JP2007121111A - Target identifying technique using synthetic aperture radar image and device therof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enhance identification accuracy by determining target domain itself with sufficient accuracy, while removing influences due to foreshortening effect in two-dimensional SAR images to enhance identification accuracy of targeting objects. <P>SOLUTION: The target identifying device is equipped with a SAR (synthetic aperture radar) section 161, a SAR image acquiring section 162 inputting radar reflection signals from the SAR section 161 to form the two-dimensional SAR images, an object extracting section 163 extracting the object from the SAR images, a feature calculating section 164 calculating the feature quantity of the object extracted, a model memory section 167 memorizing models, and an object discriminating section 165 discriminating whether features of the object coincide with features of models. Next, taking into account the deformation (foreshortening) appearing on the two-dimensional images of the object depending on flight direction of the flying object and attitude of the object, the discriminating section 165 projects the reflective intensity distribution to the radar irradiating direction and discriminates whether features of the object coincide with features of the model. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention uses a synthetic aperture radar (SAR) mounted on a flying object, and the characteristics of a two-dimensional radar image of a target that is deformed depending on the flying direction of the flying object and the attitude of the target are stored. The present invention relates to a target identifying method and apparatus for identifying whether or not a feature of a model matches. In particular, the present invention relates to a target identification method and apparatus that takes into account “falling” (four shortening) in which a tall structure falls toward a flying object in a two-dimensional SAR image.

As a technique for identifying a target based on an image acquired by a synthetic aperture radar (SAR) in remote sensing from a satellite, the technique disclosed in Patent Document 1 uses a stereo SAR (for the same target). By acquiring a plurality of images at the same time, a baseline is obtained from the phase difference of the plurality of images and a moving object is extracted with high accuracy.In the technique disclosed in Patent Document 2, divided SAR (a little time is required for imaging). A plurality of shifted images are generated, and a stable structure is extracted based on the relationship, and in the technique disclosed in Patent Document 3, a reverse synthetic aperture (a method of acquiring a focused image) (Patent Document 4). The target type is discriminated by matching the target two-dimensional pattern with respect to the image extracted in (Ref.).
Japanese Patent Laid-Open No. 10-232282 (paragraph 0006) JP 2002-214334 A (paragraph 007) JP 11-109013 A (paragraph 0016) JP 2001-221857 A (paragraph 0039)

  However, in these conventional techniques, first, for example, when identifying an object composed of a collection of structures having a height, such as a ship, a SAR-specific deformation (falling down) in the positional relationship and height relationship between the structures. (For shortening)) cannot be removed, and there is a problem that the identification accuracy is lowered.

  Therefore, the first problem (falling down) will be further described with reference to FIG. In FIG. 1A, a flying object such as an artificial satellite irradiates a ship with radar radio waves. If the radar radio wave is reflected from a high part of the ship, it returns to the flying object in a shorter time. That is, the higher the object, the closer it appears to the flying object. This is a fall, and the top of the mast 1 of the ship S1 falls to a point A close to the sea surface satellite. Further, the bridge-like point P located at a position lower than the mast falls to a point B farther from the flying object than the point A. FIG. 1B shows the collapse in the two-dimensional SAR image. When radar irradiation is performed from the ship's stomach direction, the top of the mast 1 of the ship S1 is in the ship's stomach direction (the flying object direction opposite to the radar irradiation direction). Fall down. FIG. 1 (C) shows the fall of the ship S1 that progresses obliquely in the radar irradiation direction, but the fall direction is different from that of FIG. 1 (B) in relation to other components. .

  Second, it is premised that the target area itself is accurately determined. However, the target area is not always accurately determined based on the characteristics of the SAR signal (the target area may have a portion having low reflection intensity). There was a problem that it could not be decided.

  Therefore, the second problem (determination of the target area) will be further described with reference to FIG. 2, FIG. 3, and FIG. First, FIG. 2A shows a ship model, and this ship is composed of a fuselage S, a bridge B, and masts 21 and 22. FIG. 2B conceptually shows the radar reflection from the ship. The reflection radar is shown as a spot, and has strong intensities such as reflection Q from the edge portion of the fuselage and reflection P from the edge portion of the bridge. The reflection from the masts 21 and 22 is also strong, but in the two-dimensional SAR image, it appears as a reflection having a larger area than the edge portion. FIG. 3 is a flowchart of conventional target object recognition. First, radar irradiation is performed on the ship in S30, and a two-dimensional SAR image is acquired in S31. This two-dimensional SAR image includes not only the target ship but also waves and other objects. Therefore, in S32, a high luminance point group is picked up, and in S33, the target object is cut out. However, reflection from the edge of the body, for example, the points Q1 and Q2 and the edge Q3 of the bridge may not be strong, and the target object may not be cut out correctly. Further, in S34, when the histogram of the reflection intensity signal is calculated with the long side of the cut rectangle (in the hull axis direction) as the horizontal axis without considering the radar irradiation direction (hereinafter, such calculation processing is referred to as “projection”). It may not accurately represent the target profile. 4A and 4B show that the signal intensity profile changes depending on the direction of the hull and the radar irradiation direction. 4 (C) and 4 (D) show a case where the quay Y is close to the target ship X, and the rectangle that cuts out the high-intensity point bites into a part of the quay, so that the cutting is not performed accurately. Indicates.

  Therefore, a first problem of the present invention is to remove the influence due to the fall and improve the identification accuracy of the target object. The second problem of the present invention is to accurately determine the target area itself and improve the identification accuracy of the target object.

  A first means for solving the above-described problem is to use a synthetic aperture radar mounted on a flying object, and to deform the target two-dimensionally depending on the flying direction of the flying object and the posture of the target object. A target identification method for identifying whether or not a feature of a radar image matches a feature of a stored model, the step of cutting out the target from the two-dimensional radar image, and detecting a radar radio wave as the target Projecting the reflection intensity distribution of the radar radio wave to create a histogram in the direction of irradiating the light, and identifying whether or not the projected reflection intensity distribution matches the characteristics of the model. It is.

  The second means determines, in the first means, two parallel lines surrounding a high-intensity reflection point group of the radar radio wave in a direction in which the radar radio wave is irradiated toward the target; And determining the parallelogram that maximizes the reflection point group density and identifying whether the feature of the target matches the feature of the model.

  The third means includes a first step of determining an initial cutout area for cutting out the target from the two-dimensional radar image, and a radar radio wave having an intensity equal to or greater than a predetermined threshold existing in the initial cutout area. A second step of selecting a reflection point group, and a third step of determining a new cut-out area for removing a reflection point closest to the boundary of the initial cut-out area where the distance between the reflection points is maximum in the reflection point group And repeating the third step until the dispersion of the reflection point group converges, and further comprising identifying whether the feature of the target matches the feature of the model. is there.

  The fourth means further includes, in the first means, generating a two-dimensional radar image of the model by simulation and identifying whether or not the feature of the target matches the feature of the simulation image. That is.

  The fifth means further includes a step of performing image analysis on the optical image obtained by the optical camera to create a top view of the target, and calculating a height of a component in the top view in the first means. Identifying whether the features of the target match the features of the model.

  The sixth means includes a SAR unit mounted on the flying object, a SAR image acquisition unit that inputs a radar reflection signal from the SAR unit to generate a two-dimensional SAR image, and a target segmentation that extracts a target from the SAR image. A feature calculation unit that calculates a feature amount of the cut target, a model storage unit that stores a model, and an object identification unit that identifies whether the feature of the target matches the feature of the model The target identification unit includes the target based on a deformation that appears on a two-dimensional radar image of the target depending on a flight direction of the flying object and a posture of the target. Is to identify whether or not the feature of the model matches the feature of the model.

  A seventh means is the sixth means, wherein the feature calculation unit creates a histogram in which the reflection intensity distribution of the radar radio wave is projected in a direction in which the radar radio wave is irradiated toward the target. And identifying whether the reflected intensity distribution after the projection matches the characteristics of the model.

  According to an eighth means, in the sixth means, the cutout unit determines two parallel lines surrounding a high-intensity reflection point group by the radar radio wave in a direction in which the radar radio wave is irradiated toward the target, A parallelogram including the parallel lines and maximizing the reflection point group density is determined, and the object identification unit determines whether the feature of the target cut out by the parallelogram matches the feature of the model. Is to identify.

  A ninth means is the sixth means, wherein the cut-out unit determines an initial cut-out area for cutting out the target from the two-dimensional radar image, and further, the intensity of the intensity equal to or higher than a predetermined threshold existing in the initial cut-out area. Select a reflected wave from the radar radio wave, further determine a new cutout area that removes the reflected wave closest to the boundary of the initial cutout area that has the maximum distance between the reflection points in the reflection point group, and The new cutout area is updated until the dispersion of the reflection point group converges, and the target identification unit excludes the reflected wave from other than the target, and the feature of the target cut out is the feature of the model. It is to identify whether or not they match.

  The tenth means further comprises a SAR simulation unit for generating a two-dimensional radar image of the model by simulation in the sixth means, wherein the target identification unit is a two-dimensional radar image of the target inputted from the target cutout unit Is to identify whether or not it matches the simulation image.

  The eleventh means further includes, in the sixth means, an optical camera unit mounted on the flying object, and an optical image analysis unit that analyzes an optical image obtained by the optical camera unit and creates a top view of the target. The object recognition unit distinguishes the constituent elements in the top view by reference numerals and inputs the discrimination results to the feature calculation section. The feature calculation section is configured to input the constituent elements based on the input from the target recognition section. Is calculated and the calculation result is input to the target recognition unit. The target recognition unit determines whether the feature of the target matches the feature of the model based on the input from the feature calculation unit. To identify.

  According to the present invention, it is possible to remove the influence of falling and improve the identification accuracy of the target object. In addition, the target area itself can be determined with high accuracy, and the identification accuracy of the target object can be improved.

  Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the drawings. However, even if there are specific descriptions in the dimensions, materials, shapes, relative arrangements, and the like of the components described in the present embodiment, the present invention is not intended to be limited thereto. Moreover, although this embodiment is described about a two-dimensional SAR image, it is widely applicable also to a radar image.

[Embodiment 1]
Embodiment 1 relates to a target identification method, wherein a step of extracting a target from a two-dimensional radar image, and a reflection intensity distribution of radar radio waves are projected in a direction in which the radar radio waves are emitted toward the target, and a histogram is obtained. Creating and identifying whether the projected reflection intensity distribution matches the features of the model.

  The purpose of the first embodiment is based on the premise that the target is correctly cut out, and the deformation due to the height of the target structure depends on the moving direction of the satellite (azimuth direction) and the radio wave irradiation direction (range direction). It is to generate a pattern projected in the direction (range direction) to deform the reflection intensity pattern of the target candidate region of the acquired image using information.

  The principle of the first embodiment will be described with reference to FIG. In FIG. 5A, the radar irradiation is performed obliquely with respect to the traveling direction of the ship including the masts M1, M2, and M3, and the masts M1, M2, and M3 fall down opposite to the radar irradiation direction. Therefore, the signal intensity distribution is calculated by projecting in the radar irradiation direction. FIG. 5B shows a case where the radar irradiation direction is substantially perpendicular to the ship traveling direction. In this case as well, the intensity distribution projected in the radar irradiation direction is calculated. The intensity distribution calculated in this way is an intensity distribution projected in the falling direction, and the interval between the spot positions representing the interval between the masts does not change between FIG. 5 (A) and FIG. 5 (B). Note that the signal intensity at the mast position is not necessarily the same in FIGS. 5A and 5B because it is affected by the reflected waves from the periphery of those masts.

FIG. 6 is a flowchart of the target identifying method of the first embodiment. First, in S50,
A two-dimensional SAR image is acquired. Next, in S52, a target object such as a ship is cut out. Next, in S54, a point group in the clipped image is projected in the radar irradiation direction to create an intensity distribution histogram. Next, the length of the object is normalized in S56. Finally, in S58, the object is identified by comparison with the registered data.

  Therefore, the effect of the first embodiment is that the SAR-specific reflection intensity pattern that is deformed affected by the satellite motion pattern and the target attitude can be converted into a feature that is not affected by such information (invariable). That is.

[Embodiment 2]
In the second embodiment, the step of determining two parallel lines surrounding the high-intensity reflection point group of the radar radio wave in the direction of irradiating the radar radio wave toward the target in the first embodiment, and the reflection including the parallel line is reflected. A target identification method further comprising: determining a parallelogram that maximizes the point cloud density; and identifying whether the characteristics of the target match the characteristics of the model.

  The purpose of the second embodiment is that when a target is cut out with a rectangle circumscribing the target, there is a problem that a part of the fall component protrudes from the region and valuable information is lost. It is to cut out with a “parallelogram” in consideration of (using skew affine transformation instead of region identification by affine transformation).

  The principle of the second embodiment will be described with reference to FIG. In FIG. 7A, six reflective spots are indicated by practical circles, and two dotted circles indicate the position of the mast when it is assumed that there is no collapse. Four small circles represent the edge of the bridge. Therefore, first, two great circles representing a mast are surrounded by two parallel lines L1 and L2 parallel to the laser irradiation direction. Next, referring to FIG. 7B, two parallel lines L3 and L4 that include all the spots are initialized. Then, the parallel lines L3 and L4 are rotated with respect to the parallel line L1, and the parallel lines L5 and L6 in the direction in which the point group density is maximized are determined. In this way, the target cutting parallelogram L1L2L5L6 is obtained.

  FIG. 8 is a flowchart of the target identification method according to the second embodiment. First, in S70, a two-dimensional SAR image is acquired. Next, in S72, the high-intensity reflection point group is surrounded by two straight lines L1 and L2 parallel to the radar reflection direction. Finally, in S74, a parallelogram (including straight lines L1 and L2) that maximizes the point group density is determined by skew affine transformation.

  Therefore, the operational effect of the second embodiment is that, at the time of target cutout, it is possible to cut out feature points that have moved due to falling down without being lost.

[Embodiment 3]
In the first embodiment, a first step of determining an initial cutout area for cutting out a target from a two-dimensional radar image, and a reflection point group of radar radio waves having an intensity equal to or higher than a predetermined threshold existing in the initial cutout area in the first embodiment. A second step of selecting, and a third step of determining a new cutout area for removing a reflection point that is the distance between the reflection points in the reflection point group and that is closest to the boundary of the initial cutout area, Repeating the third step until the variance of the group converges, and identifying whether or not the feature of the target matches the feature of the model.

  The purpose of the third embodiment is that there is a possibility that a strong reflection component does not necessarily exist in the boundary portion for the total length of the target, and a part of the target is cut out, or other than the target (in the case of a ship, There is a problem that a part of the reflection component is included, and therefore, the periphery of the strong reflection component is cut out based on a plurality of possibilities, and each of them is compared with the database for identification.

  The principle of the third embodiment will be described with reference to FIG. FIG. 9 shows a method of cutting out the target X from the two-dimensional SAR image including the target X close to the quay Y. First, a reflection point that is equal to or higher than a threshold, such as a point group of the target object X, a wave, or a structure on a quay, is selected and a first cutout is performed. Next, the “wave” that is the farthest reflection point from the point group (particularly, two great circles representing the mast) of the target object X is removed, and a second cutout is performed. Next, a “structure on the quay” that is the farthest reflection point from the point group (particularly, two great circles representing the mast) of the target object X is removed, and a third cutout is performed. Thereafter, the clipping is repeated until the point cloud density converges.

  FIG. 10 is a flowchart of the target identifying method of the third embodiment. First, in S90, the initial region of the cutout region is determined. Next, in S92, the initial region is deformed so as to surround all reflection points having an intensity equal to or higher than the threshold value. At this time, the radar irradiation direction is taken into consideration so as not to miss the reflection point that has moved due to falling down. Next, in S94, the cutout region is deformed so as to remove the reflection point farthest from the target. Next, in S96, it is determined whether or not the dispersion between the reflection point groups has converged. In S96, when the distribution of the point cloud converges (when the point cloud density in the cut-out area converges), the cut-out ends. In S96, if the variance has not yet converged, the process returns to S94 and the clipping is continued.

  Therefore, the operational effect of the third embodiment is that although the identification accuracy may be low depending on the target position extraction accuracy, the identification accuracy can be improved even for objects and image conditions where it is difficult to uniquely determine the extraction. .

[Embodiment 4]
The fourth embodiment further includes a step of generating a two-dimensional radar image of the model by simulation in the first embodiment and a step of identifying whether or not the feature of the target matches the feature of the simulation image. Is the method.

  The purpose of the fourth embodiment is to identify the target by pattern matching without calculating the characteristics of the target (such as projection of the reflection intensity distribution in the radar irradiation direction).

  The principle of the fourth embodiment will be described with reference to FIG. One or more models (model A, model B, model C) are prepared, and a reflection pattern for each radar irradiation direction (for example, every 10 degrees) is generated by calculation and stored in a database. On the other hand, a target is cut out from the two-dimensional SAR image acquired by actually irradiating the radar (according to the second and third embodiments), and the cut-out image and the database image are subjected to pattern matching.

  FIG. 12 is a flowchart of the target identification method according to the fourth embodiment. First, in S110, the radar reflection pattern of the model is synthesized by calculation for each radar irradiation direction. Next, in S112, the longitudinal direction of the actually obtained two-dimensional SAR image is determined, and the radar irradiation direction is estimated from the angle of collapse with respect to the hull. Next, in S114, the sizes of the cut-out image and the synthesized image are aligned, and the amount of deviation of the position of the corresponding reflection point is evaluated. Finally, in S116, if there is a model whose amount of deviation is, for example, the minimum and not more than the threshold value, the target is identified as matching the model.

  Therefore, the operational effect of the fourth embodiment is that feature quantity calculation is not required and the target can be identified at high speed.

[Embodiment 5]
A fifth embodiment is the same as the first embodiment in that a step of creating a top view of a target by image analysis of an optical image obtained by an optical camera, a step of distinguishing components in the top view by reference numerals, and heights of the components A target identification method for identifying whether the feature of the target matches the feature of the model.

  The purpose of Embodiment 5 is an optical image taken at the same time as the SAR image. Since the optical image alone is an image from directly above, identification of an object whose height (not position) is a decisive factor for identification is performed. When necessary, the position of the structure is obtained from the optical image, and the height of the structure is obtained by obtaining the amount of collapse from the SAR image, thereby improving the identification accuracy.

  The principle of the fifth embodiment will be described with reference to FIG. It is assumed that the target model A and the target model B differ only in the height of the mast and have the same geometrical arrangement in the optical image. In this case, since the amount of collapse in the two-dimensional SAR image depends on the height of the mast or the like, if the position of the mast or the like is accurately determined from the optical image and identified by a code, the two-dimensional SAR is identified. It is possible to accurately determine the extent to which the component assigned the identification number in the image has fallen. Therefore, height information can be calculated from the amount of deviation from the center of the point cloud of each reflection point after falling down. The calculation method will be described later with reference to FIGS.

  FIG. 14 is a flowchart of the target identifying method of the fifth embodiment. On the SAR image analysis side, first, in S130, a two-dimensional SAR image is acquired. Next, in S132, the target is cut out. Finally, in S134, the structure height is calculated. On the other hand, on the optical image analysis side, first, in S131, an optical image is acquired. Next, in S133, the arrangement of the target corner structure is calculated, and a top view of the target is created. Next, based on this top view, for example, the area ratio of a square corresponding to a bridge and the like, the area ratio of a circle corresponding to a mast and the like are compared with a model, and an identification model is obtained by, for example, And model B. Finally, in S135, it can be determined whether the target is the target model A or the target model B by taking in the structure height calculated in S134.

  FIG. 15 is a flowchart of the structure height calculation. First, in S140, a target (reflection point group) is cut out. Next, in S142, the main axis (the central axis in the longitudinal direction of the object) is calculated (see FIG. 13). Next, in S144, the distance between each reflection point and the principal axis is calculated. This distance is calculated in the radar irradiation direction. Then, the height is calculated based on this distance. Finally, in S146, the match between the target and the model is determined using the height of each reflection point as a feature amount.

  Therefore, the operational effect of the fifth embodiment is that the identification accuracy can be improved by combining the SAR image and the optical image.

[Embodiment 6]
Embodiment 6 is a target identification device, which includes a SAR unit mounted on a flying object, a SAR image acquisition unit that inputs a radar reflection signal from the SAR unit and generates a two-dimensional SAR image, and a target from the SAR images. A target cutout unit that cuts out an object, a feature calculation unit that calculates a feature amount of the cut out target, a model storage unit that stores a model, and whether or not the feature of the target matches the feature of the model In the target object identification device including the target identification unit, the target identification unit is configured to detect the target based on the deformation that appears on the two-dimensional radar image of the target depending on the flying direction of the flying object and the attitude of the target. Identify whether the feature matches that of the model.

  The apparatus may further include an optical camera unit. That is, the apparatus further includes an optical camera unit mounted on the flying object, and an optical image analysis unit that performs image analysis of an optical image obtained by the optical camera unit and creates a top view of the target, Based on the figure, the components of the target are identified and the identification result is input to the feature calculation unit.The feature calculation unit depends on the flying direction of the flying object and the posture of the target based on the input from the object recognition unit. The height of the component is calculated based on the deformation of the target generated in the two-dimensional radar image, and the calculation result is input to the target recognition unit. The target recognition unit is based on the input from the feature calculation unit. It is identified whether the feature of the target specified by the SAR image and the optical image matches the feature of the model.

  FIG. 16 is a block diagram of the target object identification device.

  The radar reflection signal output from the SAR unit 161 (synthetic aperture radar unit) is input to the SAR image acquisition unit 162 to form a two-dimensional SAR image. The two-dimensional SAR image is input to the target cutout unit 163, and the target target is cut out. The cut target image is input to the feature calculation unit 164, and target features (such as a histogram of the reflection intensity distribution and the height of the structure in the target) are calculated. The feature amount calculated by the feature calculation unit 164 is input to the object identification unit 164. On the other hand, the target model is stored as a database in the model storage unit 167, and the model is input to the target identification unit 165 in response to a read request from the target identification unit 165. Therefore, the object identification unit 165 collates the cut-out image with the model and obtains an identification result. The identification result is output to an identification result output unit 170 (printer, display, etc.), and the presence / absence of matching with the model and the degree of matching (correlation coefficient) are output. When a match with the model is not recognized, another model may be read out by the object identification unit 165 and collation may be repeated.

  The target cutout unit 163 cuts out the target target by the method described in the second and third embodiments.

  When executing the method described in the fourth embodiment, the target cutout unit 163 inputs the cutout image to the target identification unit 165. On the other hand, the SAR simulation unit 166 inputs a SAR simulation image of the model to the target identification unit 165 in response to a request from the target identification unit 165. The model is read from the model storage unit 167 by the SAR simulation unit 166.

  The feature calculation unit 164 calculates a target feature amount (such as a reflection intensity histogram projected in the radar direction) by the method described in the first embodiment.

  When the method described in the fifth embodiment is executed, the optical image from the optical camera unit 168 is analyzed by the optical image analysis unit 169 to calculate the geometric arrangement of the upper surface of the target, and a top view is created. The creation result is input to the object identification unit 165. The object identification unit 165 narrows down model candidates based on, for example, the area occupancy ratio of a square (corresponding to a deck) or a circle (corresponding to a mast) in the geometric arrangement. In that case, the object identification part 165 attaches | subjects a code | symbol (number, a name, etc.) to the structure in a top view, and distinguishes it. The discrimination result is input to the feature calculation unit 164. On the other hand, since the cut out two-dimensional SAR image is input from the target cutout unit 163 to the feature calculation unit, the height of the encoded structure is calculated by the method of FIGS. 13 and 15. This calculation result is input to the object identification unit 165. Thereby, the height information and plane information of the contrast are specified and compared with the model.

  The apparatus described above may be a single computer system or microcomputer except for the SAR unit 161, the optical camera unit 168, the model storage unit 167 (nonvolatile memory), and the identification result output unit 170. In that case, each operation described above is executed by a computer program (not shown).

  The present invention can be used for target recognition by a two-dimensional SAR image, and the target is not limited to a ship but can be used for a moving or stopped vehicle.

It is a figure explaining the distortion (falling down (four shortening)) of the image in a two-dimensional SAR image. It is a figure explaining the two-dimensional SAR image containing a fall. 3 is a flowchart of a general method for analyzing a two-dimensional SAR image. It is a figure explaining the influence of falling down and the difficulty of a target cutout. It is a figure explaining the principle of the method of Embodiment 1. FIG. 2 is a flowchart of a method according to the first embodiment. It is a figure explaining the principle of the method of Embodiment 2. FIG. 6 is a flowchart of a method according to the second embodiment. It is a figure explaining the principle of the method of Embodiment 3. FIG. 10 is a flowchart of a method according to the third embodiment. It is a figure explaining the principle of the method of Embodiment 4. FIG. 10 is a flowchart of a method according to the fourth embodiment. It is a figure explaining the principle of the method of Embodiment 5. FIG. 10 is an overall flowchart of a method according to the fifth embodiment. 10 is a flowchart of a part of the method of the fifth embodiment. It is a block diagram of the apparatus of Embodiment 6.

Explanation of symbols

161 SAR unit 162 SAR image acquisition unit 163 target segmentation unit 164 feature calculation unit 165 target identification unit 166 SAR simulation unit 167 model storage unit 168 optical camera unit 169 optical image analysis unit 170 recognition result output unit

Claims (11)

Using a synthetic aperture radar mounted on the flying object, the characteristics of the two-dimensional radar image of the target deformed depending on the flight direction of the flying object and the attitude of the target match the characteristics of the stored model A target identification method for identifying whether or not to
Cutting out the target from the two-dimensional radar image;
Projecting a reflection intensity distribution of the radar radio wave in a direction of irradiating the radar radio wave toward the target, and creating a histogram;
And a step of identifying whether or not the projected reflection intensity distribution matches the feature of the model. Determining two parallel lines surrounding a high intensity reflection point group of the radar radio wave in a direction of irradiating the radar radio wave toward the target;
Determining a parallelogram that includes the parallel lines and maximizes the reflection point cloud density;
The method for identifying a target according to claim 1, further comprising a step of identifying whether or not the feature of the target matches the feature of the model. A first step of determining an initial cutout region for cutting out the target from the two-dimensional radar image;
A second step of selecting a reflection point group of the radar radio wave having an intensity equal to or higher than a predetermined threshold existing in the initial cutout region;
A third step of determining a new cutout area for removing a reflection point that is the distance between the reflection points in the reflection point group and that is closest to the boundary of the initial cutout area,
Repeating the third step until the dispersion of the reflection point group converges,
The method of claim 1, further comprising the step of identifying whether the feature of the target matches the feature of the model. Generating a two-dimensional radar image of the model by simulation;
The target identification method according to claim 1, further comprising the step of identifying whether or not the feature of the target matches the feature of the simulation image. Further comprising: analyzing an optical image from an optical camera to create a top view of the target; and calculating a height of a component in the top view,
The target object identifying method according to claim 1, further comprising: identifying whether or not the feature of the target object matches the feature of the model. A SAR unit mounted on a flying object;
A SAR image acquisition unit that inputs a radar reflection signal from the SAR unit and generates a two-dimensional SAR image;
A target cutout unit for cutting out a target from the SAR image;
A feature calculation unit for calculating a feature amount of the cut target;
A model storage unit for storing the model;
A target identification device comprising: an object identification unit for identifying whether the characteristics of the target matches the characteristics of the model;
The target identification unit is configured such that the feature of the target matches the feature of the model based on a deformation that appears on a two-dimensional radar image of the target depending on a flight direction of the flying object and a posture of the target. A target identification device for identifying whether or not to perform. The feature calculation unit creates a histogram in which a reflection intensity distribution of the radar radio wave is projected in a direction in which the radar radio wave is irradiated toward the target,
The target identifying apparatus according to claim 6, wherein the target identifying unit identifies whether the reflected intensity distribution after the projection matches the feature of the model. The cutout unit determines two parallel lines surrounding a high-intensity reflection point group by the radar radio wave in a direction in which the radar radio wave is irradiated toward the target, and includes the parallel lines and the reflection point group density Determine the parallelogram that maximizes
The target identifying apparatus according to claim 6, wherein the target identifying unit identifies whether or not the feature of the target cut out by the parallelogram matches the feature of the model. The cutout unit determines an initial cutout region for cutting out the target from the two-dimensional radar image, further selects a reflected wave by the radar radio wave having an intensity equal to or higher than a predetermined threshold existing in the initial cutout region, and Determining a new cut-out area that removes the reflected wave closest to the boundary of the initial cut-out area and having the maximum distance between the reflection points in the reflection point group, and further, until the dispersion of the reflection point group converges Update the new cutout area,
7. The object identification unit according to claim 6, wherein the target identification unit identifies whether or not a feature of the target cut out by removing a reflected wave from other than the target matches a feature of the model. Target identification device. A SAR simulation unit that generates a two-dimensional radar image of the model by simulation;
The target identification apparatus according to claim 6, wherein the target identification unit identifies whether a two-dimensional radar image of the target input from the target cutout unit matches the simulation image. An optical camera unit mounted on the flying object;
An optical image analysis unit that analyzes an optical image by the optical camera unit and creates a top view of the target;
The object recognition unit distinguishes the constituent elements in the top view by a code and inputs the discrimination result to the feature calculation unit,
The feature calculation unit calculates the height of the component based on the input from the target recognition unit and inputs the calculation result to the target recognition unit;
The target identification apparatus according to claim 6, wherein the target recognition unit identifies whether the feature of the target matches the feature of the model based on an input from the feature calculation unit.

Images