JP2005208023A - Target-detecting apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To solve the problem that the conventional apparatus cannot cope with an unknown target due to the need for previously holding the spectra luminance data of a significant subject, and that detection capability degrades if a plurality of background clutter is contained due to another apparatus treating a background clutter as one variety. <P>SOLUTION: The apparatus comprises two imaging parts for capturing the images of a subject to be in the same visual filed in mutually different wavelength bands; a distance computing section for calculating the distance between a coordinate point that the outputs obtained by the two imaging parts are plotted respectively in a rectangular coordinate system with the outputs as the vertical axis and the horizontal axis and the origin of in the coordinate system; an angle computing section for calculating the angle to form with the straight line between the coordinate point and the origin, and the horizontal axis; and a subject deciding section for extracting a significant subject by treating the values obtained by the two computing sections. <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a target detection apparatus that is installed in a moving body such as an aircraft or a fixed facility and detects and determines a significant target from an image signal.

The technology of the target detection device can be roughly divided into two.
In the first technique, significant target features (luminance, spectral characteristics, etc.) are stored in advance as data, and a significant target is extracted by comparing the data with potential target with the stored data. Technology.
The second technique is a technique for detecting a significant target by analyzing characteristics of output (background clutter) data based on background from acquired data and extracting data having a property different from that of the background.

  FIG. 11 is a block diagram of a conventional target detection apparatus disclosed by Japanese Patent Laid-Open No. 08-068844 as an example of the first prior art. In FIG. 11, the infrared imaging unit 1 includes two imaging units 1 c and 1 d that can obtain subject images with the same field of view. The spectral luminance data unit 12 is a two-dimensional luminance coordinate system in which spectral luminance data of various objects (significant targets and other background clutters, etc.) is used with the luminance signals for the wavelength bands of the two imaging units 1c and 1d as orthogonal axes. As a territory. The luminance normalization units 11 a and 11 b convert the respective image signals obtained by the imaging units 1 c and 1 d of the infrared imaging unit 1 into the same luminance scale as the data held in the spectral luminance data unit 12. The target determination unit 4 compares the plurality of image signals normalized by the luminance normalization units 11a and 11b and the output of the spectral luminance data unit 12 and extracts a significant target included in the target region.

  Examples of wavelength bands used in the infrared imaging unit 1 include mid-infrared rays and far-infrared rays. Hereinafter, an operation when the mid-infrared ray and the far-infrared ray are selected as the wavelength band will be described as an example. FIG. 12A shows an example of data held in the spectral luminance data section 12. The vertical axis represents the far-infrared brightness, the horizontal axis represents the mid-infrared brightness, and the aircraft target area, cloud clutter area, and exhaust gas area are shown in the figure. The cloud data in this figure is based on data from high cumulus clouds and cirrus clouds at an altitude of about 3000 [m] in the daytime captured by an imager sensitive to mid-infrared rays from the ground and an imager sensitive to far-infrared rays. FIG. The aircraft data in the figure is a value calculated on the assumption that the temperature of the cruising aircraft increased due to aerodynamic heating. Further, the exhaust gas data in the figure is a value calculated from the exhaust gas temperature and its components.

  The target determination unit 4 creates a luminance map as shown in FIG. 12B from the outputs of the luminance normalization units 11a and 11b. In this figure, the luminance at the same position (pixels having the same coordinates) is plotted on the luminance map in the images of the far-infrared imaging unit and the mid-infrared imaging unit, and this operation is performed on all the pixels. FIG. 12B is compared with the target region of the spectral luminance data stored in the spectral luminance data unit 12 to detect a significant target. FIG. 12C is an example in which a significant target is determined by superimposing FIG. 12A on FIG. 12B, and corresponds to the plot surrounded by the ellipse of the aircraft target in FIG. A significant target can be detected by extracting pixels to be processed.

  FIG. 13 is a block diagram of a conventional target detection device according to “Multi-dimensional signal processing for electro-optical detection,” SPIE Vol. 1305 (1990) shown by AD Stocker et al. As a second prior art example. . The imaging unit 1 can obtain subject images having the same field of view in a plurality of wavelength bands. The high-pass filter 13 extracts only spatial high-frequency components by removing fluctuations due to offset components of the image output. The filter calculation unit 14 calculates and determines parameters of an adaptive two-wavelength filter that extracts a significant target based on the background pixel output distribution around the target. The target determination unit 4 extracts a significant target by applying the filter obtained by the filter calculation unit 14 to the pixel output.

  In the second prior art, the filter calculation unit 14 calculates a filter that normalizes all pixel outputs according to the characteristics of the background clutter output. 14A shows the output distribution of the imaging unit 1 obtained after applying the high-pass filter 13, and FIG. 14B shows the output obtained after applying the filter obtained by the filter calculating unit 14 to the target determining unit 4. The distribution is shown simply. When the filter calculated by the filter calculation unit 14 is applied to the pixel output by the target determination unit 4, an output having spectral characteristics different from the background clutter output which is the output of most pixels is output with the difference being increased. It is possible to extract a significant target having a property different from that of the background clutter.

Japanese Patent Laid-Open No. 08-068844 A. D. Stocker, IS Reed, X. Yu "Multi-dimensional signal processing for electro-optical detection," SPIE Vol. 1305 (1990). P218-231

  In the target detection device based on the conventional first technique as described above, it is necessary to store spectral luminance data of a significant target in advance, and it has not been possible to deal with an unknown target. In addition, since the spectral brightness changes depending on the environmental conditions such as the atmospheric temperature, visibility, and the effects of sunlight scattering even for known targets, it is a reality to store the spectral brightness data for all these conditions in advance. There was a problem that it was very difficult.

On the other hand, in the target detection device based on the second prior art, processing is performed assuming that there is always one type of background clutter. Therefore, when a plurality of background clutters are included, it is determined as a background clutter. The area expands. Also, the distribution estimate itself is inaccurate. As a result, there is a problem in that the detection ability decreases because the area detected as a significant target decreases.
The present invention has been made to solve the above-described problems, and it is not necessary to retain target characteristic data. The influence of environmental changes is small, and physical characteristics with respect to infrared wavelengths are utilized. An object of the present invention is to obtain a target detection device in which detection performance does not deteriorate even when a plurality of types of background clutter exist.

The target detection apparatus according to the present invention is:
Two imaging units that obtain subject images of the same field of view in different wavelength bands;
A coordinate point when the output obtained by the two imaging units is plotted in an orthogonal coordinate system having a vertical axis and a horizontal axis, respectively, and a distance calculation unit that calculates a distance from the origin of the coordinate system to the coordinate point;
An angle calculation unit that calculates an angle formed by a horizontal axis and a straight line connecting the coordinate point and the origin;
A target determination unit for extracting a significant target by performing threshold processing on values obtained by the two calculation units is provided.

  The target detection apparatus according to the present invention includes a straight line connecting the coordinate point and the origin of the coordinate axis when the subject image output of two imaging units having different wavelength bands is plotted on the vertical axis and the horizontal axis, respectively, and the horizontal axis. The type of object can be separated from the angle formed, and the threshold value according to this angle is used to determine the distance from the origin of the coordinate axis to the coordinate point to determine whether it is a significant target. Target detection is possible even if it is not. Further, even when a plurality of substances are included as background clutter, since they can be separated for each type depending on the angle, it is possible to prevent a decrease in detection performance due to a mixture of a plurality of background clutters.

Embodiment 1 FIG.
FIG. 1 is a block diagram showing Embodiment 1 of the present invention, and FIG. 2 is a characteristic diagram in which the output of the imaging unit is graphed. In FIG. 1, reference numeral 1 denotes an infrared imaging unit, which includes a mid-infrared imaging unit 1b and a far-infrared imaging unit 1a that can obtain a subject image having the same field of view. Reference numeral 2 denotes an angle calculation unit, in which the pixel output obtained by the far-infrared imaging unit 1a is plotted on the horizontal axis, and the pixel output obtained by the mid-infrared imaging unit 1b is plotted on the vertical axis and the coordinate point and origin on the two-dimensional graph. The angle θ of the angle formed by the connecting straight line and the horizontal axis is calculated (FIG. 2). Reference numeral 3 denotes a distance calculation unit that calculates the distance d from the origin at the coordinate point on the above-described two-dimensional graph (FIG. 2). Reference numeral 4 denotes a target determination unit that binarizes the value obtained by the distance calculation unit 3 with a threshold corresponding to the output of the angle calculation unit 2 and extracts a significant target. Specifically, when the output of the distance calculation unit 3 for an arbitrary pixel is equal to or greater than a threshold set according to the output of the angle calculation unit 2, the pixel output is determined to be based on a significant target.

  In FIG. 2, the distance d from the origin varies mainly depending on the target state such as the target distance and temperature, and the angle θ varies depending on the target type, such as when the target emissivity or the target is a flame. Therefore, even if the spectral luminance data of various objects is not provided in advance, the object types can be separated by the angle θ. Naturally, even when a plurality of substances are included as the background clutter, the detection performance can be prevented from being deteriorated due to the mixture of a plurality of background clutters because the types can be separated according to the angle θ. For example, in cloud clutter in the daytime, there are two types of output, that is, the output due to the thermal radiation of the cloud and the output due to sunlight reflection. By setting the threshold value, it is possible to improve the accuracy of the discrimination process.

  In addition, the angle θ is mainly caused by the spectral characteristics, and the spectral characteristics are inherent to the substance. Therefore, the angle θ has almost the same value for the same type of target, and the influence of the environmental change is small. For example, when there is a red target in the visible region, there is almost no change in hue due to environmental changes such as atmospheric temperature change or distance to the target, and it is the same as it looks red in any environment. is there. For example, when the atmospheric propagation distance is long as seen in the sunset, the hue of the object moves toward red, but the blue object jumps over the hue of an object close to red and moves toward red. It does not occur, and the magnitude relationship of the angle θ is relatively preserved. Therefore, the type separation based on the angle θ is always effective regardless of the environmental conditions.

  A detailed operation example is shown below. As described above, in the present embodiment, the angle θ representing the type of the object and the distance d representing the state of the object are calculated from the output obtained by the imaging unit 1. The obtained θ and d are plotted on a graph having orthogonal axes (FIG. 15). When the distribution is calculated for each type, that is, θ, most of the output is background clutter, and thus a background clutter distribution is obtained. In FIG. 15, there are three types of background clutter, which are named A, B, and C. Since the background clutter distribution is obtained for each type based on the angle θ, even if there are a plurality of types of background clutter, the background clutter is not mixed and the distribution does not become inaccurate. Of course, when there is known background clutter data, it goes without saying that the target can be determined with higher accuracy by adding the distribution to the graph of FIG. The boundary of the background clutter distribution is the threshold value. In FIG. 15, a round point outside the background clutter distribution, that is, a round point having a distance d larger than the threshold value by the background clutter B, which is a background clutter having the same angle θ, can be determined as a significant target.

  In the embodiment, the distance d calculated by the distance calculation unit 3 is two-dimensional with the output obtained by the far infrared imaging unit 1a as the horizontal axis and the output obtained by the mid infrared imaging unit 1b as the vertical axis. Although the distance from the origin to the pixel output on the graph is used, it may be the distance from the origin of the target pixel after normalizing the target pixel with the variation amount of the surrounding pixels. As the magnitude of the fluctuation amount of the peripheral pixels, for example, dispersion or PV value can be used. By using this arithmetic processing, for example, an erroneous determination due to noise is likely to occur in an apparatus with a large amount of noise, but there is an effect of normalizing the occurrence frequency of the erroneous determination and keeping it constant.

  Similarly, when the fluctuation amount of the background clutter output is large, an erroneous determination may occur at the maximum point or the minimum point of the output. However, normalizing the pixel of interest with the fluctuation amount of the surrounding pixels is an error. It is also effective in suppressing judgment. In addition, by setting the peripheral pixel area without including the neighboring pixel of the target pixel or the neighboring pixel of the size assumed to be the target, it is possible to prevent the target output from being mixed into the background clutter output. The accuracy of determination can be improved. Specifically, for example, there is a method of calculating the variance or PV value for a region on the ring several pixels away from the pixel of interest and normalizing the result as a variation amount of the surrounding pixels.

  Alternatively, the Mahalanobis general distance D obtained by normalizing the target pixel with the peripheral pixel output in the distance calculation unit 3 may be used. The Mahalanobis generalized distance D is an amount calculated by the equation (1) when the covariance matrix of the peripheral pixel output is Σ, the output of the pixel of interest is s, the average of the peripheral pixel outputs is b, and the vector is a two-wavelength output element. It is. By using this arithmetic processing, the output with different spectral characteristics from the peripheral pixel output (background clutter output) is output with the difference increased, so that it is possible to extract a significant target having a property different from the background. Yes (FIG. 14). It goes without saying that the accuracy of target determination can be improved by setting the neighboring pixels without including neighboring pixels of a size assumed to be adjacent or the target as described above.

  In the above embodiment, when the sensitivity of the mid-infrared imaging unit 1b, that is, the ratio of the output of the unit amount to the infrared input is E3, and the sensitivity of the far-infrared imaging unit 1a is E8, the target determination unit 4 has tan θ = F *. Different thresholds may be set with an angle θ that becomes E3 / E8 (where F is 0.5 <F <2.0) as a boundary. As the unit of the magnitude of the infrared input, for example, the radiance of infrared rays incident on the imaging unit from the target is used, and for example, the output current in the imaging unit is used as the unit of the output magnitude. The target smaller than θ is several tens of degrees even if the temperature is high, and it is a target that is about the same as the atmospheric temperature or aerodynamically heated. However, the target larger than θ is as high as several hundred to several thousand degrees. Heat generation or combustion state (including the case where sunlight is reflected). Since these clearly have different types of targets, it is possible to determine a target with higher accuracy by individually setting a threshold value.

  The target determination unit 4 may use not only the output of one pixel of interest but also the output of peripheral pixels. For example, an average value or a maximum value may be used as the peripheral pixel output at this time. When the corresponding pixels of the mid-infrared imaging unit 1b and the far-infrared imaging unit 1a are misaligned, an output based on a significant target that originally has a large output between the two images is erroneously associated with a small output by the background clutter. Therefore, a significant target output decrease occurs. By providing the target determination unit 4 with the above-described function, it is possible to reduce and correct a decrease in output due to the positional shift of the corresponding pixel, which is effective in improving the determination accuracy of a significant target.

  In addition, the target determination unit 4 may use an average output or a sum of outputs of a total of four pixels of two vertical pixels and two horizontal pixels. For example, when the significant target is imaged smaller than the size of one pixel, the target image captured between the pixels is output across two adjacent pixels. For this reason, by performing the process in the target determination unit 4, it is possible to evaluate the total sum of the outputs for the target image extending between the pixels, which is effective in improving the output accuracy. In addition, the noise reduction effect by taking the average of 4 pixels can be expected (when the average of 4 signals is taken with respect to a signal including white noise, the noise level is reduced to about 1/2). Needless to say.

  Furthermore, the target determination unit 4 may have a function of performing determination processing using the output result of the distance calculation unit 3 obtained from only one output of the mid-infrared imaging unit 1b and the far-infrared imaging unit 1a. As the processing of the distance calculation unit 3 at this time, the output from the normal imaging unit may be output as it is, or may be normalized by the amount of change in the surrounding pixels as described above. For example, with this function, even if one of the imaging units fails, determination processing can be performed, and a device that is resistant to failure can be obtained.

Embodiment 2. FIG.
FIG. 3 is a block diagram showing Embodiment 2 of the present invention. In FIG. 3, reference numerals 1 to 4 are the same as those of the first embodiment. Reference numeral 5 denotes a display unit having a function of displaying the determination result of the target determination unit 4. The display unit 5 has an effect that the user of this apparatus can easily recognize the target orientation.
In the present embodiment, the display unit 5 may be provided with a function for displaying the certainty of the target determination. With this display, the user can grasp the reliability of the determination process. For example, the display method may be indicated by a numerical value, or may be visually expressed such as the color, shape, brightness, and blinking of the target display.

  Further, the type and state of the target may be displayed on the display unit 5. This display allows the user to recognize the target threat level. As the display contents, for example, when the target is an aircraft, the direction of travel such as head-on and tail-on, temperature, use of an afterburner, missile launch, and the like are raised. The display method may be indicated by characters, for example, or may be visually expressed such as the color, shape, brightness, and blinking of the target display. As described above, the type and state of the target may be determined from the output from the angle calculation unit 2 and the state from the output from the distance calculation unit 3, or the presence or absence of a substance may be determined from the output from the angle calculation unit 2. May be judged. For example, when the target is an aircraft, the presence or absence of flame can be determined from the output from the angle calculation unit 2 to determine the state.

  Alternatively, the threat level may be determined from a temporal change on the two-dimensional graph. For example, for an aircraft, the temperature of the exhaust gas changes greatly before and after the afterburner, so that it appears as a time change of the angle θ as well as the kind change. In addition, since the distance d on the two-dimensional graph increases for a target whose distance is approaching, even if it is a target whose size on the image is one pixel or less and it is unknown whether the distance is close, the distance d It can be determined from the time change. Naturally, the distance d on the two-dimensional graph becomes small even for a target that is moving away, so that it can be determined from the time change of the distance d.

  In the present embodiment, the display unit 5 is provided to provide the user with information such as the target position, but the target detection device of the present invention is mounted instead of or in addition to the display unit 5. A function of providing information to the control system of the system that is operating may be provided. For example, improvement in tracking accuracy can be expected by providing the position of the significant target obtained by the target detection apparatus as information of the tracking process. When the target detection device itself enters a tracking state in which the target position is searched for after the target determination, the target position can be grasped with high accuracy. Further, by providing the position of the significant target to the control system of the other sensor, the presence / characteristic of the significant target can be confirmed also by the other sensor, and the detection accuracy can be improved. Alternatively, in a system that applies a physical impact such as a missile to the target, the launch time can be shortened by starting preparation for the launch of the missile or the like when the target is determined by the target detection device. It is possible to give an impact to the target by providing a function for aiming at the target.

Embodiment 3 FIG.
FIG. 4 is a block diagram showing Embodiment 3 of the present invention. In FIG. 4, reference numerals 1 to 4 are the same as those of the first embodiment. A filter unit 6 has a function of reducing output fluctuations due to environmental conditions.
The output from the infrared imaging unit 1 naturally varies depending on environmental conditions such as temperature changes in the infrared imaging unit 1 and visual field conditions between the mid-infrared imaging unit 1b and the far-infrared imaging unit 1a. Therefore, by providing the filter unit 6 in front of the angle calculation unit 2 and the distance calculation unit 3 to compensate or remove the output change due to the environmental condition, it is possible to obtain a target detection device that does not cause performance degradation due to the environmental condition change. .

  For example, the filter unit 6 is provided with a function of operating a high-pass filter that extracts only a spatial high-frequency component for the image output of the infrared imaging unit 1. By this processing, it is possible to remove the offset component fluctuation of the image output caused by, for example, a temperature change in the infrared imaging unit 1. Since such an offset component is a measurement error unrelated to the target output, the accuracy of target determination can be improved by removing the offset component.

  A similar effect can be obtained by providing the filter unit 6 with a function of converting the output into luminance or a target temperature. The infrared imaging unit 1 may use a device that suppresses the output within the dynamic range by artificially adding an offset component because the dynamic range is small, but even in this case, the offset component is complemented. The output can be converted into luminance or target temperature.

  Further, the filter unit 6 may perform processing for adding distortion to the image. For example, when distortion occurs in the optical system of the infrared imaging unit 1, the distortion can be corrected by applying the reverse distortion to the output image, and the target position accuracy can be improved. The magnitude of distortion necessary for correction can be obtained by analyzing from design conditions or by measuring the created optical system.

  The filter unit 6 may have a function of performing a process of translating the pixel coordinate system with respect to one output image obtained by the infrared imaging unit 1. By this process, when the mid-infrared imaging unit 1b and the far-infrared imaging unit 1a do not coincide with each other in the visual axis, the influence of the corresponding pixel can be corrected. As a result, the target position accuracy can be improved, and an error in spectral characteristics caused by the position shift of the corresponding pixel can be corrected, which is effective in improving target determination accuracy.

  As the parallel movement amount necessary for the correction, for example, the positions of the feature points of the two images obtained by the infrared imaging unit 1 may be compared and the shift amount may be used. For example, a point with high luminance or a line with substantially uniform luminance may be used as the feature point. Alternatively, the visual axis shift amount may be measured by providing a gyro in the imaging unit, and the value may be used as the parallel movement amount. Of course, when the amount of positional deviation of the corresponding pixels is always the same, a measurement value of the magnitude acquired in advance may be used.

  Alternatively, the filter unit 6 may have a function of performing enlargement / reduction processing of the pixel coordinate system on one output image obtained by the infrared imaging unit 1. By this processing, it is possible to correct the positional deviation of the corresponding pixels that occurs because the focal lengths of the optical systems do not match between the mid-infrared imaging unit 1b and the far-infrared imaging unit 1a. As a result, the effect of improving the target position accuracy and target determination accuracy is the same as the case where the filter unit 6 performs the parallel movement process. As the enlargement / reduction amount necessary for the correction, for example, the positions of the feature points of two images obtained by the infrared imaging unit 1 may be compared, and the deviation amount may be used. For example, a point or a line with high luminance may be used as the feature point. Alternatively, the design / measurement value of the focal length for the optical system may be used.

Embodiment 4 FIG.
FIG. 5 is a block diagram showing Embodiment 4 of the present invention. In FIG. 5, reference numerals 1 to 4 are the same as those of the first embodiment. Reference numeral 7 denotes a target detection unit having a function of extracting a pixel having a high possibility of being a target by simple processing.
Since the processing performed by the angle calculation unit 2 and the distance calculation unit 3 has a high calculation load, a long processing time is required when these calculations are performed on all acquired pixels. Therefore, by reducing the amount of data to be processed using the target detection unit 7 that extracts only pixels that have a high target possibility by simple processing before that, it is possible to improve the efficiency and speed of these arithmetic processes. it can. The increase in efficiency and speed of the arithmetic processing enables the processing to be realized even in a low-speed and inexpensive calculation system, and the high-speed calculation system is effective in improving the determination accuracy by increasing the number of target determination processes.

  When the pixel output obtained from one infrared imaging unit 1 in the target detection unit 7 is very small or negative, it is detected as an output that may be a target regardless of the magnitude of the other signal output. It may have a function that does not. For example, assume that the significant target is greater than the output of the background clutter, such as when the aircraft is a significant target for cloud background clutter. At this time, if the output from one infrared imaging unit is negative or if the output from one imaging unit is large and the output from the other imaging unit is very small, it is an erroneous determination caused by noise. Probability is high. Therefore, by eliminating such erroneous determinations in the target detection unit 7, the number of data for target determination processing can be efficiently reduced.

  The target detection unit 7 may use not only the output of one pixel of interest but also the output of surrounding pixels. For example, an average value or a maximum value may be used as the peripheral pixel output at this time. When the corresponding pixels of the mid-infrared imaging unit 1b and the far-infrared imaging unit 1a are misaligned, an output based on a significant target that originally has a large output between the two images is erroneously associated with a small output by the background clutter. Therefore, a significant target output decrease occurs. By providing the target detection unit 7 with the above-described function, it is possible to reduce and correct a decrease in output due to the positional shift of the corresponding pixel, which is effective in improving the accuracy of determining a significant target.

  Further, the target detection unit 7 may use an average output or a sum of outputs of a total of four pixels of two vertical pixels and two horizontal pixels. For example, when the significant target is imaged smaller than the size of one pixel, the target image captured between the pixels is output across two adjacent pixels. For this reason, by performing the above-described processing in the target detection unit 7, it is possible to evaluate the total sum of the outputs for the target image extending between the pixels, which is effective in improving the output accuracy. In addition, the noise reduction effect by taking the average of 4 pixels can be expected (when the average of 4 signals is taken with respect to a signal including white noise, the noise level is reduced to about 1/2). Needless to say.

  The target detection unit 7 may have a function of evaluating a target size and extracting only pixels having a size that may be a significant target. For example, since noise is generated by only one pixel, by removing a pixel having only one pixel size from the target determination pixel, erroneous determination due to noise can be reduced, and the determination processing time can be shortened. Further, since background clutter such as clouds and land shapes is generally larger than a significant target, only a significant target can be obtained by extracting only a target having an appropriate size. For the evaluation of the size, for example, the full width at half maximum with respect to the peak output or the diameter of a circular region including a signal of 80% of the output may be used.

  Alternatively, the target detection unit 7 may have a function of extracting a pixel having an output different from the previously acquired pixel output. In general, since the significant target moves, the pixel capturing the significant target captures the background clutter at the time of the previous pixel acquisition, and the output change is large. Therefore, a pixel having a high possibility of a significant target can be obtained by extracting a pixel whose output has greatly changed.

  In addition, the target detection unit 7 may have a function of extracting pixels having a high possibility of a significant target by lowering the detection threshold in the vicinity of pixels having a high possibility of a significant target. For example, an aircraft flying in a formation is naturally imaged in a close region within the field of view. By this processing, it is possible to detect a plurality of significant targets having such similar behavior with high sensitivity. Also, for a single significant target, for example, when the target is detected by evaluating the size as described above, the size can be evaluated with high sensitivity. When the target detection unit 7 detects a plurality of infrared images, even if a significant target is detected using one infrared image, the detection threshold in the vicinity of other wavelength band images may be lowered. The same effect can be obtained.

Embodiment 5 FIG.
FIG. 6 is a block diagram showing Embodiment 5 of the present invention. In FIG. 6, reference numerals 1 to 4 are the same as those in the first embodiment. Reference numeral 8 denotes an environment measurement unit, which has a function of measuring environmental conditions for which target determination processing is performed.
The optimal threshold value used for the target determination process in the target determination unit 4 naturally varies depending on the environmental conditions. Therefore, the environmental measurement unit 8 measures an environmental condition that greatly changes the optimum threshold value, and the threshold value is changed according to the environmental condition obtained by the target determination unit 4. It is possible to suppress the decrease, and conversely improve the target determination performance according to the environmental conditions.

For example, the environment measuring unit 8 may have a function of measuring the imaging time by the infrared imaging unit 1. During the daytime, the output value of the distance calculation unit 3 with respect to background clutter such as clouds increases due to the reflection component of sunlight. Therefore, by measuring the shooting time and raising the threshold value during the daytime, erroneous determination caused by background clutter can be reduced.
Alternatively, the environment measuring unit 8 may have a function of measuring the atmospheric temperature. An increase in the atmospheric temperature also increases the temperature of clutter such as significant targets and clouds, resulting in output fluctuations. Therefore, for example, by normalizing the output according to the atmospheric temperature or increasing the threshold value as the atmospheric temperature rises, it is possible to correct this variation and suppress a decrease in target determination performance.

  The environment measurement unit 8 may have a function of measuring atmospheric humidity. Since an increase in atmospheric humidity works as a decrease in infrared transmittance, target output performance is affected as a decrease in output signal, a decrease in apparent temperature, or an increase in apparent distance. Therefore, for example, for a far target, by determining by lowering the threshold according to the atmospheric humidity, it is possible to correct the decrease in output and suppress the decrease in detection.

  The environment measurement unit 8 may have a function of measuring the solar direction. As described above, the sunlight reflection increases the output of the background clutter due to clouds or the like, and causes erroneous determination. Therefore, for example, by increasing the threshold value for pixels that are calculated to have a high probability of occurrence of sunlight reflection from the sun direction and the visual axis direction, the erroneous determination due to sunlight reflection is reduced. Can do.

  The environment measurement unit 8 may have a function of measuring the elevation angle of the visual axis. In general, the main background clutter factors above the horizon and the horizon are sky and clouds, and the main background clutter factors below are the sea surface, ground, and plants. That is, since the type of the background clutter varies depending on the condition of the visual axis, the determination accuracy is improved if the threshold value of the target determination process is changed accordingly. For example, the cause of misjudgment above the horizon / horizon is mainly the background clutter caused by clouds, but the output is small because the clouds are cold, and the detection performance is improved by lowering the threshold for target decision processing in this region. be able to.

  The environment measurement unit 8 may have a function of measuring the distance to the target. Since the target size on the image changes depending on the distance to the target, for example, when the configuration for evaluating the target size is used in the target detection unit 7 shown in the fourth embodiment, the target size is set. It is possible to improve detection accuracy by correcting. Alternatively, since the amount of attenuation of the output due to the atmosphere changes according to the distance, for example, for a far target, the target determination performance can be corrected by lowering the threshold and performing the target determination process. At this time, since the attenuation amount of the signal varies depending on the wavelength band, the accuracy can be further improved by individually correcting the influence of the outputs obtained by the mid-infrared imaging unit 1b and the far-infrared imaging unit 1a. The target distance can be measured by, for example, triangulation using a laser or laser propagation time. Alternatively, it can be estimated based on the size of the target on the image, and in the case of a target that can be regarded as performing constant velocity motion, it is also possible to estimate from the moving speed.

  When the target detection device is mounted on an aircraft, the environment measurement unit 8 may have a function of measuring the altitude of the aircraft. For background clutter such as clouds, sea surface, and land, the main cause of background clutter is known by altitude, for example, the sea surface is near 0 m altitude. Therefore, the factor of the background clutter at the altitude can be specified by measuring its own altitude, and the discrimination accuracy can be improved by correcting the threshold according to the clutter. For example, when the altitude of the aircraft exceeds 3000 m above the background, the background clutter can be regarded as a substantially cold cloud, and as described above, the detection performance can be improved by lowering the threshold value of the target determination process at this altitude. Or since atmospheric temperature and humidity can be estimated from an altitude, the fall of the detection performance by offset and humidity by a temperature change can be suppressed as mentioned above. In addition, by providing the above-described function of measuring the elevation angle of the visual axis, it is possible to specify the background clutter with higher accuracy.

  The environment measuring unit 8 may have a function of measuring the target altitude. The target temperature is substantially equal to the ambient atmospheric temperature when the target itself is not generating heat or cooling, and the temperature can be estimated from the moving speed in the case of aerodynamic heating even when the target is generating heat. As described above, since the atmospheric temperature can be estimated from the altitude, the target temperature can also be estimated. Since an optimum threshold value can be set in accordance with the target temperature, for example, for a high temperature target, the output also increases. Therefore, by setting the threshold value high, erroneous determination can be reduced. The target altitude can be calculated by measuring, for example, the own altitude, the visual axis elevation angle, and the distance to the target.

  The environment measurement unit 8 may have a function of measuring the latitude and longitude of the own device. Atmospheric temperature and humidity vary with climate. By measuring the latitude and longitude of the aircraft, the position on the earth can be identified and the climate of the land can be known. Therefore, as described above, fluctuations in offset components due to temperature changes and deterioration in detection performance due to humidity can be suppressed. In order to measure latitude and longitude, for example, GPS (Global Positioning System) may be used. In addition, by providing map information, it is possible to identify the cause of background clutter because it is possible to determine whether the ground surface is the sea surface / land or its boundary, and to reduce misjudgment by correcting the threshold according to the background clutter. be able to. In addition, by providing a function to measure the visual axis direction and map information for a fixed heat source such as a power plant, it is possible to increase the accuracy by confirming the output characteristics for known significant targets, or increase the speed by omitting detection. be able to. Alternatively, since a known target having a low threat level can be excluded from a significant target, for example, this system has a practical effect in a system used for the purpose of extracting a target having a high threat level.

  The environment measurement unit 8 may have a function of measuring the horizontal direction of the acquired image. Many of the causes of background clutter, such as clouds and water, take a shape that extends horizontally. Therefore, for example, as shown in the fifth embodiment, the target detection unit 7 is provided, and the target detection unit 7 removes the target extending in the horizontal direction, thereby reducing erroneous determination. For the measurement in the horizontal direction, for example, an inclinometer or GPS may be used. The horizontal direction can also be obtained by identifying a horizontal line, a horizon line, or the like from the acquired image. Alternatively, by measuring the polarization state of the water surface using a polarizer, the p-wave component of the reflected light can be minimized in the horizontal direction.

Embodiment 6 FIG.
FIG. 7 is a block diagram showing Embodiment 6 of the present invention. In FIG. 7, reference numerals 1 to 4 are the same as those of the first embodiment. A data recording unit 9 has a function of holding determination data after the target determination unit 4 performs the target determination process.

  Since the past determination data includes the data of the significant target and the background clutter currently being determined, for example, the horizontal axis represents the output obtained by the far-infrared imaging unit 1a by analyzing the past data. The distribution of significant target / background clutter can be extracted on a two-dimensional graph with the output obtained by the mid-infrared imaging unit 1b as the vertical axis. Then, the target determination accuracy can be improved by resetting a threshold value for separating both based on the distribution. For example, since the distribution of an unknown significant target is also unknown, a target having characteristics different from those of a normal background clutter is extracted as a significant target. That is, the threshold for target determination is determined only by the background clutter distribution. However, if there is past determination data, the distribution of an unknown significant target can also be known. Therefore, it is possible to improve the target determination performance by setting an optimum threshold for the distribution of the significant target.

  Alternatively, a function may be provided in which a target is determined by lowering a threshold value around a pixel from which a target has been extracted in the previous determination based on past determination data. Since the target extracted last time is naturally likely to appear near this time as well, the target can be extracted with high sensitivity by lowering the threshold value. Since the area where the threshold value is lowered is limited to the vicinity of the area where the previous target is recognized, an increase in erroneous determination due to the lowered threshold value can be suppressed. In addition, a function may be provided that re-determines the target by going back past determination data and lowering the threshold with respect to the peripheral pixels of the pixel from which the target is extracted this time. This function makes it possible to perform target determination with high sensitivity and increase past data of significant targets. The increase in the past data of the significant target gives information on the target characteristic distribution, and for example, the position that appears at the next determination can be predicted from the position change, especially when this target detection device is used in a tracking system. It is valid. Alternatively, when a significant target is not extracted in the target peripheral pixels in the past, the target is likely to be erroneously determined, and the reliability of the determination can be grasped. Limiting the target redetermination region to the periphery of the target is effective in suppressing an increase in erroneous determination as described above, and is effective in shortening the processing time.

Other Embodiments
In the above embodiment, the determination data in the data storage unit 9 is used for the processing in the target determination unit 4. However, the target detection device is configured as shown in the block diagram of FIG. May be used. For example, when the Mahalanobis general distance is calculated by the distance calculation unit 3 as shown in the first embodiment, the target determination accuracy improves as the covariance matrix Σ represents the background clutter distribution with higher accuracy. . Since the accuracy of the covariance matrix Σ increases as the number of data increases, the target determination accuracy can be improved by increasing the number of background clutter data using past determination data.

Other Embodiment 2
Alternatively, the target detection device may be configured as a block diagram including the target detection unit 7 of the fourth embodiment shown in FIG. 9, and the determination data in the data storage unit 9 may be used in the calculation in the target detection unit 7. As described above, the past determination data includes data of the significant target currently being determined and the background clutter. For example, according to the information obtained from the data storage unit 9, the target detection unit 7 selects only a target having a spatial size, output size, and spectral characteristics equivalent to a pixel determined as a significant target in the past. Thus, a significant target can be extracted efficiently.

  Note that it is not necessary to store all the data in the data storage unit 9, and it is only necessary to store only data necessary for comparison with the current data. For example, if it is used to increase the accuracy of the distribution of significant targets, it is only necessary to store the pixel outputs determined as significant targets, and if there is no need for output due to defective pixels or noise, these There is no need to save the data. Further, it is not necessary to store the image output itself. For example, when the distribution is stored, the average value and variance, the number of data, and the like may be stored. By saving the processed data in this way, the amount of data to be saved can be reduced, the storage capacity can be reduced, the number of storage can be increased with the same capacity, and the access time to the data can be increased. There are advantages such as shortening.

Embodiment 7 FIG.
FIG. 10 is a block diagram showing Embodiment 7 of the present invention. In FIG. 10, reference numerals 2 to 4 are the same as those of the first embodiment. Reference numeral 10 denotes an imaging unit having the same field of view, which includes a mid-infrared imaging unit 1b, a far-infrared imaging unit 1a, and a visible region imaging unit 1e.
In this embodiment, by including the visible region imaging unit 1e in addition to the mid-infrared imaging unit 1b and the far infrared imaging unit 1a, it is possible to improve the determination accuracy by adding color information to the significant target determination conditions.
Also, instead of the imaging unit 1e in the visible region, it is also possible to subdivide the wavelength band of the mid-infrared imaging unit 1b and divide it into a plurality of imaging units to acquire detailed spectral characteristics and add them to the significant target determination conditions. This is effective in improving the determination accuracy. Of course, subdividing the far-infrared imaging unit 1a is also effective.
Further, in addition to the mid-infrared imaging unit 1b, the far-infrared imaging unit 1a, and the visible region imaging unit 1e, the wavelength band of the mid-infrared imaging unit 1b is subdivided and divided into a plurality of imaging units. Similarly, the far-infrared imaging unit 1a is subdivided. It is effective to improve the accuracy of target determination by acquiring detailed spectral characteristics and adding them to the determination conditions for significant targets.

  Moreover, you may use multiple wavelength bands other than the said mid-infrared or far-infrared as an imaging part. For example, a near-infrared imaging unit in the 1-2 μm band is more sensitive to high temperature changes such as several hundred to several thousand degrees than a far-infrared imaging unit in the 8-12 μm band. Therefore, it is possible to accurately determine a high-temperature target by using a near-infrared imaging unit in the 1-2 μm band instead of the far-infrared imaging unit in the 8-12 μm band for each of the above embodiments. It becomes.

Embodiment 8 FIG.
The present embodiment provides a target detection apparatus that performs threshold processing in which a noise component is included in the output of the imaging unit 1 and the detection performance does not deteriorate even if the target distribution spreads.
The configuration of this embodiment is the same as the block diagram shown in FIG. FIG. 16 is a characteristic diagram in which the threshold processing of the target determination unit 4 is graphed. In FIG. 16, the horizontal axis of the graph is the Mahalanobis general distance obtained by the distance calculator 3, and the vertical axis is the angle obtained by the angle calculator 2.

  Since the output of the imaging unit 1 includes a component due to noise, the output is not constant even if the same target is imaged. Also, when the environmental conditions such as temperature change change, the target signal changes in the same manner, so that the output fluctuates. The spread of the target distribution increases as the Mahalanobis general distance decreases. FIG. 17 is a characteristic diagram in which the output of the imaging unit 1 with respect to a signal from a significant target including noise or the like is graphed. The signal from the target spreads as described above due to noise or the like, but since the amount and sensitivity of noise of the two imaging units 1a and 1b are generally different, the distribution becomes an elliptical shape having different sizes on the vertical and horizontal axes. The angle at which this distribution is expected from the origin is the distribution Δθ in the vertical axis direction in FIG. 16, and Δθ becomes smaller in inverse proportion to the distance from the origin. Since the Mahalanobis general distance is roughly proportional to the distance from the origin, assuming that the Mahalanobis general distance is D, a relationship of Δθ = 2G / D is obtained using the constant G. If the target angle when there is no noise or the like, that is, the true target angle is θo, the significant target distribution in FIG. 16 is θo−G / D ≦ θ ≦ θo + G / D. By setting the threshold processing so as to be included in the determination region, it is possible to obtain a target detection device in which the detection performance does not deteriorate even if the target distribution spreads due to noise or the like.

When there are multiple targets, by setting a region where the angular distribution becomes smaller in inverse proportion to the Mahalanobis general distance D described above for each target, detection is possible even if the target distribution is widened due to noise or the like. It is possible to obtain a target detection device whose performance does not deteriorate. At this time, when the target distribution is located at a unit distance from the origin, the value of G for the target distribution is proportional to the angle at which the target distribution is expected from the origin. That is, the value of G is larger as the prospective angle is larger than the target distribution at the same distance. Assuming that the horizontal axis output is a1 and the vertical axis output is a2 when the signal output of noise generated in the two imaging units is plotted in the orthogonal coordinate system, the expected angle of the target distribution in unit distance is Roughly √ ((a1 * sin θo) ² + (a2 * cos θo) ² ). Therefore, since a relationship of G / √ ((a1 * sin θo) ² + (a2 * cos θo) ² ) = constant holds for each target distribution, this relationship causes another target distribution from the value of G of one target distribution. The value of G can be easily calculated.

In the present embodiment, it is used as a threshold value for determination processing in the target determination unit 4 for a region of a significant target distribution including noise or the like, but may be used as a threshold value for determining target detection performance. As described above, the signal output from the target including noise has an elliptic distribution as shown in FIG. For this reason, the graph also shows the threshold processing of the target determination unit 4 in which the horizontal axis of the graph is the Mahalanobis general distance obtained by the distance calculation unit 3 and the vertical axis is the angle obtained by the angle calculation unit 2. It has an elliptical distribution (FIG. 18). Therefore, by using the edge (periphery) on the side where the Mahalanobis generalized distance is smaller than the center of this elliptic distribution as a threshold value, for example, a target at the far limit distance of the detection ability, noise, etc. This makes it possible to determine a target without causing a decrease in detection performance due to the influence of. With respect to the target distribution of the ellipse, this threshold processing is D ≧ Do−H√ (1− (Do (θ−θo)) ² / J ² ) using constants H, J, and Do. Do is equal to the Mahalanobis general distance when no distribution due to noise or the like occurs with respect to a target condition that determines the limit of detection performance. Alternatively, the calculation may be simplified by approximating the threshold processing to D ≧ Do−H (1− (Do (θ−θo)) ² / J ² ) / 2.

  This process of using a significant target distribution including noise or the like as a threshold for determining the target detection performance may be used alone or in combination with the above-described threshold of the determination process for the target distribution area. Good (FIG. 19). At this time, the constant J is proportional to the angle at which the target distribution is expected from the origin of the characteristic diagram in which the output of the imaging unit is graphed, and is substantially equal to the constant G. Therefore, by using the value of the constant G as the value of the constant J, the calculation of the threshold value can be simplified.

Embodiment 9 FIG.
The present embodiment provides a target detection apparatus in which the target determination unit 4 is less likely to reduce target detection performance and efficiently reduce false alarms.
The configuration of this embodiment is the same as the block diagram shown in FIG. FIG. 20 is a characteristic diagram in which the threshold processing of the target determination unit 4 is graphed. In FIG. 20, the horizontal axis of the graph is the Mahalanobis general distance obtained by the distance calculation unit 3, and the vertical axis is the angle obtained by the angle calculation unit 2.
As described above, the background clutter output is suppressed by using the Mahalanobis pan-distance D. However, since an error output generated due to noise or a change in environmental conditions appears, this causes a false alarm. In order to reduce the false alarm efficiently while reducing the degradation of the target detection performance, it is necessary to perform threshold processing in accordance with the characteristic of the false alarm. FIG. 21 is a characteristic diagram in which the output of the imaging unit 1 with respect to a signal generated due to noise or the like is graphed. Since the noise amounts and sensitivities of the two imaging units 1a and 1b are generally different, the signal distribution due to noise or the like has an elliptical shape with different sizes on the vertical and horizontal axes. As a result, the horizontal axis used for threshold processing is as shown in FIG. 20 on the graph with the Mahalanobis general distance D and the vertical axis the angle θ. The shape of the distribution of noise or the like in FIG. 20 is expressed as D = K√ ((L * cos θ) ² + (M * sin θ) ² ) using constants K, L, and M. By performing the process, the false alarm can be efficiently reduced.

  The constants L and M can be determined by analyzing data acquired by the system, but this parameter is mainly determined by the noise of the system, so it may be determined using its design value or measurement value. This makes it possible to determine parameters even before the target exploration device is in operation and before acquiring data, and when determining parameters for another system, it is only necessary to convert them using design values. Will be able to. At this time, L / M = a1 / a2 is satisfied.

  This threshold process may be performed alone or may be set in combination with the process of the eighth embodiment. When combined with the eighth embodiment, even if the significant target distribution has a spread of noise or the like, the detection performance does not deteriorate, and a process for efficiently increasing the detection performance can be performed.

  The present invention is applicable to a target detection device installed in a moving body such as an aircraft or a fixed facility, so that it can cope with an unknown target that cannot be detected by a conventional device. Can be expanded.

It is a block diagram which shows Embodiment 1 of this invention. It is the characteristic view which graphed the output of the imaging part. It is a block diagram which shows Embodiment 2 of this invention. It is a block diagram which shows Embodiment 3 of this invention. It is a block diagram which shows Embodiment 4 of this invention. It is a block diagram which shows Embodiment 5 of this invention. It is a block diagram which shows Embodiment 6 of this invention. FIG. 38 is a block diagram illustrating another example in the sixth embodiment. FIG. 38 is a block diagram showing another example of the sixth embodiment. It is a block diagram which shows Embodiment 6 of this invention. It is a block diagram of the target detection apparatus by a prior art. It is a brightness | luminance characteristic view by the far infrared rays and middle far infrared rays of an object. It is a block diagram of the target detection apparatus by another prior art. It is an output distribution characteristic figure after filter application in other conventional technology. It is the characteristic view which displayed the kind of object and the state of the object on the coordinate. FIG. 10 is a characteristic diagram in which threshold processing of a target determination unit according to Embodiment 8 is graphed. FIG. 10 is a characteristic diagram in which the output of the imaging unit with respect to a signal from a significant target including noise according to the eighth embodiment is graphed. It is the characteristic view which graphed the threshold value processing result of the target determination part. FIG. 19 is a characteristic diagram in which a threshold processing result obtained by combining the threshold processing of FIG. 16 and FIG. 18 is graphed. FIG. 20 is a characteristic diagram in which threshold processing of a target determination unit according to Embodiment 9 is graphed. FIG. 10 is a characteristic diagram in which an output of an imaging unit with respect to a signal generated by noise or the like according to Embodiment 9 is graphed.

Explanation of symbols

1: infrared imaging unit, 1a: far infrared imaging unit, 1b: middle infrared imaging unit, 1e: visible region imaging unit, 2: angle calculation unit, 3: distance calculation unit, 4: target determination unit, 5: display unit, 6: filter unit, 7: target detection unit, 8: environment measurement unit, 9: data recording unit, 10: imaging unit.

Claims (22)

Two imaging units that obtain subject images of the same field of view in different wavelength bands;
A coordinate point when the output obtained by the two imaging units is plotted in an orthogonal coordinate system having a vertical axis and a horizontal axis, respectively, and a distance calculation unit that calculates a distance from the origin of the coordinate system to the coordinate point;
An angle calculation unit that calculates an angle formed by a horizontal axis and a straight line connecting the coordinate point and the origin;
A target detection apparatus comprising: a target determination unit that extracts a significant target by performing threshold processing on values obtained by the two calculation units.   The target detection apparatus according to claim 1, wherein each of the two imaging units has a far-infrared wavelength band and a mid-infrared wavelength band.   The target detection apparatus according to claim 1, further comprising a display unit that displays a significant target extracted after the target determination unit.   The target detection apparatus according to claim 3, wherein the display unit has a function of displaying a type or state of a significant target.   5. The target detection apparatus according to claim 1, further comprising a filter unit that corrects or removes a change in signal amount due to an environmental condition before the distance calculation unit and the angle calculation unit.   The target detection apparatus according to claim 5, wherein the filter unit is a high-pass filter that extracts only a spatial high-frequency component.   The target detection apparatus according to claim 1, wherein the distance calculation unit outputs a Mahalanobis general distance calculated using a covariance matrix with respect to a pixel output distribution around the target.   A data storage unit for storing data after processing by the target determination unit is provided, and past data stored in the data storage unit is referred to when the distance calculation unit calculates the distance. The target detection apparatus according to claim 7.   9. The apparatus according to claim 1, further comprising a target detection unit that extracts a pixel having a high target possibility from the output of the imaging unit and outputs the output of the pixel to a distance calculation unit and an angle calculation unit. The target detection apparatus according to any one of the above.   A data storage unit for storing data after processing of the target determination unit is provided, and the target detection unit refers to past data stored in the data storage unit for extraction processing of a pixel having a high target possibility. The target detection device according to claim 9.   The target according to any one of claims 1 to 10, further comprising an environment measuring unit that measures environmental conditions around the apparatus and the target, wherein the target determining unit adjusts the threshold according to a measurement result of the environment measuring unit. Detecting device.   The target detection apparatus according to claim 11, wherein the environment measurement unit has a function of measuring a direction of the sun with respect to a field of view of the imaging unit.   In the target determination unit, when the sensitivity of the mid-infrared imaging unit is E3 and the sensitivity of the far-infrared imaging unit is E8, tan θ = F * E3 / E8 (where F is 0.5 <F <2.0) The target detection device according to claim 1, wherein different threshold values are set with a value of θ that satisfies the condition as a boundary.   The target determination unit includes a function of performing a determination process using an output result of a distance calculation unit obtained from only one of the outputs obtained by the two imaging units. The target detection apparatus according to any one of 1 to 13.   A data storage unit for storing data after processing by the target determination unit is provided, and the target determination unit refers to past data stored in the data storage unit when extracting a significant target. The target detection apparatus according to claim 1.   The target detection apparatus according to claim 1, further comprising an imaging unit for a wavelength band other than the two imaging units for different wavelength bands.   In the target determination unit, when the true target angle obtained from the angle θ obtained by the angle calculation unit is θo and the Mahalanobis general distance obtained by the distance calculation unit is D, θo−G using a constant G 8. The target detection apparatus according to claim 7, wherein a threshold value is set so that / D ≦ θ ≦ θo + G / D is a significant target region. In the target determination unit, when there are a plurality of target distributions, the horizontal axis output is a1 and the vertical axis output when plotted in the orthogonal coordinate system with respect to the noise signal output generated by the two imaging units. 18. A target according to claim 17, wherein a2 is set to a2, and a constant G is set so that G / √ ((a1 * sin θo) ² + (a2 * cos θo) ² ) = constant for each target distribution. Detecting device. In the target determining unit, the true angle of the target obtained from the angle θ obtained by the angle calculating unit is θo, the Mahalanobis general distance obtained by the distance calculating unit is D, and the Mahalanobis generalized when no target distribution is generated. The threshold is set so that D ≧ Do−H√ (1− (Do (θ−θo)) ² / J ² ) is a significant target area using constants H and J when the distance is Do. The target detection apparatus according to claim 7, 17 or 18.   The target detection apparatus according to claim 19, wherein the target determination unit satisfies J = G. In the target determination unit, when the Mahalanobis general distance obtained by the distance calculation unit is D, D = K√ ((L * cos θ) ² + (M * sin θ) ² using constants K, L, and M. 21. The target detection apparatus according to claim 7, wherein a threshold value is set so as to be a significant target region.   The target detection apparatus according to claim 21, wherein the target determination unit satisfies L / M = a1 / a2.

Images