ISAR Image Data Learning Device, Target Classification Device, and Radar Device

Abstract

To obtain an ISAR (Inverse Synthetic Aperture Radar) image data learning device which generates a learning result so as to be able to limit a range of learning results used for classification, and to provide a target classification device using the same, and a radar device.SOLUTION: An ISAR image data learning device includes: an ISAR image input part to which the ISAR image data is inputted for each target; a difference information input part to which the difference information equal to the information due to the ISAR image data different in the same target is inputted for each of the ISAR image data; and a learning part in which learning is performed in association with the ISAR image data corresponding to the difference information inputted to the difference information input part and the target type.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to an ISAR image data learning device for learning ISAR image data, a target classification device using the same, and a radar device.

Background Art

[0002] Conventionally, an inverse synthetic aperture radar (ISAR) has been used for a process of performing target identification using an ISAR image created from radar echoes. For example, there are a method of simulating and collating a target captured by ISAR with a target shape model stored in a database in advance, and a method of extracting a plurality of feature points from both the ISAR image and the target shape model.

[0003] As a sensor device for obtaining image data for classifying a target, there are an imaging device having an imaging element such as a camera or a telescope (see, for example, Patent Document 1), and a radar device such as a synthetic aperture radar (SAR) and the inverse synthetic aperture radar described above (see, for example, Patent Document 2).

[0004] In the telescope which is an imaging device disclosed in Patent Document 1, an observation device for observing a satellite is provided with a function of identifying the satellite from the position of the satellite in the image. This telescope is characterized by being able to identify a moving object even in a time zone when the luminance of the emitted light or reflected light of the moving object is low. When identifying a target from an ISAR image, there is one that generates a pseudo-ISAR image based on a target model selected from a target model database, environmental conditions such as a sheath state obtained at the time of actual observation, and observation conditions such as radar specifications (see, for example, Patent Document 3).

[0005] Conventionally, in order to always accurately identify a target from an ISAR image, some target classification devices extract the overall length of the target and the size of the structure from the ISAR image and perform target identification based on them (see, for example, Patent Document 4). In addition, some target classification devices generate a pseudo-ISAR image from a three-dimensional model of the target to be identified, using the attitude change due to the rotational motion of the target to be identified as a parameter (see, for example, Patent Document 5).

[0006] Furthermore, in the classification of ISAR images, there are those that compensate for the influence of the discrete distribution of observation parameters, generate a highly compatible filter, and perform highly accurate type determination (see, for example, Patent Document 6). On the other hand, there are those that create a learning model by machine learning using AI (Artificial Intelligence) or the like for use in the classification of ISAR images and the like (see, for example, Patent Document 2).

[0007] Note that there has conventionally been a method of performing extraction processing after integrating the ISAR image over time (see, for example, Patent Document 7). When the ISAR image is integrated over time, a plurality of images are superimposed according to the integration time, so there is an effect of improving the apparent blurriness. In Patent Document 7, rather than improving the apparent blurriness caused by superimposing a plurality of images according to the integration time of the ISAR image, a method of identifying a target using a single ISAR image without integration is disclosed, because the overall shape becomes blurred.

[0008] Of course, in order to avoid the overall shape becoming blurred, an appropriate integration time (observation time) is set (see, for example, Patent Document 8). Usually, if observed for a long time, weak signals can be accumulated due to the integration effect. However, in the case of an ISAR image, if the observation time exceeds an appropriate integration time (observation time), the motion of each reflection point of the target becomes complex, and the change in speed within the integration time (observation time) becomes large. Therefore, the Doppler spread becomes large, and blurring occurs in the image, so it is necessary to set an appropriate observation time. The appropriate observation time is determined depending on the relative motion between the ISAR and the target.

Prior Art Documents

Patent Documents

[0009]

Patent Document 1

Patent Document 2

Patent Document 3

Patent Document 4

Patent Document 5

Patent Document 6

Patent Document 7

Patent Document 8

Summary of the Invention

Problems to be Solved by the Invention

[0010] However, conventional machine learning as disclosed in Patent Document 6 has a problem that a learning model that restricts the range of learning results used for classification has not been considered. Note that the method disclosed in Patent Document 1 does not suggest using the aspect angle for the learning model due to the limitation. Also, the method disclosed in Patent Document 3 does not suggest using the sheath state for the learning model due to the limitation. Furthermore, the methods disclosed in Patent Documents 7 and 8 do not suggest using the integration time for the learning model due to the limitation.

[0011] The present disclosure has been made to solve the above-described problems, and an object thereof is to obtain an ISAR image data learning device that generates learning results, a target classification device using the same, and a radar device so that the range of learning results used for classification can be restricted.

Means for Solving the Problems

[0012] The ISAR image data learning device according to the present disclosure is an ISAR image data learning device that generates a learning result for classifying a target from ISAR image data obtained by an inverse synthetic aperture radar installed at an observation point, and includes an ISAR image input unit to which the ISAR image data is input, a difference information input unit to which at least two of an aspect angle that is an angle of a transmission beam of the inverse synthetic aperture radar with respect to the target, information on the wave height of the sea area where the target exists, and an integration time from which the ISAR image data is derived are input as difference information, and a learning unit that learns to classify the target from the ISAR image data corresponding to the difference information input to the difference information input unit. , the integration time is such that an integration start time and an integration end time are set between the observation start time and the observation end time of the target It is such a thing.

[0013] The target classification device according to the present disclosure includes an ISAR image input unit to which ISAR image data obtained by an inverse synthetic aperture radar installed at an observation point is input, a difference information input unit to which at least two of an aspect angle that is an angle of a transmission beam of the inverse synthetic aperture radar with respect to the target, information on the wave height of the sea area where the target exists, and an integration time from which the ISAR image data is derived are input as difference information, and a learning unit that learns to classify the target from the ISAR image data corresponding to the difference information input to the difference information input unit. , the integration time is such that an integration start time and an integration end time are set between the observation start time and the observation end time of the target It includes an ISAR image data learning device, a new ISAR image input unit to which the newly obtained new ISAR image data is input, a new difference information input unit to which the difference information corresponding to the new ISAR image data is input, and a target classification unit that classifies the target from the new ISAR image data based on the learning result learned by the learning unit.

[0014] The radar device according to the present disclosure includes an ISAR image input unit into which ISAR image data obtained by an inverse synthetic aperture radar installed at an observation point is input, an aspect angle which is the angle of a transmission beam of the inverse synthetic aperture radar with respect to a target, information on the wave height of the sea area where the target exists, and at least two of the integration times from which the ISAR image data was derived are input as difference information. And a learning unit that learns to classify a target from the ISAR image data corresponding to the difference information input to the difference information input unit. , the integration time is such that an integration start time and an integration end time are set between the observation start time and the observation end time of the target An ISAR image data learning device, a new ISAR image input unit into which the newly obtained new ISAR image data is input, a new difference information input unit into which the difference information corresponding to the new ISAR image data is input, and based on the learning result learned by the learning unit. A target classification device having a target classification unit that classifies the target from the new ISAR image data, a tracking device that tracks the target, and the inverse synthetic aperture radar that transmits a transmission beam to the target tracked by the tracking device in order to acquire the new ISAR image data. It is equipped with.

Effects of the Invention

[0015] According to the present disclosure, By inputting at least two of the aspect angle, which is the angle of the transmission beam of the inverse synthetic aperture radar with respect to the target, the information on the wave height of the sea area where the target exists, and the integration time when the ISAR image data is derived, as difference information, the range of the learning results used for classification is restricted An ISAR image data learning device that can be obtained, a target classification device using the same, and a radar device can be obtained.

Brief Description of Drawings

[0016] [Figure 1] It is a functional block diagram of an ISAR image data learning device according to Embodiment 1. [Figure 2] It is a flowchart for explaining the operation (ISAR image data learning method) of the ISAR image data learning device according to Embodiment 1. [Figure 3] It is a functional block diagram of an ISAR image data learning device according to Embodiment 1 (with an ISAR image generation unit 13). [Figure 4] Functional block diagrams of the ISAR image data learning device, target classification device, and radar device according to Embodiment 1. [Figure 5] Radar device according to Embodiment 1 for acquiring ISAR image data input to the ISAR image data learning device according to Embodiment 1 (new ISAR image data input to the target classification device according to Embodiment 1), its scanning range, and an exemplary diagram showing a target. [Figure 6] Functional block diagrams of the ISAR image data learning device, target classification device, and radar device according to Embodiment 1. [Figure 7] Radar device according to Embodiment 1 for acquiring ISAR image data input to the ISAR image data learning device according to Embodiment 1 (new ISAR image data input to the target classification device according to Embodiment 1), its scanning range, and an exemplary diagram showing a target. [Figure 8] Functional block diagrams of the ISAR image data learning device, target classification device, and radar device according to Embodiment 1. [Figure 9] Functional block diagrams of the ISAR image data learning device, target classification device, and radar device according to Embodiment 1. [Figure 10] Functional block diagrams of the ISAR image data learning device, target classification device, and radar device according to Embodiment 1. [Figure 11] Flowchart for explaining the operations (target classification method) of the ISAR image data learning device and target classification device according to Embodiment 1.

Modes for Carrying Out the Invention

[0017] Embodiment 1. Hereinafter, an ISAR image data learning device according to Embodiment 1, a target classification device using the same, and a radar device (the target classification device according to Embodiment 1, the radar device according to Embodiment 1) will be described with reference to FIGS. 1 to 11. In the figures, the same reference numerals indicate the same or corresponding parts, and detailed descriptions thereof will be omitted. As described above, ISAR is an abbreviation for Inverse Synthetic Aperture Radar, which is an inverse synthetic aperture radar.

[0018] The target 1 to be observed is, for example, a moving object such as a ship or an aircraft. The fuselage of the target 1 preferably has a linear structure. That is, when the observation point relative to the target 1 changes, it is preferable that the shape of the target 1 changes. Note that, if it has a linear structure, it can be easily understood that the shape is different when observed from the side and when observed from the front or the rear. Also, the observation point itself may move, or the target 1 may move. In the present application, learning is performed on the premise that there are a plurality of types of targets 1. Therefore, as the amount of learning progresses, when the new ISAR image data shooting described later is for an unknown target 1, it is also possible to determine the presence or absence of a similar type and to determine that there is no similar one.

[0019] In FIGS. 1 and 3, when the ISAR image data learning device 2 classifies the corresponding target 1 from the ISAR image data obtained by the inverse synthetic aperture radar 12 installed at the observation point and determines the type corresponding to the target 1, it learns different ISAR image data for the same target 1, and generates a learning result so as to be able to limit the range of the learning result used for classifying the target 1 (the ISAR image data learning device according to Embodiment 1). The ISAR image input unit 3 is configured to input ISAR image data for each target 1. The difference information input unit 4 is configured to input difference information for each of the ISAR image data input to the ISAR image input unit 3. The learning unit 5 learns by associating the ISAR image data corresponding to the difference information input to the difference information input unit 4 with the type of the target 1. Information on the type of the target 1 may be attached to the ISAR image data. Also, the learning unit 5 may generate a learning result for each difference information.

[0020] Preferably, the difference information input unit 4 has the input difference information being any two or more of the aspect angle, wave height information, and integration time (observation time) among the three. Further, specifically, among the input difference information, the aspect angle is the angle of the transmission beam of the inverse synthetic aperture radar 12 with respect to the target 1, the wave height information is the information on the wave height of the sea area where the target 1 exists, and the integration time is the time derived by integrating the ISAR image data. Since a plurality of ISAR images are superimposed according to the integration time (observation time) as the difference information to form the ISAR image data, the ISAR image data will change if the integration time for obtaining the same ISAR image data of the same target 1 changes.

[0021] The aspect angle as the difference information can be said to be the angle of the transmission beam of the inverse synthetic aperture radar 12 with respect to the linear structure of the target 1 when the fuselage of the target 1 has a linear structure. Also, the aspect angle can be said to be the angle formed by the transmission beam direction of the ISAR 12 from the observation point and the moving direction of the target 1 relative to the ISAR 12 at the observation point. In any case, the aspect angle is composed of the azimuth angle and the elevation angle. The angle in the azimuth direction can be fixed (or set to no value), and the aspect angle can be the information of only the elevation angle, or the angle in the elevation direction can be fixed (or set to no value), and the aspect angle can be the information of only the azimuth angle. The wave height information as the difference information is the value of the class set according to the wave height. The value of the class set according to the wave height is, for example, the value of the sea state class that divides the wave height into 10 classes from 0 to 9.

[0022] As described above, the difference information input unit 4 has the aspect angle calculated from the position of the observation point and the information on the moving direction of the target 1 which is a moving object being input as the difference information. Also, the difference information input unit 4 has the aspect angle calculated from the position of the observation point and the information on the moving direction obtained by tracking the target 1 being input as the difference information. Further, the difference information input unit 4 has the value of the class set according to the wave height as the wave height information being input as the difference information.

[0023] Also, although the integration time and the observation time have been equivalently explained as difference information, strictly speaking, they can be treated differently. Here, the integration time is the time delimited by the integration start time and the integration end time. Similarly, the observation time is the time delimited by the observation start time and the observation end time. For example, in the difference information input unit 4, as long as the integration time that is the ISAR image data of the same target 1 and in which the integration start time and the integration end time fall between the observation start time and the observation end time of the target 1 is input. Of course, even if the integration times are the same, the times from the observation start time to the observation end time of the target 1 may be different. In an example included in this, even if the integration times are different, the difference information input unit 4 may input integration times in which the integration start times are the same. Similarly, the difference information input unit 4 may input integration times in which the integration start times are different.

[0024] Furthermore, the difference information input unit 4 may input, as difference information, an integration time in which the integration start time is the same and the integration end time comes within the time until one round trip of the fluctuation period of the target 1. Next, the difference information input unit 4 may input, as difference information, an integration time in which the time from the integration start time to the integration end time corresponds to one round trip of the fluctuation period of the same target 1. Of course, in the ISAR image input unit 3, ISAR image data of the same target 1 in which the integration times are different may be input as difference information.

[0025] As the type of the target 1, information including at least one of the model name of the target 1, the dimensions of the airframe, the performance of the airframe, and the country / organization to which the airframe belongs may be sufficient. Also, even if the types of the same target 1 are the same, if the difference information for each ISAR image data is different, the ISAR image data is different. That is, the ISAR image data input to the ISAR image input unit 3 may be different even for the same target 1 when the difference information input to the difference information input unit 4 is different. The ISAR image data may be simulation data instead of the data actually acquired by the ISAR 12. The difference information such as the aspect angle and the wave height information and the integration time (observation time) may be simulation data instead of the data actually acquired.

[0026] FIG. 1(A) shows an ISAR image data learning apparatus 2 in which the learning unit 5 does not learn the information on the distance from the observation point to target 1. On the other hand, FIG. 1(B) shows an ISAR image data learning apparatus 2 in which the learning unit 5 learns the information on the distance from the observation point to target 1. As shown in FIG. 1(B), the input of the distance information to be learned by the learning unit 5 may be via the ISAR image input unit 3 or via the difference information input unit 4. That is, the ISAR image input unit 3 may be configured to receive ISAR image data associated with the information on the distance from the observation point to target 1. Also, the difference information input unit 4 may be configured to receive difference information associated with the information on the distance from the observation point to target 1. If there is information on the distance from the observation point to target 1, the learning unit 5 can learn the ISAR image data at a predetermined scale from the distance information.

[0027] FIG. 3 illustrates a case where the ISAR image input unit 3 and the difference information input unit 4 receive the ISAR image data and the integration time output from the ISAR image generation unit 13, respectively. That is, it is an additional explanation of the case where the integration time is input as the difference information to the difference information input unit 4. FIG. 3(A) corresponds to FIG. 1(A), and FIG. 3(B) corresponds to FIG. 1(B). That is, FIG. 3(A) shows an ISAR image data learning apparatus 2 in which the learning unit 5 does not learn the information on the distance from the observation point to target 1. FIG. 3(B) shows an ISAR image data learning apparatus 2 in which the learning unit 5 learns the information on the distance from the observation point to target 1. Note that the ISAR image data learning apparatus 2 may include the ISAR image generation unit 13 or may have the ISAR image generation unit 13 provided externally.

[0028] At the observation point for observing Target 1, a radar device 10 (tracking radar 11, ISAR 12) is installed. The observation point may move. That is, the radar device 10 may be mounted on a moving object such as a ship or an aircraft. The scanning range is the range that can be scanned from the radar device 10 (tracking radar 11, ISAR 12) centered on the observation point. It is conceivable that the radar device 10 can scan 360 degrees in the azimuth direction. The observation point (radar device 10) assumes a ship or an aircraft that moves by itself, but the observation point may be a fixed point.

[0029] Even if it is the same type of Target 1, if the aspect angle at which the ISAR image data is acquired is different, the shape of Target 1 represented in the ISAR image data is different. If Target 1 has the aforementioned linear structure, this is prominent. That is, the ISAR image data input to the ISAR image input unit 3 may have a different aspect angle as the difference information input to the difference information input unit 4 even for the same Target 1. The ISAR image data may be simulation data instead of the data actually acquired by the ISAR 12. The aspect angle may be simulation data instead of the data actually acquired by the tracking radar 11 (described later). Although Target 1 is preferably a linear structure, Target 1 may be defined such that if the aspect angle is different, the shape of Target 1 represented in the ISAR image data is different. Also, even for the same Target 1, it may be possible to select and learn the aspect angles for which the shape of Target 1 represented in the ISAR image data becomes different.

[0030] Similarly, even for the same type of target 1, the wave height information as the difference information input to the difference information input unit 4 may be different. For example, when the wave height in the sea area where target 1 exists increases, the sway of target 1 becomes larger, and the movement amount per unit time of target 1 becomes larger. Therefore, the speed of target 1 increases, and the Doppler frequency included in the radar echo of target 1 becomes higher. The ISAR image data is generated based on the Doppler frequency included in the radar echo of target 1. Therefore, when the wave height in the sea area where target 1 exists is high and low, the shape of target 1 represented in the generated ISAR image data is different. That is, even for the same type of target 1, when the wave height in the existing sea area is different, the shape of target 1 represented in the generated ISAR image data is different. In particular, in target 1, in parts such as masts that are far from the water surface, compared with the parts near the water surface, the sway due to waves is larger, so the change in Doppler frequency is significant.

[0031] Next, the operation of the ISAR image data learning device according to Embodiment 1 (the ISAR image data learning method according to Embodiment 1) will be described with reference to FIG. 2. In FIG. 2, step 1 is a processing step in which ISAR image data is input to the ISAR image input unit 3 for each target 1. Step 2 is a processing step in which difference information for each piece of ISAR image data input to the ISAR image input unit 3 is input to the difference information input unit 4. The order of steps 1 and 2 does not matter. They may be simultaneous. Step 3 is a processing step in which, based on the ISAR image data and the difference information, the learning unit 5 associates and learns the ISAR image data corresponding to the difference information input to the difference information input unit 4, which is the basis for classification (determination), with the type of target 1. Information on the type of target 1 may be given to the ISAR image data when step 1 is performed.

[0032] In FIGS. 4, 6, 8, 9, and 10 described below, preferably, the new difference information input to the new difference information input unit 8 is any two or more of the aspect angle, wave height information, and integration time (observation time). More specifically, among the new difference information input to the new difference information input unit 8, the aspect angle is the angle of the transmission beam of the inverse synthetic aperture radar 12 with respect to target 1, the wave height information is the wave height information of the sea area where target 1 exists, and the integration time is the time when new ISAR image data is derived by integration. Since a plurality of new ISAR images are superimposed according to the integration time (observation time) as the difference information to form new ISAR image data, even for the new ISAR image data of the same target 1, if the integration time for obtaining it changes, the new ISAR image data will change.

[0033] In FIG. 4, the target classification device 6 uses the learning result (learning model) of the ISAR image data learning device 2 shown in FIGS. 1 and 4 (the target classification device according to Embodiment 1). This is a suitable configuration when using the integration time as the difference information and the new difference information. When used in combination with the target classification device 6 shown in FIGS. 6, 8, 9, and 10 described below, at least one of the aspect angle or the sea state can be used as the difference information and the new difference information.

[0034] In FIG. 4, the new ISAR image input unit 7 is for inputting newly obtained ISAR image data. The new difference information input unit 8 inputs the integration time for each newly obtained ISAR image data input to the new ISAR image input unit 7 as the difference information. The new ISAR image data (newly obtained ISAR image data) and the new integration time (newly obtained integration time) here may be obtained by the radar device 10 (the radar device according to Embodiment 1) described below or the calculation of the subsequent circuit. The calculation of the subsequent circuit may be, for example, using the ISAR image generation unit 13 for new calculations or using the integration processing unit 14 described below.

[0035] In FIG. 4, the target classification unit 9 classifies target 1 corresponding to newly obtained ISAR image data based on the learning result learned by the learning unit 5, and determines the type corresponding to target 1. Further, when the learning unit 5 has also learned the information on the distance from the observation point to target 1, the target classification unit 9 may convert the newly obtained ISAR image data into a predetermined scale based on the learning result learned by the learning unit 5, and then classify the corresponding target 1 to determine the type corresponding to target 1.

[0036] In order to speed up the processing, the target classification unit 9 may limit the range of the learning result used for classification from the difference information for each newly obtained ISAR image data. When the learning unit 5 generates a learning result for each difference information, the target classification unit 9 can classify target 1 and determine the type corresponding to target 1 based on the learning result learned by the learning unit 5 and generated for each difference information.

[0037] As the learning of the learning unit 5 progresses, when there is no difference information for each newly obtained ISAR image data that matches the difference information input to the difference information input unit 4, or when the error from the difference information input to the difference information input unit 4 is not within a predetermined range, it becomes possible to classify target 1 and determine the type corresponding to target 1 based on the learning result of the closest angle.

[0038] Further, the new ISAR image input unit 7 may be configured to input a plurality of newly obtained ISAR image data in which target 1 is the same and the integration time, which is the difference information, is different. Further, the new ISAR image input unit 7 may be configured to input a plurality of newly obtained ISAR image data in which target 1 and the integration time are respectively the same and the integration start time and the integration end time of the integration time are different. In these cases, since the target classification unit 9 can classify target 1 corresponding to each of the plurality of newly obtained ISAR image data and determine the type corresponding to target 1 based on the learning result learned by the learning unit 5, the accuracy of the determination is improved.

[0039] Of course, in these cases (when a plurality of newly obtained ISAR image data with the same target 1 but different integration times are input, or when a plurality of newly obtained ISAR image data with the same target 1 and integration time but different integration start times and integration end times for the integration time are input), if the results of classifying the corresponding target 1 for each of the plurality of newly obtained ISAR image data all match, the target classification unit 9 may determine the type corresponding to target 1. Also, if the results of classifying the corresponding target 1 for each of the plurality of newly obtained ISAR image data include those that do not match, and when the number of those that do not match is less than the number of those that match, the target classification unit 9 may determine the type corresponding to target 1 as the result of classifying the matching ones.

[0040] If the learning unit 5 comprehensively learns using a plurality of ISAR image data with different integration times, the target classification unit 9 can also classify comprehensively. Since the integration time can freely generate data for various cases in signal processing, it is clear that it is more advantageous to use the integration time during both the learning of the learning unit 5 and the classification of the target classification unit 9 than to use natural environmental conditions.

[0041] Similarly, in FIG. 4, the radar device 10 having the target classification device 6 (the ISAR image data learning device 2 and the target classification device 6) has the following configuration. The tracking radar 11 is the same radar as described above and tracks target 1. The inverse synthetic aperture radar 12 (ISAR 12) transmits a transmission beam to target 1 tracked by the tracking radar 11 and receives the radio wave reflected from target 1. The integration processing unit 14 integrates the video signal obtained from the radio wave reflected from target 1 in order to generate newly obtained ISAR image data and send it to the new ISAR image input unit 7. That is, it can be said that the integration processing unit 14 (similarly for the ISAR image generation unit 13) generates ISAR image data by superimposing a plurality of ISAR images.

[0042] As described above, the new ISAR image input unit 7 is configured to receive a plurality of newly obtained ISAR image data in which the target 1 is the same but the difference information is different. Therefore, based on the learning result learned by the learning unit 5, the target classification unit 9 can classify the corresponding target 1 for each of the plurality of newly obtained ISAR image data and determine the type corresponding to the target 1, thereby improving the accuracy of the determination.

[0043] The details of the case where the aspect angle calculated from the position of the observation point and the information on the moving direction of the target 1 which is a moving object is input to the difference information input unit 4 as difference information will be described. As shown in FIG. 5, the difference information input unit 4 may be configured to receive, as difference information, the aspect angle calculated from the position of the observation point and the information on the moving direction obtained by tracking the target 1 with the tracking radar 11. In FIG. 5, a radar device 10 (tracking radar 11, ISAR 12) is grounded at the observation point. The observation point may move. That is, the radar device 10 may be mounted on a moving object such as a ship or an aircraft. The scanning range 15 (imaging range 15) is a virtual indication by a dotted line of the range within which the radar device 10 (tracking radar 11, ISAR 12) centered on the observation point can scan. The radar device 10 is exemplified as being capable of scanning 360 degrees in the azimuth direction.

[0044] In FIG. 5, an example is shown where target 1 is a ship. FIG. 5(A) shows the case where target 1 is observed from the side, and FIG. 5(B) shows the case where target 1 is observed from behind. Since target 1 is a ship (one having a normal linear structure), it can be easily understood that the ISAR image data obtained depending on the respective situations of FIG. 5(A) and FIG. 5(B) differ in shape even though they are the same target 1. Also, as described above, the observation point (radar device 10) in FIG. 5 assumes a ship or an aircraft that moves by itself, but the observation point may be a fixed point. Note that in FIG. 5, for the sake of simplification of the explanation, the angle in the elevation direction is fixed at 0°, that is, the elevation angle is fixed at 0°, and the aspect angle is virtually illustrated as planar information with only the azimuth angle as information. Here, the elevation angle refers to the angle rising from the horizontal. On the other hand, the angle falling from the vertical direction is called the incident angle. The relationship between the elevation angle and the incident angle is elevation angle = 90° - incident angle. Therefore, it can be said that the aspect angle is composed of the azimuth angle and the incident angle.

[0045] Even when a plurality of newly obtained ISAR image data with different aspect angles, which are different information but for the same target 1, are input to the new ISAR image input unit 7, the target classification unit 9 can classify the corresponding target 1 for each of the plurality of newly obtained ISAR image data based on the learning result learned by the learning unit 5 and determine the type corresponding to target 1, so the accuracy of the determination is improved. Of course, when the results of classifying the target 1 corresponding to each of the plurality of newly obtained ISAR image data all match, the target classification unit 9 may determine the type corresponding to target 1. Also, when the results of classifying the target 1 corresponding to each of the plurality of newly obtained ISAR image data include those that do not match and the number of those that do not match is small compared to the number of those that match, the type corresponding to target 1 may be determined based on the result of classifying those that match.

[0046] In FIG. 6, the target classification device 6 uses the learning result (learning model) of the ISAR image data learning device 2 shown in FIGS. 1 and 6 (the target classification device according to Embodiment 1). This is a configuration suitable for using the aspect angle as the difference information and the new difference information. If used in combination with the target classification device 6 shown in FIGS. 4, 8, 9, and 10 described later, at least one of the integration time and the sheath state can be used as the difference information and the new difference information.

[0047] In FIG. 6, the new ISAR image input unit 7 is for inputting newly obtained ISAR image data. The radar device 10 having the target classification device 6 (the ISAR image data learning device 2 and the target classification device 6) has the following configuration. The tracking radar 11 is the same radar as described above and tracks the target 1. The inverse synthetic aperture radar 12 (ISAR 12) is the same radar as described above and transmits a transmission beam to the target 1 tracked by the tracking radar 11 in order to send the newly obtained ISAR image data to the new ISAR image input unit 7. Specifically, the tracking radar 11 tracks the target 1 in order to send the aspect angle for each newly obtained ISAR image data to the new difference information input unit 8.

[0048] An exemplary diagram of the radar device 10, the scanning range 15, and the target 1 when the radar device 10 acquires new ISAR image data and a new aspect angle is the same as FIG. 5. In FIG. 5, the target 1 is a moving ship moving from FIG. 5(A) to FIG. 5(B) as an example. Thus, when moving from FIG. 5(A) to FIG. 5(B), as described above, the new ISAR image input unit 7 inputs a plurality of newly obtained ISAR image data in which the target 1 is the same and the aspect angle as the difference information is different. Therefore, the target classification unit 9 can classify the target 1 corresponding to each of the plurality of newly obtained ISAR image data based on the learning result learned by the learning unit 5 and determine the type corresponding to the target 1, so that the accuracy of the determination is improved.

[0049] The details will be described when the information on the wave height calculated from the signal intensity of the sea clutter in the sea area where the target 1 obtained from the inverse synthetic aperture radar 12 (ISAR12) exists is input to the difference information input unit 4 as difference information. The difference information input unit 4 may be one to which the information on the wave height obtained from a device for measuring the wave height of the sea area is input. Further, the difference information input unit 4 may be one to which the information on the wave height obtained from the meteorological information in the sea area where the target 1 exists is input as the information on the wave height. In FIG. 7, a radar device 10 (ISAR12) is installed at the observation point. The observation point itself may move. That is, the radar device 10 may be mounted on a moving body such as a ship or an aircraft. The scanning range 15 is virtually shown by a dotted line as the range that the radar device 10 (ISAR12) centered on the observation point can scan. The radar device 10 is exemplified as one that can scan 360 degrees in the azimuth direction.

[0050] In FIG. 6, the target classification unit 9 has the same configuration as the target classification unit 9 shown in FIG. 4, and classifies the target 1 corresponding to the newly obtained ISAR image data based on the learning result learned by the learning unit 5, and determines the type corresponding to the target 1. Also, in the same configuration shown in FIG. 4, when the learning unit 5 has also learned the information on the distance from the observation point to the target 1, the target classification unit 9 may convert the newly obtained ISAR image data into a predetermined scale based on the learning result learned by the learning unit 5, and then classify the corresponding target 1 and determine the type corresponding to the target 1.

[0051] In FIG. 6, in the same configuration shown in FIG. 4, the target classification unit 9 may limit the range of the learning result used for classification from the difference information for each newly obtained ISAR image data in order to speed up the processing. Note that, in the same configuration shown in FIG. 4, when the learning unit 5 has generated a learning result for each difference information, the target classification unit 9 can classify the target 1 and determine the type corresponding to the target 1 based on the learning result generated by the learning unit 5 for each difference information learned.

[0052] In FIG. 6, with the same configuration as shown in FIG. 4, as the learning in the learning unit 5 progresses, when there is no match between the difference information for each newly obtained ISAR image data and the difference information input to the difference information input unit 4, or when the error from the difference information input to the difference information input unit 4 is not within a predetermined range, the target classification unit 9 can classify target 1 based on the learning result at the closest angle and determine the type corresponding to target 1.

[0053] In FIGS. 8, 9, and 10, the target classification device 6 uses the learning results (learning models) of the ISAR image data learning device 2 shown in FIGS. 1 and 8, FIGS. 1 and 9, and FIGS. 1 and 10 (the target classification device according to Embodiment 1). It is a configuration suitable for using the sheath state as the difference information and the new difference information. When used in combination with the target classification device 6 shown in FIGS. 4 and 6, at least one of the integrated information or the aspect angle can be used as the difference information and the new difference information.

[0054] In FIGS. 8, 9, and 10, the new ISAR image input unit 7 is for inputting newly obtained ISAR image data. The new difference information input unit 8 inputs the wave height information for each newly obtained ISAR image data input to the new ISAR image input unit 7 as the difference information. The new wave height information (newly obtained wave height information) here may be acquired from a device for measuring the wave height, or from the inverse synthetic aperture radar 12 (ISAR12), or from meteorological information. FIG. 8 is a functional block diagram of the ISAR image data learning device, the target classification device, and the radar device when the new wave height information (newly obtained wave height information) is acquired from a device for measuring the wave height. FIG. 9 is a functional block diagram of the ISAR image data learning device, the target classification device, and the radar device when the new wave height information (newly obtained wave height information) is acquired from the inverse synthetic aperture radar 12 (ISAR12). FIG. 10 is a functional block diagram of the ISAR image data learning device, the target classification device, and the radar device when the new wave height information (newly obtained wave height information) is acquired from meteorological information.

[0055] Specifically, as shown in FIG. 8, the new difference information input unit 8 may receive wave height information obtained from a device (wave height measuring device 16) that measures the wave height in the sea area where target 1 exists. The wave height measuring device 16 may be, for example, a radar different from the ISAR 12, a pressure type wave height meter, a buoy type wave height meter, etc. Note that the wave height may be measured visually instead of the wave height measuring device 16. The wave height measuring device 16 may be configured to be provided in the radar device 10, but as shown in FIG. 8, it may be configured outside the radar device 10. For example, the radar device 10 (ISAR 12) may be mounted on a moving body such as a ship or an aircraft, and the wave height measuring device 16 may be fixed on land. Thus, the radar device 10 and the wave height measuring device 16 may be provided at different locations.

[0056] Also, as shown in FIG. 9, the new difference information input unit 8 may receive, as difference information, wave height information calculated from the signal intensity of sea clutter in the sea area where target 1 exists and obtained from the inverse synthetic aperture radar 12 (ISAR 12). Also, as shown in FIG. 9, the new difference information input unit 8 may receive, as difference information, wave height information obtained from meteorological information in the sea area where target 1 exists as wave height information.

[0057] In FIGS. 8, 9, and 10, the radar device 10 having the target classification device 6 (ISAR image data learning device 2 and target classification device 6) has the following configuration. The inverse synthetic aperture radar 12 (ISAR 12) is the same radar as described above, and transmits a transmission beam to target 1 in order to send the newly obtained ISAR image data to the new ISAR image input unit 7. The illustration of the radar device 10, the scanning range 13, and target 1 when the radar device 10 acquires new ISAR image data is the same as that in FIG. 7.

[0058] In FIGS. 8, 9, and 10, the target classification unit 9 has the same configuration as the target classification unit 9 shown in FIGS. 4 and 6, and classifies the target 1 corresponding to the newly obtained ISAR image data based on the learning result learned by the learning unit 5, and determines the type corresponding to the target 1. Also, in the same configuration shown in FIGS. 4 and 6, when the learning unit 5 has also learned the information on the distance from the observation point to the target 1, the target classification unit 9 may convert the newly obtained ISAR image data to a predetermined scale based on the learning result learned by the learning unit 5, and then classify the corresponding target 1 and determine the type corresponding to the target 1.

[0059] In FIGS. 8, 9, and 10, with the same configuration as shown in FIGS. 4 and 6, for the purpose of speeding up the processing, the target classification unit 9 may limit the range of the learning result used for classification from the difference information for each newly obtained ISAR image data. In addition, in the same configuration shown in FIGS. 4 and 6, when the learning unit 5 has generated a learning result for each difference information, the target classification unit 9 can classify the target 1 and determine the type corresponding to the target 1 based on the learning result generated by the learning unit 5 for each difference information learned.

[0060] In FIGS. 8, 9, and 10, with the same configuration as shown in FIGS. 4 and 6, as the learning of the learning unit 5 progresses, when there is no difference information for each newly obtained ISAR image data that matches the difference information input to the difference information input unit 4, or when the error from the difference information input to the difference information input unit 4 is not within a predetermined range, it becomes possible to classify the target 1 and determine the type corresponding to the target 1 based on the learning result at the closest angle.

[0061] Finally, the operation of the main target classification device according to Embodiment 1 (the target classification method according to Embodiment 1) will be described with reference to FIG. 11. FIG. 11 shows the operation of the target classification device 6 shown in FIGS. 4, 6, 8, 9, and 10.

[0062] In FIG. 11, step 11 is a processing step in which the newly obtained ISAR image data is input to the new ISAR image input unit 7. Step 12 is a processing step in which the newly obtained ISAR image data is input to the new difference information input unit 8. Steps 11 and 12 may be in any order of processing. They may be simultaneous. The new ISAR image data (newly obtained ISAR image data) and the new difference information (newly obtained difference information) mentioned here may be acquired by the radar device 10 (the radar device according to Embodiment 1).

[0063] Step 13 is a processing step of inputting the new ISAR image data (newly obtained ISAR image data) and the new difference information (newly obtained difference information) from the new ISAR image input unit 7 and the new difference information input unit 8 to the learning unit 5 and using the learning model. Step 14 is a processing step of classifying the target 1 corresponding to the newly obtained ISAR image data based on the learning result (learning model) learned by the learning unit 5 and determining the type corresponding to the target 1. Since the target classification method according to other Embodiment 1 is the same as the operation of the target classification device mainly according to the aforementioned Embodiment 1, the description thereof is omitted.

[0064] In the ISAR image data learning device according to Embodiment 1, the target classification device using the same, and the radar device, the ISAR image data and the difference information learned by the learning unit 5 may be simulated data obtained by simulation. Similarly, the new ISAR image data and the difference information classified by the target classification unit 9 may also be simulated data obtained by simulation. By classifying the target classification unit 9 with this simulated data, the performance of the learning unit 5 and the target classification unit 9 can be confirmed. That is, the ISAR image data and the difference information generated by the ISAR image generation unit 13 may also be simulated data obtained by simulation.

[0065] In the ISAR image data learning device according to Embodiment 1, the target classification device using the same, and the radar device, in order for the learning unit 5 to learn, if the difference information input to the difference information input unit 4 is any two or more of the aspect angle, wave height information, and integration time, the performance will be improved. On the other hand, in the target classification device and the radar device using the ISAR image data learning device according to Embodiment 1, if the difference information input to the new difference information input unit 8 is set to any two or more of the aspect angle, wave height information, and integration time, the accuracy will be improved when the target classification unit 9 classifies Target 1 using the learning result of the learning unit 5. Even if the difference information corresponding to the ISAR image data being learned by the learning unit 5 is the aspect angle and wave height information, and the new difference information input to the new difference information input unit 8 is the wave height information and integration time, classification can be performed using the wave height information.

[0066] Furthermore, in the ISAR image data learning device according to Embodiment 1, the target classification device using the same, and the radar device, in order for the learning unit 5 to learn, using all three of the aspect angle, wave height information, and integration time as the difference information input to the difference information input unit 4 can reduce the burden on the new difference information input unit 8. Strictly speaking, the more types of difference information to be learned, the more the burden on the new difference information input unit 8 can be reduced. That is, even if the difference information input to the new difference information input unit 8 is only one of the aspect angle, wave height information, and integration time, since the learning unit 5 is surely learning the difference information, the target classification unit 9 can perform the classification of Target 1. Of course, regardless of the learning status of the learning unit 5, the difference information input to the new difference information input unit 8 may be any one of the aspect angle, wave height information, and integration time.

[0067] As described above, the ISAR image data learning device according to Embodiment 1, the target classification device using the same, and the radar device associate and learn the ISAR image data corresponding to the difference information with the type of Target 1, or utilize the learning result thereof. Therefore, the type of Target 1 can be determined using the difference information. Further, the ISAR image data learning device according to Embodiment 1, the target classification device using the same, and the radar device use, as the difference information, any two or more of the aspect angle, the wave height information, and the integration time (observation time), thereby further improving the learning efficiency and the classification accuracy.

[0068] The difference information and the new difference information used in the ISAR image data learning device according to Embodiment 1, the target classification device using the same, and the radar device can be implemented with any one of the aspect angle, the wave height information, and the integration time. However, the difference information and the new difference information are not limited thereto. The difference information may be information on the factors that result in different ISAR image data for the same target for each ISAR image data. Similarly, the new difference information may be information on the factors that result in different new ISAR image data for the same target for each new ISAR image data.

Explanation of Reference Numerals

[0069] 1 Target 1, 2 ISAR image data learning device, 3 ISAR image input unit, 4 Difference information input unit, 5 Learning unit, 6 Target classification device, 7 New ISAR image input unit, 8 New difference information input unit, 9 Target classification unit, 10 Radar device, 11 Tracking radar, 12 Inverse synthetic aperture radar (ISAR), 13 ISAR image generation unit, 14 Integration processing unit, 15 Scanning range (imaging range), 16 Wave height measuring device.

Claims

1. An ISAR image data learning device that generates a learning result for classifying a target from ISAR image data obtained by an inverse synthetic aperture radar installed at an observation point, an ISAR image input unit into which the ISAR image data is input; a difference information input unit into which at least two of the aspect angle, which is the angle of the transmission beam of the inverse synthetic aperture radar with respect to the target, information on the wave height of the sea area where the target exists, and the integration time from which the ISAR image data was derived are input as difference information; and a learning unit that learns to classify the target from the ISAR image data corresponding to the difference information input to the difference information input unit. The integration time is an ISAR image data learning device in which an integration start time and an integration end time are set between the observation start time and the observation end time of the target.

2. The ISAR image data learning device according to claim 1, wherein the learning unit learns by associating the ISAR image data corresponding to the difference information input to the difference information input unit with the type of the target.

3. The ISAR image data learning device according to claim 2, wherein the type of the target is information including at least one of the name of the target, the dimensions of the target, the performance of the target, the country to which the target belongs, and the organization to which the target belongs.

4. The ISAR image data learning device according to any one of claims 1 to 3, wherein the aspect angle is calculated from the position of the observation point and information on the moving direction of the target, which is a moving object.

5. The ISAR image data learning device according to claim 4, wherein the aspect angle is calculated from the position of the observation point and the information on the moving direction obtained by tracking the target.

6. The ISAR image data learning device according to any one of claims 1 to 5, wherein the information on the wave height is a value of a wind wave class set according to the wave height of the sea area where the target exists.

7. The ISAR image data learning device according to any one of claims 1 to 6, wherein the integration time corresponds to one round trip of the period of the target fluctuation from the integration start time to the integration end time.

8. An ISAR image input unit into which ISAR image data obtained by an inverse synthetic aperture radar installed at an observation point is input, an aspect angle which is an angle of a transmission beam of the inverse synthetic aperture radar with respect to a target, information on the wave height of a sea area where the target exists, and at least two of the integration times from which the ISAR image data is derived are input as difference information. A difference information input unit, and a learning unit that learns to classify the target from the ISAR image data corresponding to the difference information input to the difference information input unit. The integration time is an ISAR image data learning device in which an integration start time and an integration end time are set between an observation start time and an observation end time of the target, A new ISAR image input unit into which newly obtained new ISAR image data is input, A new difference information input unit into which the difference information corresponding to the new ISAR image data is input, A target classification device including a target classification unit that classifies the target from the new ISAR image data based on a learning result learned by the learning unit.

9. The target classification device according to claim 8, wherein the target classification unit restricts a range of the learning result used for classification from the difference information for each of the ISAR image data.

10. The target classification device according to claim 9, wherein the target classification unit classifies the target based on the learning result generated by the learning unit for each of the difference information, and determines a type corresponding to the target.

11. When there is no match between the difference information corresponding to the new ISAR image data and the difference information input to the difference information input unit, or when the error from the difference information input to the difference information input unit is not within a predetermined range, the target classification unit classifies the target based on the learning result at the closest time. The target classification device according to any one of claims 8 to 10.

12. The new ISAR image input unit receives a plurality of new ISAR image data with the same target and different difference information. The target classification unit classifies the target corresponding to each of the plurality of new ISAR image data based on the learning result learned by the learning unit. The target classification device according to any one of claims 8 to 11.

13. When the results of classifying the targets corresponding to each of the plurality of new ISAR image data all match, the target classification unit determines the type corresponding to the target. The target classification device according to claim 12.

14. If the results of classifying the targets corresponding to each of the plurality of new ISAR image data include those that do not match, and the number of those that do not match is less than the number of those that match, the target classification unit determines the type corresponding to the target as the result of classifying the matching ones. The target classification device according to claim 13.

15. An ISAR image input unit into which ISAR image data obtained by an inverse synthetic aperture radar installed at an observation point is input, an aspect angle that is the angle of the transmission beam of the inverse synthetic aperture radar with respect to the target, information on the wave height of the sea area where the target exists, and at least two of the integration times for deriving the ISAR image data are input as difference information to a difference information input unit, and a learning unit that learns to classify the target from the ISAR image data corresponding to the difference information input to the difference information input unit. The integration time is an ISAR image data learning device in which an integration start time and an integration end time are set between the observation start time and the observation end time of the target. A target classification device having a new ISAR image input unit into which newly obtained new ISAR image data is input, a new difference information input unit into which the difference information corresponding to the new ISAR image data is input, and a target classification unit that classifies the target from the new ISAR image data based on the learning result learned by the learning unit. A tracking device that tracks the target. A radar device including the inverse synthetic aperture radar that transmits a transmission beam to the target tracked by the tracking device in order to acquire the new ISAR image data.

Images