JP4472978B2 - Target extraction apparatus, target extraction method, and computer-readable recording medium recording program - Google Patents

Description

  The present invention relates to a technique for detecting a target such as a ship from digital image data such as a satellite photograph, and more particularly to a technique capable of accurately detecting a target.

  A conventional technique for detecting a target such as a ship from digital image data such as satellite photographs will be described. FIG. 12 is a diagram showing a target detection gate of the prior art. This is a technique disclosed in Japanese Patent Laid-Open No. 2000-353234. In this example, the ratio of the average brightness of the outer frame and the inner frame is calculated, and the presence / absence of the target point is determined by comparing the ratio with a determination threshold value.

  FIG. 13 is a diagram illustrating a background area and a target candidate constituent pixel determination area in the related art. This is a technique disclosed in Japanese Patent Laid-Open No. 2000-180543. Statistical values (average value, standard deviation, etc.) are calculated from all the pixels in the background area, and a determination value is calculated from these statistical values. The presence / absence of the target point is determined by comparing the determination value with the luminance of each pixel in the target candidate constituent pixel determination region.

  In any of the above examples, paying attention to the fact that the brightness of the target is higher than the sea level, the target point is extracted (presence / absence determination) using a gate consisting of the center (inner side) and outer frame. Depending on the size and arrangement of the target (collection of target points) existing in the gate, the luminance average of the outer frame part or the statistical value of the background region may be affected, and extraction omission may occur.

  As described above, in the target point extraction using the conventional gate, the luminance difference (ratio) between the central portion and the outer frame is used. The following introduces cases that prevent accurate extraction.

  FIG. 14 is a diagram showing an image of a large target with respect to the gate. If the target (collection of target points) is larger than the gate as shown in this figure, the brightness of the entire outer frame will increase, so the difference from the center will not be clear and will not be extracted as target points There was a fear.

  Moreover, the land area brightness included in the satellite image may be converted to 0 and distinguished from the water area as the land area removal process. In such a case, if the average luminance of the gate is calculated as it is, the average decreases as a whole. FIG. 15 is a diagram illustrating an image of a gate including a land area. As described above, if the luminance average B of the gate is lowered, the luminance ratio (= A / B) with the luminance A in the central portion is increased, so that a target point that is not originally extracted may be erroneously extracted by comparison with the threshold value. There is.

Although it is possible to minimize the influence of the size of the target in the gate by using only the outer peripheral portion as the outer frame portion of the gate, in the case of the case as shown in FIG. 16 or FIG. As a result, the average luminance of the gate increases as a whole, and the target point may not be extracted. FIG. 16 is a diagram illustrating an image in which the target itself overlaps the gate. FIG. 17 is a diagram illustrating an image in which another target overlaps the gate. As described above, when the ship is larger than expected or when the ship is dense, the determination of the presence of the target by the luminance ratio (or difference) may not be performed normally.
JP 2000-353234 A JP 2000-180543 A

  The present invention has been made in order to eliminate the above-described drawbacks of the prior art, and the object of the present invention is to have a large target located in the center and overlap with the gate or close to it. It is an object of the present invention to accurately calculate a luminance average of a background region and accurately extract a target even when the target overlaps with a gate.

The target object extraction device according to the present invention includes:
(1) An image input processing unit that inputs an image composed of a pixel group having luminance. (2) A determination target pixel that is any pixel in the input image. An average luminance is obtained for each of the plurality of gates having a predetermined relative positional relationship with the determination target pixel, and a gate luminance average calculation processing unit (3) that selects the gate luminance average of the determination target pixel from the average luminance group Target point determination processing unit (4) for determining whether or not the determination target pixel is included in the target area based on the luminance of the determination target pixel and the average gate luminance of the determination target pixel. An extraction result output processing unit that outputs coordinate information of pixels determined to be included in the region.

  The determination target pixels are all pixels included in the input image.

  The target object extracting apparatus includes a pixel selection processing unit that compares the luminance of a pixel group included in an input image with a predetermined threshold value and selects a determination target pixel based on the comparison result.

  The plurality of gates are similar layered gates centered on the determination target pixel.

  The layered gates have a predetermined width, and the layered gates have a predetermined interval.

  The layered gate is square.

  The layered gate is circular.

  The gate luminance average calculation processing unit selects a minimum value from the average luminance group as a gate luminance average in the determination target pixel.

  The gate luminance average calculation processing unit selects a maximum value from the average luminance group as a gate luminance average in the determination target pixel.

  The target point determination processing unit determines the ratio of the luminance of the determination target pixel and the gate luminance average by comparing with a predetermined threshold value.

  The target point determination processing unit determines the difference between the luminance of the determination target pixel and the gate luminance average with a predetermined threshold value.

The target extraction method according to the present invention includes:
(1) A step of inputting an image composed of a pixel group having luminance. (2) For a determination target pixel that is any pixel in the input image, the determination target. Step (3) of determining the average luminance for each of a plurality of gates having a predetermined relative positional relationship with the pixel, and selecting the average gate luminance of the determination target pixel from the average luminance group; Step (4) for determining whether or not the determination target pixel is included in the target area based on the gate luminance average in the determination target pixel (4) Coordinate information of the pixel determined to be included in the target area Is output.

A computer-readable recording medium on which a program according to the present invention is recorded,
A computer-readable recording medium that records a program for causing a computer serving as a target extraction device to execute the following processing. (1) Processing for inputting an image composed of a pixel group having luminance (2) For a determination target pixel that is one of the pixels in the input image, an average luminance is obtained for each of the plurality of gates that have a predetermined relative positional relationship with the determination target pixel, and the average luminance group The process of selecting the gate luminance average in the determination target pixel (3) Whether the determination target pixel is included in the region of the target based on the luminance of the determination target pixel and the gate luminance average in the determination target pixel Processing for determining whether or not (4) Processing for outputting coordinate information of pixels determined to be included in the target region.

The program according to the present invention is:
It is a program for causing a computer as a target object extracting apparatus to execute the following procedure. (1) Procedure for inputting an image composed of a pixel group having luminance (2) Any of the input images Average luminance is obtained for each of the plurality of gates having a predetermined relative positional relationship with the determination target pixel, and the average luminance of the gates in the determination target pixel is determined from the average luminance group. (3) Procedure for determining whether or not the determination target pixel is included in the target area based on the luminance of the determination target pixel and the gate luminance average of the determination target pixel (4) A procedure for outputting coordinate information of pixels determined to be included in the target region.

  In the present invention, the average luminance is obtained for each of the plurality of gates having a predetermined relative positional relationship with the target pixel, and the average of the gate luminance at the target pixel is selected from the average luminance group. Even if an unexpected object (including the target itself) overlaps the gate, the luminance average of the gate is excluded, the appropriate luminance average of the background area is obtained, and based on this luminance average Thus, it is possible to accurately extract points (target points) within the target area.

Embodiment 1 FIG.
Hereinafter, the present invention will be described based on embodiments shown in the drawings. FIG. 1 is a diagram showing an image of a gate according to the present invention. In the present invention, it is assumed that the influence of the size of the target is minimized by setting the outer frame portion of the gate only to the outer periphery.

  Then, a plurality of such gates are set to calculate the respective luminance averages, and the smallest luminance average among them is set as the luminance average of the gates. This minimizes the increase in luminance average due to unexpected circumstances.

  FIG. 2 is a diagram illustrating an image in which the target itself overlaps a plurality of gates. In this example, the minimum luminance average can be obtained with the gate 3 where the targets themselves do not overlap. This luminance average indicates the luminance of the water area. Thus, the gate is a group of similar layered gates centered on the determination target pixel. In this example, a square (rectangular example) gate is used in particular. The layered gates have a predetermined width, and the layered gates have a predetermined interval.

  FIG. 3 is a diagram illustrating an image in which another target overlaps a plurality of gates. In this example, the minimum luminance average can be obtained at the gate 2 where other objects do not overlap.

  In order to perform processing in this way, the target object extraction device performs processing as follows. FIG. 4 is a diagram showing an overall processing flow.

  In S10, the target image reading processing unit performs the target image reading process. In this process, a target image from which a target point is extracted is read from a CD-ROM or a hard disk. The target image is a digital image such as a satellite image, and is an example of an image including a pixel group having luminance. This image may be an original image or a preprocessed image such as land removal processing or luminance conversion.

  In S20, the target point extraction processing unit performs target point extraction processing. In this processing, a plurality of gates having different sizes are scanned on the image, the average luminance of each gate at each scanning position is calculated, and the ratio (or difference) between the minimum value and the central luminance of the gate is calculated. By comparing with the determination threshold value, the central part of the gate is extracted as a target point. Details will be described later with reference to FIG.

  In S30, the extraction result writing processing unit performs extraction result writing processing. In this process, the position (coordinates) of the extracted target point on the image and the total number of extractions are written to the hard disk or temporary area.

  The target point extraction process in S20 will be described in detail. FIG. 5 is a diagram showing a target point extraction processing flow.

  In the loop process of S201, in order to extract the target point on the image, the luminance is sequentially acquired for the positions of all the pixels, and the following process is repeated. Therefore, the loop counter corresponds to the pixel number. The pixel to be processed (determination target pixel) is assumed to be the center of the gate in the following processing, and the luminance acquired here is assumed to be the luminance of the center of the gate.

  In S202, the pixel selection processing unit performs a cut-off determination process. In this process, the center luminance of the gate is compared with a preset cut-off luminance, and if it is smaller than the cut-off luminance, control is performed so that the subsequent extraction processing is not performed. Since it is assumed that the video portion of the target has a luminance that is at least larger than the cut-off luminance, this processing is effective. In other words, the minimum brightness that cannot be the target point is set as a cut-off value in advance, and pixels with brightness less than or equal to the cut-off value are excluded in the previous stage of the extraction process, thereby improving the efficiency of the extraction process. Yes.

In S203, the gate luminance average calculation processing is performed by the gate luminance average calculation processing unit. In this process, the luminance average is calculated for each gate, and the smallest one of the calculated luminance averages is set as the gate luminance average at the scanning position.
Details of this processing will be described later with reference to FIG.

  In S204, the luminance ratio calculation processing unit performs luminance ratio calculation processing. In this process, the ratio (= A / B) between the central luminance A of the gate and the average luminance B of the gate is calculated.

  In S205, the luminance ratio determination processing unit performs luminance ratio determination processing. In this process, the threshold value set in advance is compared with the luminance ratio, and if the luminance ratio> threshold value, the central pixel of the gate is determined as the target point. That is, it is determined that the determination target pixel is included in the target area.

  In S206, the extraction result storage processing unit performs extraction result storage processing. The brightness and coordinates of the target point on the image are stored in the target point information array.

  Here, the gate luminance average calculation processing in S203 will be described in detail. FIG. 6 is a diagram showing a gate luminance average calculation processing flow.

  In the loop 1 process of S2031, the luminance average of the gates is calculated in order for the number of gates set in advance. Therefore, the loop counter corresponds to the set number of gates.

  In S2032, a gate size reading process is performed. Read the corresponding gate size from the settings table.

  In the loop 2 process of S2033, the luminance total value of the pixels constituting the gate is calculated. Therefore, the loop counter corresponds to the number of pixels constituting the gate.

  In S2034, a gate configuration pixel position calculation process is performed. The position on the image of the pixels constituting the gate is calculated from the center position of the gate and the gate size.

  In S2035, gate configuration pixel position determination processing is performed. It is determined whether or not the position of the pixels constituting the gate is within the image. If it is outside the image, it is removed from the luminance average calculation target. FIG. 7 is a diagram illustrating the concept of the gate configuration pixel position determination process. As shown in this figure, the part outside the image is excluded from the calculation.

  In S2036, a gate configuration pixel luminance determination process is performed. It is determined whether or not the luminance of the pixels constituting the gate is 0. If the luminance is 0, the pixel is excluded from the luminance average calculation target. FIG. 8 is a diagram illustrating the concept of the gate configuration pixel luminance determination process. This process is based on a preprocessed image that identifies a land area with 0 brightness.

  In S2037, a gate constituent pixel luminance total calculation process is performed. In this process, the total luminance value of the pixels constituting the gate is calculated, and the number of luminance average target pixels is incremented.

Total luminance = total luminance + the luminance of the Jth pixel Luminance average target pixel number = luminance average target pixel number + 1
In S2038, luminance average calculation processing is performed. In this process, the average luminance of the gate is calculated. Divide the luminance total by the number of luminance average target pixels to obtain the quotient as the gate luminance average.
Average gate luminance = total luminance / number of luminance average target pixels In S2039, luminance average minimum value calculation processing is performed. The minimum luminance average is selected from the average luminance of the gate calculated for each gate size, and is set as the average luminance of the gate at the scanning position. FIG. 9 is a diagram showing the concept of the luminance average minimum value calculation process. In the example of this figure, since the luminance average of the gate 2 is the minimum, the luminance average is set as the luminance average of the gate at the scanning position.

Embodiment 2. FIG.
In the above-described embodiment, since it is assumed that the luminance of the target area is larger than the luminance of the background area, the minimum luminance average is selected in S2039. However, if it is assumed that the brightness of the target area is smaller than the brightness of the background area, it is effective to select the maximum brightness average in S2039.

  In S205, it is determined whether the luminance ratio is equal to or less than a threshold value.

Embodiment 3 FIG.
In the above embodiment, the luminance ratio is used, but a luminance difference may be used.

  Therefore, in S204, the difference (= A−B) between the central luminance A of the gate and the average luminance B of the gate is calculated.

  In step S205, it is determined whether the luminance difference is greater than or equal to a threshold value.

Embodiment 4 FIG.
When the luminance of the target area is assumed to be smaller than the luminance of the background area, and a luminance difference is used, in S204, the difference between the average luminance B of the gate and the central luminance A of the gate (= B−A). In step S205, it is determined whether the luminance difference is equal to or greater than a threshold value.

Embodiment 5 FIG.
In the above example, concentric square gate groups are used, but the gate shape may be other shapes. For example, a concentric circular gate group may be used.

  The target extraction apparatus is a computer, and each element can execute processing by a program. Further, the program can be stored in a storage medium so that the computer can read the program from the storage medium.

  FIG. 10 is a diagram illustrating a hardware configuration example of the target object extraction device. A display device 1001, an arithmetic device 1002, a data storage device 1003, an image output device 1004, an operation input device 1005, a memory 1006, and an image input device 1007 are connected to the bus. The data storage device 1003 is, for example, a hard disk, the image output device 1004 is, for example, a printer, and the image input device 1007 is, for example, a scanner.

  The program is normally stored in the data storage device 1003 and is loaded into the arithmetic device 1002 in order to be processed while being loaded in the memory 1006.

  FIG. 11 is a diagram showing the main elements and data flow of the target extracting apparatus. The target image reading processing unit 1101 is an example of an image input processing unit, and is configured to perform the process of S10 in FIG. The gate luminance average calculation processing unit 1102 is configured to perform the process of S203 shown in FIG. 5 among the processes of S20. The target point determination processing unit 1105 includes a luminance ratio calculation processing unit 1103 that performs the processing of S204 and a luminance ratio determination processing unit 1704 that performs the processing of S205. The extraction result writing processing unit 1106 is an example of an extraction result output processing unit, and is configured to perform the process of S206.

It is a figure which shows the image of the gate which concerns on this invention. It is a figure which shows the image in which the target itself overlaps with several gates. It is a figure which shows the image which another target object overlaps with several gates. It is a figure which shows the whole processing flow. It is a figure which shows a target point extraction process flow. It is a figure which shows a gate brightness | luminance average calculation processing flow. It is a figure which shows the concept of a gate structure pixel position determination process. It is a figure which shows the concept of a gate structure pixel brightness | luminance determination process. It is a figure which shows the concept of a brightness | luminance average minimum value calculation process. It is a figure which shows the hardware structural example of a target object extraction apparatus. It is a figure which shows the main elements and data flow of a target object extraction apparatus. It is a figure which shows the target detection gate of a prior art. It is a figure which shows the background area | region and target candidate structure pixel determination area | region of a prior art. It is a figure which shows the image of a big target with respect to a gate. It is a figure which shows the image of the gate containing a land area. It is a figure which shows the image in which the target itself overlaps with a gate. It is a figure which shows the image in which another target object overlaps with a gate.

Explanation of symbols

  1001 Display device, 1002 Arithmetic device, 1003 Data storage device, 1004 Image output device, 1005 Operation input device, 1006 Memory, 1007 Image input device, 1101 Target image reading processing unit, 1102 Gate luminance average calculation processing unit, 1103 Luminance ratio calculation Processing unit, 1104 luminance ratio determination processing unit, 1105 target point determination processing unit, 1106 extraction result writing processing unit.

Claims (14)

Target extraction device characterized by having the following elements: (1) An image input processing unit for inputting an image consisting of a pixel group having luminance (2) Extracting one pixel included in the input image as a determination target pixel A plurality of regions having similar shapes but different sizes and having the determination target pixel as a center are set as a plurality of gates, and a plurality of pixels constituting the gate among all the pixels included in the image are obtained for each gate. The average brightness of the plurality of pixels is obtained as the average brightness of the gate, the average brightness of each of the plurality of gates is compared, and the average brightness of a specific gate among the average brightness of each of the plurality of gates is determined based on the comparison result. based on the gate luminance average and in the luminance and the determination target pixel gate average luminance calculation section (3) the determination target pixel selected as gate luminance average of pixel Extraction result output process for outputting the coordinate information of the determination target pixel the determination target pixel is determined to be included in the area of the target point determination processing unit determines whether or not included in the region of the target (4) target Department. The target object extraction device according to claim 1, wherein the gate luminance average calculation processing unit sets all the pixels included in the input image as determination target pixels . The target extraction device further includes:
Comparing all pixels each luminance included in the input image and Jo Tokoro threshold, the pixel selection unit for selecting a pixel in the determination target pixel among all the pixels included in the image based on the comparison result Yes, and
The target extraction apparatus according to claim 1, wherein the gate luminance average calculation processing unit uses, as a determination target pixel, a pixel selected by the pixel selection processing unit among all the pixels included in the input image. . The target extraction apparatus according to any one of claims 1 to 3 , wherein the plurality of gates are similar layered gates having the determination target pixel as a center.   5. The target extracting apparatus according to claim 4, wherein the layered gates have a predetermined width, and the layered gates have a predetermined interval.   6. The target object extracting apparatus according to claim 5, wherein the layered gate is a square.   6. The target object extracting apparatus according to claim 5, wherein the layered gate is circular. The gate luminance average calculation processing unit, a minimum of the average luminance of the average Luminance of each of the plurality of gates, any claims 1 to 7, characterized in that selected as gate luminance average in the determination target pixel target extracting apparatus according to. The gate luminance average calculation processing unit, the maximum average luminance of the average Luminance of each of the plurality of gates, any claims 1 to 7, characterized in that selected as gate luminance average in the determination target pixel target extracting apparatus according to. The target point determination processing unit compares said determined target pixel luminance and the determination ratio to Jo Tokoro threshold gate luminance average and the target pixel, in the region of the determination target pixel is the target on the basis of the comparison result It is determined whether it is contained , The target object extraction apparatus in any one of Claims 1-9 characterized by the above-mentioned. The target point determination processing unit compares a difference between a luminance of the determination target pixel and a gate luminance average of the determination target pixel with a predetermined threshold, and the determination target pixel is included in a target region based on a comparison result. 10. The target object extracting device according to claim 1, wherein it is determined whether or not the target is extracted. A target extraction method characterized by having the following elements: (1) a step of inputting an image composed of a pixel group having luminance; (2) one pixel included in the input image is extracted as a determination target pixel ; A plurality of regions having different shapes and sizes, each having a plurality of areas centered on the pixel to be determined as a plurality of gates, and a plurality of pixels included in the image being determined for each gate. The average luminance of the gates is obtained as the average luminance of the gates, the average luminances of the plurality of gates are compared, and the average luminance of the specific gates among the average luminances of the plurality of gates is compared in the determination target pixel based on the comparison result Ryo step (3) the determination target pixel said determination object pixel is the target on the basis of the gate luminance average in the determination target pixel and the luminance of selecting a gate luminance average Determining whether or not step (4) a step of outputting the coordinate information of the determination target pixel determined to be included in the region of the target contained in the. A computer-readable recording medium that records a program for executing the following processing in a computer that is a target extraction device. (1) A process of inputting an image composed of a pixel group having luminance. (2) An input image . One included pixel is extracted as a determination target pixel, a plurality of regions having similar shapes and different sizes and centered on the determination target pixel are set as a plurality of gates, and included in the gate among all the pixels included in the image For each gate, determine the average luminance of the calculated pixels as the average luminance of the gate, compare the average luminance of each of the gates, and calculate the average luminance of each of the plurality of gates based on the comparison result. out the process of selecting the average luminance of the specific gate as a gate luminance average in the determination target pixel (3) the determination and the brightness of the determination target pixel The determination target pixel is determined whether or not included in the area of the target (4) coordinate information of the determination target pixel determined to be included in the area of the target on the basis of the gate brightness mean and the elephant pixel Processing to output. The computer as a target extractor, a determination target pixel one pixel included in the procedure (2) the input image for inputting an image consisting of pixel groups having a program (1) brightness for executing the following steps Extract, set a plurality of regions centered on the determination target pixel different in size in a similar shape, a plurality of pixels included in the gate among all the pixels included in the image, for each gate, The average luminance of the obtained plurality of pixels is obtained as the average luminance of the gate, the average luminance of each of the plurality of gates is compared, and the average luminance of the specific gate among the average luminance of each of the plurality of gates is calculated based on the comparison result. determining step (3) selecting as the gate luminance average of the target pixel before on the basis of the gate luminance average of the luminance and the determination target pixel of the determination target pixel Procedure determination target pixel and outputs the coordinate information of the determination target pixel determined to be included in the area of whether the procedure determines (4) target included in the area of the target.

Images