y + Yc + dy) + Vt ′ (x, y))

  Next, a normalized correlation coefficient is calculated (S206). Specifically, in the same manner as described above, the size of the template image is S × S, and the gradation value at the pixel (x, y) (0 ≦ x ≦ S-1, 0 ≦ y ≦ S-1) of the template image Is Vt (x, y) (0 to 127), the mask value is Mt (x, y) (0,1,2,3), and the search target image position (Xc + dx, Yc + dy) is the starting point. An image gradation value corresponding to the template position (x, y) in the region is V (x + Xc + dx, y + Yc + dy) (1 to 127). Within the range of S × S pixels, using C pixels (x, y) with Mt (x, y)> 0, the average gradation value Va of the search target image and the average gradation value Vat of the template image Is calculated by executing the processing according to the above [Equation 5] (where Mt (x, y)> 0).

  Then, within the range of S × S pixels, using a pixel (x, y) with Mt (x, y)> 0, a process according to the following [Formula 9], which is a well-known calculation formula, is executed. To calculate a normalized correlation coefficient. However, when the normalized correlation coefficient is a negative value (negative correlation), it is treated as 0 (no correlation).

[Formula 9]
(Normalized correlation coefficient) = 1000 · Σ _{y = 0, S-1} Σ _{x = 0, S-1 | Mt (x, y) ≠ 0} (V (x + Xc + dx, y + Yc + dy) -Va) ・ (Vt (x, y) -Vat) / [Σ _{y = 0, S-1} Σ _{x = 0, S-1 | Mt (x, y) ≠ 0} (V (x + Xc + dx, y + Yc + dy) -Va) ^2. Σ _{y = 0, S-1} Σ _{x = 0, S-1 | Mt (x, y) ≠ 0} (Vt (x, y) -Vat) ² ] ^1/2

  Subsequently, each variable is updated (S207). Specifically, the luminance histogram difference calculated in S201 is smaller than the minimum value of the currently set luminance histogram difference, and the normalized correlation coefficient calculated in S206 is the normalized correlation coefficient currently set. If greater than the maximum value, the minimum value of the luminance histogram difference independent of the angle, the minimum value of the luminance difference independent of the angle, the minimum value of the luminance histogram difference dependent on the angle, the minimum value of the luminance difference dependent on the angle, Six variables (similarity evaluation values) of the minimum value of the gray gradation pixel difference and the maximum value of the normalized correlation coefficient are replaced with the values calculated this time. Then, the position, template ID, and angle values are updated with the position, template ID, and angle values at this time.

  When the update of each variable is completed, it is determined whether or not the processing of all angle patterns has been completed for one template image (S208). If the processing has not been completed for all the angle patterns, the process returns to S203 to calculate the angle-dependent luminance histogram difference, the angle-dependent luminance difference, the gray gradation pixel difference, and the normalized correlation coefficient. . If the process is completed for all angle patterns, the process proceeds to S209. Therefore, in the present embodiment, the processing of S203 to S207 is repeated for eight angles for one template image.

  When processing is completed for all angles of one template image, it is determined whether or not processing has been completed for all template images (S209). If the processing has not been completed for all template images, the processing of S201 to S208 is repeated. If the processing is completed for all template images, the process proceeds to S210. Therefore, in the present embodiment, the processing of S201 to S208 is repeated for eight template images for one pixel position.

  When processing is completed for all template images, it is determined whether processing has been completed for all pixel positions (S210). All pixel positions mean the positions of pixels included in a predetermined pixel range in the x-axis and y-axis directions from the designated base point (Xc, Yc). As the predetermined pixel range, about 5% of each of the x-axis and y-axis directions of the template image is set. For example, when the template image is 100 × 100 pixels, the process shown in FIG. 9 is repeated for a total of 25 pixels of 5 × 5. If the processing has not been completed for all pixel positions, the processing of S201 to S209 is repeated. When the process is completed for all pixel positions, each variable is output and the matching process is terminated. Each variable is output for each template image.

  Next, details of setting the determination reference value in S103 will be described. The determination reference value is set by multiplying each variable output in the matching process by a predetermined value by the determination reference value setting means 22. In the present embodiment, for the normalized correlation coefficient, a value obtained by multiplying the maximum value of the output normalized correlation coefficient by 0.9 is used as a determination reference value, and for the other five variables, the output value A value obtained by multiplying 1.1 by 1.1 is used as a criterion value. Specifically, the calculation is performed by executing processing according to the following [Equation 10].

[Formula 10]
(Criteria value for angle-independent luminance histogram difference) = (minimum value of angle-independent luminance histogram difference) × 1.1
(Criteria value of angle-independent luminance dispersion difference) = (minimum value of angle-independent luminance dispersion difference) × 1.1
(Judgment reference value of angle-dependent luminance histogram difference) = (minimum value of angle-dependent luminance histogram difference) × 1.1
(Criteria value for angle-dependent luminance dispersion difference) = (minimum value of angle-dependent luminance dispersion difference) × 1.1
(Determination reference value of gray gradation pixel difference) = (Minimum value of gray gradation pixel difference) × 1.1
(Criteria value of normalized correlation coefficient) = (maximum value of normalized correlation coefficient) × 0.9

  Next, details of the template update in S104 will be described. In S104, the template update unit 23 updates each template image. Specifically, the image of each angle of each template image is updated using the pixel block on the search target image at the optimum position of each template image. First, among the template images, the one with the optimum angle and the pixel block on the search target image are synthesized. Since the outer mask area is not combined in the template image, only the inner mask area and the inside of the effective image area in the template image are combined.

  As a specific method of synthesis, an integer pseudo-random number is sequentially generated for each pixel in the inner mask area and the effective image area, and when a predetermined number is obtained, the corresponding pixel of the pixel block on the search target image is calculated. Replace with a value. For example, if an odd number appears, if it is set to replace the corresponding pixel block pixel on the search target image, the pixels in the inner mask area and the effective image area become pixels of the search target image with approximately half the probability. Replaced. FIG. 11 is a diagram illustrating an example in which the template image is updated by combining the search target image and the template image. In the example of FIG. 11, the image size is simplified to 5 × 5 pixels for convenience of explanation. FIG. 11B shows a template image, where T indicates a pixel inside the outer polygon (effective image area and inner mask area), and a blank indicates a pixel in the outer mask area. FIG. 11A shows a pixel block on the search target image determined to be the optimum position with respect to the template image, and W indicates a pixel on the search target image. As a result of performing the above-described combining processing on these two images, as shown in FIG. 11C, about half of the pixels T in the inner mask area and the effective image area of the template image are pixels W of the search target image. A composite image replaced with is obtained. This composite image is registered as an updated template image. Note that the seeds are changed each time so that the random number generation method becomes a different series each time processing is performed.

  When the update process is completed for the optimum angle pattern, the other angle patterns are also updated. For other angle patterns, obtain the difference between the angle and the optimum angle, rotate the pixel block on the search target image by that angle, and rotate the pixel unit inside the effective image area to rotate the pixel block on the search target image. Replace with the corresponding pixel. In this way, the template image is updated for all angles. All template images are updated by performing the same process on all template images prepared in advance. FIG. 12 is a diagram showing how the template image is updated. As shown in FIG. 12, the image of all angles is updated using the pixel block at the optimum position on the search target image.

(4. Search process)
As described above, when the preparation of the template image is completed, the target object search unit 40 searches for the target object using the prepared template image. This target searching process will be described with reference to the flowchart of FIG. When the user instructs the target searching process, the angle-independent similarity determining unit 41 and the angle-dependent similarity determining unit 42 of the target searching unit 40 cooperate to sequentially perform the matching process (S301). Specifically, pattern matching processing is sequentially executed from the start pixel position in the search target image to find a position for determining similarity.

  FIG. 14 is a diagram showing an outline of sequential matching processing. In the sequential matching process, first, S × S square pixel blocks from the start pixel position (0, 0) in the search target image to (S−1, S−1), the template image, The comparison is performed by moving one pixel at a time in the X-direction scan range, such as a square region starting from (1, 0), a square region starting from (2, 0), and so on. When the x-axis direction is finished, the entire search target image is sequentially moved by one pixel in the Y-direction scan range, and a square area starting from (0, 1), a square area starting from (1, 1), and so on. Compare with the template image.

  The matching processing at each pixel position includes angle-independent template matching for matching with any one of the eight angle patterns of the template image with the same ID, and all eight of the template images with the same ID. This is divided into angle-dependent template matching for matching with individual angle patterns. If it is determined that the angle-independent template matching is incompatible, the position shifts to the next pixel position without performing the angle-dependent template matching.

  The details of the sequential matching process when the pixel position is specified are shown in the flowchart of FIG. First, as in the matching process in S102, each variable that becomes a similarity evaluation value is initialized. Specifically, the minimum value of the luminance histogram difference independent of the angle, the minimum value of the luminance difference independent of the angle, the minimum value of the difference of the luminance histogram dependent on the angle, the minimum value of the luminance difference dependent on the angle, gray Six variables (similarity evaluation values) of the minimum value of the gradation pixel difference and the maximum value of the normalized correlation coefficient are initialized. All of the six similarity evaluation values used in the search process also take a value of 0 to 1000. For the normalized correlation coefficient, the initial value is set to the minimum value 0, and for the other five variables, the initial value is set to the maximum value 1000. And

  When the similarity evaluation value is initialized, first, the search target area at the pixel position is confirmed (S400). Since the first pixel position is (0, 0), it is confirmed whether or not each of S × S pixels in the range of (S−1, S−1) is set as a search target region. In this embodiment, since the search area is set by the value of the first bit of each pixel, whether this is a search area or a non-search area is determined by checking this value. When a majority of S × S pixels starting from (0, 0) are search non-target regions (number of pixels whose first bit is 0> S × S / 2), the next pixel position (1 , 0) (nonconformity end), and the search target area in S400 is confirmed. By the process in S400, the process after S401 is not performed for the search non-target area, and the efficiency of the search process as a whole is improved.

  In S400, when there is a pixel set as a search target region even in one pixel, the luminance histogram is collated (S401). Specifically, as in S201 above, the process according to [Formula 4] is executed to calculate the luminance histogram difference. Then, the luminance histogram difference is compared with the determination reference value set in S103, and if it is larger than the determination reference value, it is not suitable. If it is equal to the determination reference value, it may be set to be either, but in this embodiment, it is set to be determined as non-conforming. The same applies to other similarity evaluation values. Nonconformity means that there is no similarity. If it does not match, it is determined that the currently used template image does not match, and it is determined whether or not the processing with all template images has been completed (S411). If processing with all template images has not been completed, the template image is changed to another template image (S412), and the process returns to S400.

  If the luminance histogram difference is smaller than the determination reference value in S401, the luminance dispersion is checked (S402). Specifically, as in S202 above, the process according to [Formula 7] is executed to calculate the luminance dispersion difference. Then, the luminance dispersion difference is compared with the determination reference value set in S103. Processing in the case of nonconformity is the same as that in the case of nonconformity in S401. The processes in S401 and S402 are performed on the common effective image area, similar to the processes in S201 and S202. Since the processes of S401 and S402 are processes that do not depend on the angle, if they do not match even once after collating with each template image, the process proceeds to a process with another template image.

  In S402, when the luminance dispersion difference is smaller than the determination reference value, collation of the luminance histogram depending on the angle is performed (S403). Specifically, as in S401, the process according to [Formula 4] is executed to calculate the luminance histogram difference. However, unlike the processing in S401 and S402, the same template image is processed for all angles. The luminance histogram depending on the angle is updated to the minimum one, and the angle at that time is recorded. When the calculation of the luminance histogram difference is completed for all angles (S404), the minimum luminance histogram difference is compared with the determination reference value set in S103, and the luminance histogram difference is equal to or greater than the determination reference value. In that case, it becomes nonconforming. Processing in the case of nonconformity is the same as that in the case of nonconformity in S401.

  When the luminance histogram depending on the angle is matched, the luminance distribution depending on the angle is collated (S405). Specifically, similarly to S402, the process according to [Formula 7] is executed to calculate the luminance dispersion difference. However, unlike the processing in S401 and S402, the same template image is processed for all angles. The luminance dispersion difference depending on the angle is updated to the minimum, and the angle at that time is recorded. When the calculation of the luminance dispersion difference is completed for all angles (S406), the minimum luminance dispersion difference is compared with the determination reference value set in S103, and the luminance dispersion difference is greater than or equal to the determination reference value. In that case, it becomes nonconforming. Processing in the case of nonconformity is the same as that in the case of nonconformity in S401.

  When the luminance histogram depending on the angle is matched, the gray gradation pixel difference is calculated (S407). Specifically, similarly to S205 described above, processing according to [Equation 8] is executed to calculate the gray gradation pixel difference. Also in the gray tone pixel difference calculation process, the same template image is processed for all angles. The gray gradation pixel difference is updated to the minimum, and the angle at that time is recorded. When the calculation of the gray gradation pixel difference is completed for all angles (S408), the gray gradation pixel difference that has been minimized is compared with the determination reference value set in S103, and the gray gradation pixel difference is determined. If is more than the criterion value, it is not conforming. Processing in the case of nonconformity is the same as that in the case of nonconformity in S401.

  If the gray gradation pixel difference is met, a normalized correlation coefficient is calculated (S409). Specifically, the normalized correlation coefficient is calculated by executing the processing according to [Equation 5] and [Equation 9] as in S206. In the normalized correlation coefficient calculation process, the same template image is processed for all angles. The normalized correlation coefficient is updated to the maximum value, and the angle at that time is recorded. When calculation of the normalized correlation coefficient is completed for all angles (S410), the normalized normalized correlation coefficient is compared with the criterion value set in S103, and the normalized correlation coefficient is calculated. Is greater than the criterion value, it is compliant, and when the normalized correlation coefficient is less than the criterion value, it is not compliant.

  When the processing is completed for all template images (S411), it is determined whether or not there is a template in which all items, that is, all six similarity evaluation values are suitable (S413). If there is a template in which all items are compatible, it is determined whether or not the angle coincides with the template ID when the gray gradation pixel difference is minimum and when the normalized correlation coefficient is maximum (S414). . That is, the template and angle that give the minimum value in the calculation processing of the gray gradation pixel difference in S407 and S408 and the template and angle that give the maximum value in the calculation processing of the normalized correlation coefficient in S409 and S410 respectively match. If it is determined in S401, S402, S403, S405, S407, and S409 that the matching templates are all matched, matching ends, and each variable, that is, template ID, pixel position (Xc, Yc), angle, and six similarities. The sex evaluation value is output. In any case, in the case of non-conformity, the non-conformance ends, and matching processing is sequentially performed on the next pixel position.

  In the sequential matching process shown in FIG. 15, the search area setting unit 62 confirms the search target area in S400, the angle-independent similarity determination unit 41 collates the luminance histogram independent of the angle in S401, and the angle in S402. Brightness-dependent collation that does not depend on the angle, the angle-dependent similarity determination means 42 collates the brightness histogram that depends on the angle in S403, the brightness dispersion that depends on the angle in S405, the calculation of the gray-scale pixel difference in S407, S409 A total of seven suitability determination processes for calculating the normalized correlation coefficient are performed, and these suitability determination processes are set so as to be performed in order from the smallest processing load. In other words, in the present embodiment, checking the search target area in S400 has the smallest processing load, followed by matching of the brightness histogram that does not depend on the angle has the second smallest processing load, and the calculation of the normalized correlation coefficient is the most processed. The load is large. This is because, when it is determined that the non-conformity is determined in one process, it is more efficient to perform the process starting from the one with the smallest processing load if it is determined as dissimilar as a whole. In particular, the search target area in S400 does not need to be checked against the template, and it is only necessary to count the pixels set by the search area setting means 62 in the area that is the target of the matching process in sequence. Since the processing load is smaller than that of any of the sequential matching processes, and the ratio of pixels set by the search area setting unit 62 is generally 10% or less, the subsequent sequential matching processes are not executed in an area of more than 90%. Can be judged as non-conforming. Subsequently, in the matching process, the angle-dependent luminance histogram matching, the angle-dependent luminance dispersion matching, the gray-scale pixel difference calculation, and the normalized correlation coefficient calculation are executed for each angle. For this reason, the processing load is significantly increased as compared to the luminance histogram matching that is executed for only one angle and the luminance dispersion matching that is independent of the angle. In addition, among the processes depending on the angle, a process for calculating and collating one value for the entire image such as a luminance histogram and luminance variance is performed in units of pixels such as a gray gradation pixel difference and a normalized correlation coefficient. Since the processing load is small compared to the processing of collating with the above, the processing load is reduced as a whole because the processing load is performed first.

  When the sequential matching process in S301 is completed, the target object search unit 40 performs a local matching process (S302). The sequential matching process in S301 searches for a rough position of the target in the entire search target image, whereas the local matching process in S302 searches for a detailed position of the target. Specifically, template matching is performed by moving a predetermined minute width (ΔX, ΔY) in the pixel scan progression direction (rightward, downward) starting from the position (Xc, Yc) determined to be optimal by sequential matching processing. And the six similarity evaluation values are updated to the minimum value and the maximum value, and the template ID, pixel position (Xm, Ym), and angle Am at that time are output. The details of the matching process are the same as the matching process in S102 of FIG. 8, specifically the process shown in FIG. 10, unlike the sequential matching process shown in FIG. The sequential matching process only has to determine whether or not the matching is performed at a certain position (Xc, Yc), but the local matching process searches for a matching position under the best condition within a minute range in which the matching is determined. This is necessary.

  When the local matching process is completed, the search result output unit 43 of the target object search unit 40 outputs the search result (S303). Specifically, the template ID, position (Xm, Ym), angle Am, and six similarity evaluation values are output as a list.

  Subsequently, the detection target erasing unit 44 erases the detection area (S304). Specifically, in the search target image, a portion corresponding to the inside of the effective image area (inside the inner polygon) shown in FIG. 5 of the template image at the optimum position is deleted. That is, the values of the pixels on the search target image located inside the effective image area are all set to 0 (the first bit indicating the search target is not changed and the bit values of the second to eighth bits are all 0). FIG. 16 is a diagram illustrating the optimum position of the search result on the search target image and the erased area. As a result of the search, when the position as shown in FIG. 16A is the optimum position of the inner polygon and the outer polygon of the template image with respect to the target on the search target image, FIG. As shown, the portion corresponding to the effective image area (inside the inner polygon) on the search target image is deleted. In FIG. 16B, the outer polygon is indicated by a dotted line. However, if the inner mask area is deleted in addition to the inside of the outer polygon, that is, the effective image area, a part of the adjacent target is deleted. Therefore, it is difficult to detect an adjacent target. Thus, even when the region corresponding to the inside of the inner polygon of the template image is deleted, in order to respond to a request to graphically display the search position on the search target image after the end of the search process, the target object search unit 40 Backup the original search target image before erasure.

  The processing of S301 to S304 is performed on all pixels on the search target area. At this time, since the area corresponding to the inside of the effective image area is deleted in S304, the similarity determination is not performed even if the matching processing is sequentially performed on the neighboring pixels, and one target is overlapped. Can be prevented. On the other hand, since the area on the search target image corresponding to the inner mask area used in the angle-dependent template matching is not deleted, it is possible to prevent the adjacent target from being deleted.

  The result output in S303 is output as a search result by the search result output means 43 as described above. Specifically, the position / direction of the detected object is output as a list. In this case, for the position, the pixel position is converted into information on latitude and longitude held in the satellite image and output. It is also possible to graphically indicate the position of the target object detected on the search image. In this case, since all the searched parts are deleted, it is necessary to reload the original search target image stored in the backup. In addition, the outline of the inner polygonal shape is rotated at a specified position at a specified angle and displayed as an overlay, and the detected position / direction of the target object is clearly indicated. Even if the detection area erasing process of S304 is performed, if there are overlapping results, an operation is performed in which one is manually deleted and the others are deleted.

  The search result output means 43 can output search results in a form other than list output. In the present embodiment, a list output is performed, and an output in which an inner rectangular shape corresponding to the target detection position is arranged on the image is also performed. FIG. 17 is a diagram illustrating an example of a searched screen output by the search result output unit 43. A large number of rectangular frames correspond to the inner rectangular shape of the template image, and indicate locations where the target is detected. In addition, dark shading indicates a search target area, and the other indicates a search non-target area. As is clear from FIG. 17, the rectangular frame exists only in the search target region. This is because the search is not performed in the search non-target region. In practice, the search result output means 43 performs a process of outputting the search target area in red and outputting a rectangular frame in green in order to perform color output.

  In the above embodiment, the case where the template image is used without being updated at the time of searching has been described. However, at the time of searching, the template image is updated as in the search preparation stage. Also good. In this case, after the local matching process in S302, the second template updating unit 45 executes the same process as in S104. Specifically, the image at each angle of the template image specified by the template ID determined in S302 is updated using the pixel block at the pixel position (Xm, Ym) determined in S302. Here, as in S104, first, among the template images, the optimal angle determined in S302 and the pixel block on the search target image are synthesized. Since the mask area is not combined in the template image, only the effective image area and the inner mask area in the template image are combined.

  As a specific method of synthesis, as described with reference to FIG. 11, as in S104, an integer pseudorandom number is sequentially generated for each pixel inside the outer polygonal shape, and when a predetermined number is obtained, a search is performed. Replace with the value of the corresponding pixel in the pixel block on the target image. When the update process for the optimum angle pattern is completed, the second template update unit 45 updates the other angle patterns as in S104. In the present embodiment, the second template update unit 45 functions as a part of the target object search unit 40, but its specific function is the same as the template update unit 23 in the template preparation unit 20. Actually provided as a software module. Therefore, after the processing by the angle-independent similarity determination unit 41 and the angle-dependent similarity determination unit 42, the template update unit 23 may execute a template image update process according to an instruction from the target object search unit 40. good. In this case, the template update unit 23 functions as a second template update unit.

It is a block diagram of the target object detection system which concerns on this invention. It is a figure which shows the pixel block extracted from the pixel block to which a pixel evaluation rule is applied, and a search object image. It is a figure which shows an example of the search object image in which the target object is reflected. It is a figure which shows the process which expands the pixel of search object. It is a figure which shows the relationship between a template image area | region and a double polygon shape. It is a figure which shows the template image which has eight angle patterns. It is a figure which shows the relationship between a template image, the effective image area | region of each angle, and a common effective image area | region. It is a flowchart which shows the process outline | summary of the calculation of a determination reference value, and the update of a template image. It is a figure which shows the mode of the process on a search object (sample) image. It is a flowchart which shows the detail of a matching process. It is a figure which shows the example which updates a template image by the synthesis | combination of a search object (sample) image and a template image. It is a figure which shows the mode of the update of a template image. It is a flowchart which shows the outline | summary of a search process. It is a figure which shows the outline of a sequential matching process. It is a flowchart which shows the detail of a sequential matching process. It is a figure which shows the optimal position of the result of the search on a search object image, and the erased area. It is a figure which shows an example of the searched screen which the search result output means 43 output.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 ... Search object image storage part 20 ... Template preparation part 21 ... Template preparation means 22 ... Determination reference value setting means 23 ... Template update means 30 ... Template storage part 40 ... Target search unit 41... Angle-independent similarity determination means 42... Angle-dependent similarity determination means 43... Search result output means 44. Update means 50 ... search result storage section 60 ... search area setting section 61 ... pixel judgment reference value setting means 62 ... search area setting means

Claims (15)

A system for detecting a target from a search target image, which is a search target, using a template image including the target,
At least two or more pixel groups are defined for a pixel block of a predetermined size, a predetermined pixel evaluation calculation formula for each pixel group or between different pixel groups, and a pixel determination reference value corresponding to the pixel evaluation calculation formula The predetermined number of pixel blocks are sequentially extracted from the search target image using a pixel evaluation rule in which a setting condition based on a relationship between a pixel evaluation value calculated by a pixel evaluation calculation formula and a pixel determination reference value is defined. The pixel evaluation value is calculated by the pixel evaluation calculation formula using the pixel in the pixel block, the pixel evaluation value is compared with the pixel determination reference value, and whether the comparison result satisfies the setting condition, Search area setting means for setting a pixel in the extracted pixel block to either a search target or a search non-target;
Extracting a block of pixels having the same size as each of the template images from the search target image while sequentially moving the extraction position from the start pixel position to the final pixel position with respect to the set search target region, Similarity determination means for calculating a similarity evaluation value of the pixel block and determining the similarity by comparing the similarity evaluation value with the set determination reference value;
A search for outputting the template image ID, the pixel position for specifying the position of the pixel block on the search target image, and the rotation angle of the template image when the similarity determination unit determines that there is similarity. A result output means;
A target detection system comprising: In claim 1,
The search area setting means sets all pixels in the extracted pixel block as search targets when a predetermined pixel in the pixel block extracted from the search target image is a search target. Target detection system. In claim 1,
The pixel block has a square shape of M × M pixels, and the search area setting means includes a central pixel group of the center (M−2) × (M−2) pixels and the remaining surroundings as the pixel group. Two peripheral pixel groups are defined, and the minimum, maximum, and average values of the central pixel group are V _imin , V _imax , V _iave , and the minimum, maximum, and average values of the peripheral pixel group are V _omin , V _omax, when formed into a V _OAVE, the pixel evaluated value _{V iave, V oave, V imax} -V omin, V omax -V imin, calculates _V iave -V oave, pixel evaluation value V _Iave, the V _OAVE is Each of the two types of pixel determination reference values, pixel evaluation values V _imax −V _omin , V _omax −V _imin , and V _iave −V _oave is characterized in that one type of pixel determination reference value is set. Target detection system. In any one of Claims 1-3,
The pixel block of the predetermined size corresponding to the position specified in the sample image in which the sample of the target object is shown is extracted, a pixel evaluation value is calculated for the pixel block, and a predetermined calculation is performed on the pixel evaluation value. A target object detection system, further comprising: a pixel determination reference value setting unit that calculates a pixel determination reference value by applying the pixel determination reference value. In any one of Claims 1-4,
The similarity determination means includes
Extracting a pixel block of the same size as each template image from the search target image while sequentially moving the extraction position from the start pixel position to the final pixel position with respect to the search target image, and rotating the angle of the template image Without comparing, the pixel included in the template image and the corresponding pixel in the pixel block are collated with a predetermined collation method to calculate a similarity evaluation value that does not depend on the angle, Angle-independent similarity determination means for determining similarity by comparing with a determination reference value set to
When the angle-independent similarity determining unit determines that there is similarity, the template image is rotated by a predetermined angle, and the pixels included in the template image and the corresponding pixels in the pixel block are An angle-dependent similarity determination unit that calculates a similarity evaluation value by collating with a predetermined collation method, and determines similarity by comparing with a preset criterion value;
A target detection system comprising: In claim 5,
The angle-independent similarity determination means determines a similarity corresponding to each similarity evaluation value by calculating a plurality of similarity evaluation values and comparing each similarity evaluation value with a corresponding criterion value. A target detection system characterized in that if there is no similarity in any one of the similarity evaluation values, it is determined that there is no similarity. In claim 5 or claim 6,
The angle-dependent similarity determination means determines a similarity corresponding to each similarity evaluation value by calculating a plurality of similarity evaluation values and comparing each similarity evaluation value with a corresponding criterion value, A target object detection system characterized in that if there is no similarity in any one of the similarity evaluation values, it is determined that there is no similarity. In any one of Claims 5-7,
The angle-independent similarity determination means calculates a luminance histogram for each of the pixels included in the template image and the corresponding pixels in the pixel block, and calculates the Euclidean distance between the two luminance histograms as the similarity evaluation. A target detection system characterized by being used as a value. In any one of Claims 5-8,
The angle-independent similarity determining means calculates a luminance variance for each of the pixels included in the template image and the corresponding pixels in the pixel block, and the difference between the two luminance variances is calculated as the similarity evaluation value. A target object detection system characterized by being used as an object. In any one of Claims 5-9,
The angle-dependent similarity determining means calculates a luminance histogram for each of the pixels included in the template image and the corresponding pixels in the pixel block, and calculates the Euclidean distance between the two luminance histograms as the similarity evaluation value. A target object detection system characterized by being used as an object. In any one of Claims 5-10,
The angle-dependent similarity determination unit calculates a luminance variance for each of the pixels included in the template image and the corresponding pixels in the pixel block, and the difference between the two luminance variances is used as the similarity evaluation value. A target detection system characterized by being used. In any one of Claims 5-11,
The angle-dependent similarity determination means uses a difference value between a gray gradation value of a pixel included in the template image and a corresponding pixel in the pixel block as the similarity evaluation value. Target detection system. In any one of Claims 5-12,
The angle-dependent similarity determination means uses a normalized correlation coefficient based on a gradation value of a pixel included in the template image and a corresponding pixel in the pixel block as the similarity evaluation value. A target detection system. In any one of Claims 1-13,
The search result output means distinguishes and displays the search target area and the search non-target area on the search target image, and the rotation angle of the template image for the target detected by the similarity determination means. A target object detection system, wherein a polygonal shape defined inside a template image corresponding to is overlapped on the search target image and output.   The program for functioning a computer as a target object detection system in any one of Claims 1-14.