WO2011024865A1

WO2011024865A1 - Image target identification device, image target identification method, and image target identification program

Info

Publication number: WO2011024865A1
Application number: PCT/JP2010/064403
Authority: WO
Inventors: 岳夫野崎
Original assignee: 日本電気株式会社
Priority date: 2009-08-27
Filing date: 2010-08-25
Publication date: 2011-03-03
Also published as: JPWO2011024865A1; JP5664551B2

Abstract

In order to reliably identify a target object in a captured infrared image, a luminance histogram calculation unit (6) which generates a luminance histogram indicating the appearance frequencies of luminance values of pixels included in the infrared image, a luminance shift calculation unit (7) which sets the luminance value at which the appearance frequency is maximum in the luminance histogram to the intermediate value of the luminance range width of the infrared image, and generates a luminance-shifted image obtained by linearly shifting the respective luminance values of the infrared image on the basis of the intermediate value, a reverse image processing unit (8) which generates a reverse shifted image obtained by reversing luminance levels of the luminance- shifted image, and a luminance calculation processing unit (9) which generates a calculation-processed image by performing calculation processing based on the difference between luminance values at respective corresponding positions of the luminance-shifted image and reverse shifted image are provided.

Description

Image target identification device, image target identification method, and image target identification program

The present invention relates to an image processing apparatus for detecting a target such as a ship using an infrared image.

In maritime search and rescue by aircraft etc., pilots search for people floating in the waves and rescue boats using radio signals by beacons such as rescue signals, visual search while flying in the search sea area in the low sky, Perform searches based on images from infrared cameras.
Especially at night and in bad weather, the visibility and visibility are worse than when searching for clear sky and daytime, so it is possible to discover the target objects such as the victims floating on the sea surface, people on ships and ships, etc. with the naked eye. Or shooting as an image that can be detected by an infrared camera becomes difficult.
As a result, there is a risk that in the situation related to lifesaving, it takes time to discover the person, so the probability of survival becomes low, or the invasion of ships and suspicious persons of other countries due to delay in discovery of the search target There is an inconvenience that increases.

On the other hand, by performing contrast improvement in the captured infrared image, the searcher generates a field of view support information (image) for identifying and finding the target object at an early stage. Image information processing technology, image identification support technology, and the like that are effectively supported have been developed.

For example, an infrared camera that captures an infrared image by an uncooled infrared sensor that is downsized by omitting a cooling function and has a significantly improved S / N ratio with respect to noise, an image processing system using the infrared camera, and the like are known. ing.

In addition, as a related technology, a miniature uncooled infrared camera is installed on an aircraft or flying object, an infrared image is taken, the obtained infrared image is processed, and a target object is extracted and identified from the infrared image. An infrared image processing system for display is disclosed (Patent Document 1).

In Patent Document 1, infrared light incident from the target and the background of the target is separated into plane polarized light orthogonal to each other, and the infrared light is received by an image sensor in which two light receiving elements are two-dimensionally arranged, and converted into an electric signal amount. Then, a polarization difference image signal obtained by taking the difference of the electric signal amount at the same coordinate position of the two-dimensional coordinates among the output image signals is output.

In addition, a polarization image pickup device that outputs a luminance image signal obtained by adding electric signal amounts at the same coordinate position of two-dimensional coordinates, a combination of a target and a background, a target and its background, Is a binarized image signal obtained by binarizing the electric signal amount of the polarization difference image signal based on a comparison with the threshold value in combination with a threshold database that stores the threshold value of the polarization difference image signal that can be separated in advance. Is output.
Further, a target candidate is extracted from an image obtained by multiplying the binarized value of the binarized image signal and the electric signal amount of the luminance image signal at the same coordinate position of the two-dimensional coordinates. Yes.

In the invention described in Patent Document 1, when extracting targets such as floating objects floating on the sea surface, ships, flying objects in the airspace, and obstacles with low visibility in the low airspace, generally the target object and its background Is equipped with an optical system consisting of surface deflection light receiving elements that separate P waves and S waves in front of the infrared sensor, and an appropriate brightness calculation processing unit for captured images. Infrared image processing is performed to extract a target object on the assumption that a location above a predetermined temperature (luminance) is the target object.
Furthermore, as this related technique, infrared rays from various directions are collected using a dome-shaped lens composed of a large number of lenses (Patent Document 2), and the captured image is binarized based on the temporal change in luminance of the captured image. And a system for detecting a target moving from a background of a complicated luminance distribution in an infrared image is disclosed (Patent Document 3).
In addition, an image device for improving the visibility of a target object in an infrared image taken when fog occurs by combining an infrared image and a radar image is disclosed (Patent Document 4).

JP 2001-056860 A Japanese Patent Laid-Open No. 06-242304 Japanese Patent Laid-Open No. 06-303485 JP-A-10-62531

However, the invention described in Patent Document 1 has a disadvantage that it is difficult to reduce the size because the optical system for separating the P wave and the S wave before the infrared sensor has a large-scale configuration.
In addition, when the background is partially heated by sunlight, etc., when the temperature difference between the target and the background becomes minute, or when there is a lot of noise due to splashing on the water surface, the infrared sensor is used. Since the captured infrared image has a lot of background noise, the contrast between the target object and the background cannot be obtained depending on the optical system, making it difficult to correctly recognize and capture the target object. There were inconveniences such as erroneous recognition of (target) and oversight.

In addition, in the inventions described in Patent Documents 2 to 4 described above, and the contents obtained by combining these, for example, white waves are generated due to bad weather on the sea in the imaging region, or artificial wave splash is generated due to helicopter hovering. If the target object appearing in the infrared captured image is buried in white waves and the shape of the target appears to change, the target object is identified by image identification based on this infrared captured image. In addition, there may be inconveniences such as missing a target or selecting (screening) an incorrect target object.

[Object of invention]
The present invention provides an image target identification device, an image target extraction method, and an image target extraction program that can improve the inconvenience of the related technology and can reliably identify a target object in a captured infrared image. For that purpose.

In order to achieve the above object, an image target identification device according to the present invention condenses infrared light from a photographing object including a target object and acquires a specific wavelength band from the condensed infrared light. An optical unit that performs, an uncooled infrared image sensor that generates an infrared image corresponding to the temperature of the imaging target based on the wavelength acquired by the optical unit, and a calculation in which the target object is highlighted based on the infrared image An image target identification device including an image processing unit that generates a processed image, wherein the image processing unit generates a luminance histogram indicating a frequency of appearance of luminance values of pixels included in the infrared image And the luminance value having the highest appearance frequency in the luminance histogram is set to an intermediate value of the luminance range width in the infrared image, and the previous value based on the intermediate value A luminance level adjustment unit that generates a luminance shift image in which each luminance value of the infrared image is linearly shifted; and an inverted shift image that inverts the luminance level of the luminance shift image, and the luminance shift image and the inverted shift image And a luminance calculation processing unit that performs calculation processing based on a difference in luminance value at each corresponding position and generates the calculation processing image.

In addition, the image target identification method according to the present invention condenses infrared light from an imaging target including a target object and generates an infrared image based on the temperature of the imaging target from the condensed infrared light. An image target identification method for generating an arithmetic processing image in which the target object is highlighted on the basis of a brightness histogram indicating a frequency of appearance of brightness values of pixels included in the infrared image, and the brightness histogram A luminance value that maximizes the appearance frequency in the infrared image is set to an intermediate value of the luminance range width in the infrared image, and a luminance shift image is generated by linearly shifting each luminance value of the infrared image based on the intermediate value, A reverse shift image in which the brightness level of the brightness shift image is inverted is generated, and brightness values at corresponding positions of the brightness shift image and the reverse shift image are generated. It is characterized by generating the processing image by performing arithmetic processing based on the difference.

In addition, the image target identification program according to the present invention condenses infrared light from a photographing object including a target object and generates an infrared image based on the temperature of the imaging object from the condensed infrared light. A luminance histogram for generating a luminance histogram indicating a frequency of appearance of luminance values of pixels included in the infrared image, the image target identification program for generating an arithmetic processing image in which the target object is highlighted based on An arithmetic function and a luminance value with the highest appearance frequency in the luminance histogram are set to an intermediate value of the luminance range width in the infrared image, and each luminance value of the infrared image is linearly shifted based on the intermediate value Luminance level adjustment function for generating a luminance shift image and inversion for generating an inverted shift image obtained by inverting the luminance level of the luminance shift image And a brightness calculation processing function for generating the calculation processing image by performing calculation processing based on a difference in luminance values at corresponding positions of the luminance shift image and the inverted shift image. It is characterized by.

Since the present invention is configured and functions as described above, according to this, the luminance value having the maximum appearance frequency is set to the intermediate value of the display gradation number from the luminance histogram indicating the appearance frequency of the luminance value of the pixel of the infrared image. The brightness value difference between the means for generating a brightness shifted image that is set and linearly shifting other brightness values based on this intermediate value, and the inverted shift image obtained by inverting the brightness level of this image and the brightness shifted image Since the calculation processing image is generated by performing the calculation processing based on the target object, the target object in the captured infrared image is highlighted, and the calculation processing image that can reliably identify the target object is generated. An image target identification device, an image target identification method, and an image target identification program can be provided.

It is a schematic block diagram which shows one Embodiment in the image target identification device by this invention. It is a schematic block diagram which shows an example of an internal structure of the image process part in the image target identification device disclosed in FIG. FIG. 3A is a schematic explanatory diagram of one frame data of an infrared image in the image target identification device disclosed in FIG. 1, and FIG. 3B is a diagram showing a luminance shift image of the infrared image and its inverted shift image. FIG. 3C is an explanatory diagram showing a luminance profile, and FIG. 3C is an explanatory diagram showing a luminance profile of a processed image that is a result of the subtraction process for the image shown in FIG. It is explanatory drawing which shows the brightness | luminance histogram of 1 frame data of the infrared image in the image target identification device disclosed in FIG. 3 is a luminance histogram of a luminance shift image obtained by linearly shifting the luminance of each pixel of an infrared image in the image target identification device disclosed in FIG. 1. FIG. 6A is a schematic explanatory diagram of one frame data of an infrared image shown by the image target identification device disclosed in FIG. 1, and FIG. 6B is a luminance shift image of the infrared image and its inverted shift image. FIG. 6C is an explanatory diagram showing a brightness profile of a processed image that is a result of division processing on an infrared image. It is explanatory drawing which shows an example of the infrared image which imaged the sea surface suspended | floating matter which concerns on the image target identification device disclosed in FIG. It is explanatory drawing which shows an example of the brightness | luminance histogram of the infrared image disclosed in FIG. It is explanatory drawing which shows an example of the process image obtained by dividing the infrared image of FIG. 7 with the image target identification apparatus disclosed in FIG. It is explanatory drawing which shows an example of the brightness | luminance histogram of the infrared image disclosed in FIG. It is explanatory drawing which shows an example of the infrared image which imaged the ship which concerns on the image target identification device disclosed in FIG. It is explanatory drawing which shows an example of the process image obtained by carrying out the subtraction process of the infrared image disclosed in FIG. 11 by the image target identification apparatus disclosed in FIG.

[Embodiment]
Next, the basic configuration content of the embodiment of the present invention will be described.

As shown in FIG. 1, an image identification support device (image target identification device) 100 according to the present embodiment includes a far-infrared light collecting dome 1 (hereinafter referred to as “light collecting dome”) for collecting infrared light, The optical unit 2 that receives the emitted infrared light through a preset condenser lens, and the heat corresponding to the temperature of the imaging object based on the wavelength (band) obtained through the optical unit 2 An image processing is performed on an uncooled infrared image sensor unit (uncooled infrared sensor unit) 3 that generates an image (infrared image) and an infrared image generated by the uncooled infrared sensor unit 3, It has a configuration including an image processing unit 4 that generates a processed image in which a target object such as a ship is highlighted on a background region such as the sky, and an image display unit 5 that outputs and displays the generated processed image.

This will be described in detail below.
The condensing dome 1 that condenses the infrared light includes a condensing lens that guides the condensed infrared light to the uncooled infrared sensor unit 3 of the optical unit 2.
The optical unit 2 includes an optical filter that detects a desired wavelength band set in advance.

The uncooled infrared sensor unit 3 generates an infrared image indicating the amount of heat corresponding to the infrared rays that have passed through the optical filter of the optical unit 2. Here, it is assumed that the generated infrared image is processed in the image processing unit 4 for each frame included in the infrared image.

As shown in FIG. 2, the image processing unit 4 generates a luminance histogram of each frame of the infrared image, and determines a luminance level (mode frequency level) at which the appearance frequency (appearance frequency value) is maximum based on the luminance histogram. The appearance frequency value corresponding to each luminance level in each frame of the infrared image is linearly converted so that the histogram calculating unit (luminance histogram calculating unit) 6 to detect and the mode luminance level become an intermediate value of the number of display gradations. A luminance shift calculation unit 7 that generates a luminance shift image and an inverted image processing unit 8 that generates an inverted shift image obtained by inverting the luminance level of the luminance shift image are included.

Further, the image processing unit 4 divides (or subtracts) the luminance in the corresponding pixels based on the inverted shift image and the luminance shift image, and performs an operation based on an appropriate gain coefficient, thereby performing an intermediate processing image. A luminance calculation processing unit 9 that generates (subtraction processing image, division processing image), a feature detection unit 10 that performs appropriate filtering processing on specific luminance information in the intermediate processing image, and a target object and background in the intermediate processing image A contrast expansion unit 11 that performs a process for improving the contrast with the image, and a binarization processing unit 12 that performs binarization at a certain luminance level on the processed image and extracts a target object. .

The histogram calculation unit 6 has an original image receiving function for receiving an infrared image (frame data group) imaged from the uncooled infrared sensor unit 3 and each frame data in units of each frame (frame data) 6a of the infrared image. Has a luminance level detection function for detecting the luminance level of each pixel included in the.
The luminance level of each pixel included in the frame data indicates a luminance level proportional to the temperature of the heat source, detected for each pixel of the infrared sensor included in the uncooled infrared sensor unit 3.

In addition, the histogram calculation unit 6 correlates with the frame number of each captured frame, based on the image data relating to the captured image stored in the memory of the uncooled infrared sensor unit 3, and the luminance level in each frame A histogram generation function for generating a luminance histogram (FIG. 4) representing the relationship with the appearance frequency is provided.
Further, the histogram calculation unit 6 has a maximum frequency extraction function for extracting the mode luminance level 6b as the luminance level having the maximum appearance frequency (referred to as “mode”) based on the luminance histogram in each frame. .

The luminance shift calculation unit (luminance level adjustment unit) 7 sets the mode value of the mode luminance level extracted by the histogram calculation unit 6 to an intermediate value of the luminance range width that is the number of display gradations of each frame. The luminance shift image is obtained by performing a process of linearly shifting the frequency value of the other luminance level in the image (frame) while maintaining the contrast relationship with the mode of the mode luminance level (intermediate value). A luminance shift image generation function is provided.

For example, when the luminance level of the infrared sensor image (infrared image) is 256 gradations, the luminance shift calculation unit 7 shows the most frequent luminance level in the histogram (FIG. 4) of the generated frame data 6a. As described above, the intermediate value is set to 128, and the other luminance level in the frame data 6a is linearly shifted (converted) using this as a reference to generate the luminance shifted image 7a.
Thereby, in the brightness shift image 7a, the visibility of the target object is improved.
It is assumed that the brightness exceeding 255 is calculated as 255 and the brightness level at which the minimum brightness is 0 or less is 0.

In addition, the luminance shift calculation unit 7 duplicates the generated luminance shift image 7 a and stores it in a luminance shift image storage unit set in advance in the image processing unit 4, as well as the inverted image processing unit 8 and the luminance calculation processing unit. 9 each has a luminance shift image transfer function to transmit to each.

The inverted image processing unit 8 holds the luminance shift image 7a in the imaging memory and has an inverted image generation function for generating an inverted shift image in which the luminance level of the luminance shift image 7a is inverted (for example, black and white inversion). ing.

The luminance calculation processing unit 9 receives the inverted shift image 8a sent from the inverted image unit 8, and each of the average luminance values in the frames of the luminance shifted image 7a and the inverted shifted image 8a stored in the luminance shifted image storage unit, respectively. In addition to the calculation, among the luminance shift image 7a and the inverted shift image 8a, the image with the larger (higher) in-frame luminance average value is the larger luminance image, and the image with the smaller (lower) luminance average is the smaller luminance image. Have a luminance average value calculation setting function to be set respectively.

In addition, as shown in [Equation 1] below, the luminance calculation processing unit 9 calculates an image having a larger luminance average (large luminance) with respect to the luminance value of the position address corresponding to each of the luminance shifted image 7a and the inverted shifted image 8a. A luminance value division processing function for performing processing (division processing) to divide the luminance value of the image) by the luminance value of the image having the smaller average luminance (small luminance image).
Here, when performing the above-described division processing, the luminance calculation processing unit 9 ensures that the most frequent luminance level before the calculation and the intermediate value level that is the calculation (division) result are equal to each other as shown in Expression 1. The division result is multiplied by a preset gain K (for example, K = 128).

[Formula 1]
Id (x, y) = K * A (x, y) / B (x, y)

In addition, the luminance calculation processing unit 9 generates a division processing image 9a composed of Id (x, y) in which the contrast with respect to the background of the target object in the image is enhanced, and the division processing image 9a is used as the feature detection unit 10. Is sent to.

Further, the luminance calculation processing unit 9 uses the image with the larger average luminance (large luminance) as shown in the following [Equation 2] for the luminance value of the position address corresponding to each of the luminance shifted image 7a and the inverted shifted image 8a. It may be set to perform processing for subtracting the luminance value of the image (small luminance image) having the smaller average luminance from the luminance value of the image (subtraction processing).
When performing the subtraction process, the luminance calculation processing unit 9 equals the mode luminance level before calculation at each position address and the intermediate value level as the calculation (subtraction) result, as shown in Expression 2. It is assumed that a preset gain value (for example, K = 128) is added to the subtraction result.

[Formula 2]
Is (x, y) = | A (x, y) −B (x, y) | + K

Thereby, the luminance calculation processing unit 9 generates a subtraction processing image 9a ′ composed of Is (x, y) and in which the contrast of the target object in the image is emphasized, and the subtraction processing image 9a ′ is used as the feature detection unit. 10 is sent out.

It should be noted that the selection of whether to perform the division process or the subtraction process in the luminance calculation processing unit 9 preferably employs the calculation with the larger luminance difference with respect to the target target background, but the final result The setting may be selected by comparing the processed images obtained in the above.

The feature detection unit 10 applies a region (pixel) having a luminance frequency lower than a preset luminance frequency in the image to the division processing image 9a (or the subtraction processing image 9a ′) sent from the luminance calculation processing unit 9. A filtering process function for generating the filtered image 10a by performing the filtering process to be removed is provided. Thereby, the brightness | luminance of the mode value vicinity in an image (filtering image 10a) can be increased relatively.

The contrast extension unit 11 selects a region (referred to as “luminance region”) having a luminance value equal to or higher than a preset luminance value of the filtered image 10a, and based on the luminance mode value in the selected region, the vicinity of the mode value. Is provided with a contrast expansion processing function for generating a contrast expansion image 11a in which the contrast is expanded by linearly expanding the luminance range of the luminance region so that the maximum is.

The binarization processing unit 12 has a binarization processing function for generating a binarized image 12a by performing preset threshold processing on the luminance of each pixel of the contrast expansion image 11a.
Further, the binarization processing unit 12 may be configured to extract a target object from the binarized image 12a output to the image display unit 5 and to perform image highlight display of the target object.
As a method for the above-described image emphasis display, the binarization processing unit 12 is configured to display, for example, a mark set corresponding to the extracted target object so as to be superimposed on the target object in the binarized image 12a. Also good.

As described above, in the present embodiment, the luminance shift image in which the most frequent luminance level of the luminance distribution of the infrared image (original image) captured by the uncooled infrared sensor unit 3 is matched with the intermediate value of the luminance dynamic range, The inverted shift image is created, and the image obtained by dividing or subtracting the luminance level at the same pixel position of the two images is subjected to filtering processing and contrast expansion, whereby the original image is compared with the original image. Image processing that suppresses the luminance variation of the background that occupies most of the luminance distribution in the image and enlarges the luminance difference of the target object with respect to the background can be performed.

As a result, the image identification support device 100 improves the target object and background contrast of the image captured by the uncooled infrared sensor unit 3, and supports the identification of the target object (target object) in the image.
For this reason, the image identification support device 100 can be used, for example, to find a floating object between waves such as a person or a rescue boat for rescue of a marine accident, and the floating in an infrared image photographed by the uncooled infrared sensor unit 3. By effectively extracting target objects such as objects and ships, separating the target object and background noise such as white waves, and removing (suppressing) the background noise, the image contrast with respect to the background of the target object can be improved. .

[Description of Operation of Embodiment]
Next, an outline of the operation of this embodiment will be described.
First, the luminance histogram calculation unit 6 of the image processing unit 4 generates a luminance histogram indicating the appearance frequency of the luminance value (luminance level) of the pixels included in the infrared image (luminance histogram calculation step), and then the luminance shift calculation. The unit 7 sets the luminance value having the highest appearance frequency in the luminance histogram to an intermediate value of the number of display gradations in the infrared image, and maintains a contrast relationship with the frequency value of the luminance level set to the intermediate value. Meanwhile, a luminance shift image is generated by linearly shifting the frequency value of each other luminance level (luminance level adjustment step).

Next, the inverted image processing unit 8 generates an inverted shift image in which the luminance level of the luminance shift image is inverted (inverted shift image generating step), and the luminance calculation processing unit 9 corresponds to the corresponding positions of the luminance shift image and the inverted shift image. An arithmetic processing image is generated by performing arithmetic processing based on the difference in luminance values at (luminance arithmetic processing step), and the image target extraction unit 13 extracts a target object from the arithmetic processing image.

Here, with respect to the luminance histogram calculation step, the luminance level adjustment step, the inverted shift image generation step, and the luminance calculation processing step, the execution contents may be programmed and executed by a computer.

Next, a processing operation for extracting a target object by the image processing unit 4 of the apparatus 100 will be described with reference to FIG.

First, the image processing unit 4 receives an infrared image (frame data 6a) imaged from the uncooled infrared sensor unit 3, and the luminance histogram calculation unit (histogram calculation unit) 6 receives each frame (frame data) of the infrared image. ) The luminance level of each pixel included in 6a is detected in units of 6a.
The luminance level imaged by the non-cooling infrared sensor unit 3 is a luminance level proportional to the temperature of the heat source detected by the infrared sensor included in the non-cooling infrared sensor unit 3 in units of pixels.
Here, FIG. 3A is an explanatory diagram schematically showing one frame of the infrared image.

Next, the histogram calculation unit 6 generates a luminance histogram (FIG. 4) that represents the relationship between the luminance level and the appearance frequency in each frame in association with the frame number of each captured image.
Further, the histogram calculation unit 6 transfers the mode-level luminance level value 6b and the corresponding infrared image frame data 6a to the luminance shift calculation unit 7 based on each luminance histogram (step S1).

Next, the luminance shift calculation unit 7 uses the luminance level of the mode 6b of the luminance histogram corresponding to each frame data 6a of the infrared image as an intermediate value of the luminance display range (number of display gradations) in each frame data 6a. Based on this intermediate value, a process of linearly shifting the frequency values of other luminance levels is performed.

For example, when the luminance level of the infrared image is 256 gradations, the luminance shift calculation unit 7 indicates the luminance level of the mode value in the luminance histogram (FIG. 4) of the generated frame data 6a as shown in FIG. Then, the intermediate value is set to 128, and the frequency value of the other luminance level in the frame is linearly shifted (converted) using this as a reference to generate the luminance shifted image 7a.
It is assumed that the brightness exceeding 255 is calculated as 255 and the brightness level at which the minimum brightness is 0 or less is 0.

In addition, the luminance shift calculation unit 7 duplicates the generated luminance shift image 7 a and stores it in a luminance shift image storage unit set in advance in the image processing unit 4, as well as the inverted image processing unit 8 and the luminance calculation processing unit. 9 is transmitted to each of them (step S2).

Next, the inverted image unit 8 generates an inverted shift image 8a in which the luminance level of the luminance shift image 7a is inverted (for example, black and white inversion), and sends it to the luminance calculation processing unit 9 (step S3).

Next, the luminance calculation processing unit 9 receives the inverted shift image 8a sent from the inverted image unit 8, and averages the in-frame luminance of each of the luminance shifted image 7a and the inverted shifted image 8a stored in the luminance shifted image storage unit. Each value is calculated, and an image having a larger luminance average of both images (luminance shift image 7a and inverted shift image 8a) is set as A (x, y), and one of the images is set as B (x, y). . Note that (x, y) indicates the position address of the pixel in the frame.

Here, as shown in the following [Equation 1], the luminance calculation processing unit 9 determines the image with the larger luminance average (larger) for the luminance value of the position address corresponding to each of the luminance shifted image 7a and the inverted shifted image 8a. A process of dividing the luminance value of the luminance image) by the luminance value of the image having the smaller average luminance (small luminance image) is performed (division processing).
Further, when performing the division process, the luminance calculation processing unit 9 equals the most frequent luminance level before calculation at each position address and the intermediate value level that is the calculation (division) result as shown in Expression 1. In such a manner, a preset gain K (for example, K = 128) is multiplied to the calculation result.

[Formula 1]
Id (x, y) = K * A (x, y) / B (x, y)

As a result, the luminance calculation processing unit 9 generates a division processing image (9a) made of Id (x, y) in which the contrast with respect to the background of the target object in the image is emphasized, and the division processing image 9a is characterized. The data is sent to the detection unit 10 (step S4).

Note that the luminance calculation processing unit 9 replaces the division processing with respect to the luminance value of the position address corresponding to each of the luminance shift image 7a and the inverted shift image 8a, as shown in [Equation 2] below, It may be set to perform processing for subtracting the luminance value of the small luminance image from the luminance value (subtraction processing). In addition, when performing this subtraction process, the calculation result is set in advance so that the most frequent luminance level before calculation at each position address is equal to the intermediate value level that is the calculation (subtraction) result. A gain value (for example, K = 128) is added.

[Formula 2]
Is (x, y) = | A (x, y) −B (x, y) | + K

Thereby, the luminance calculation processing unit 9 is a subtraction consisting of Is (x, y), in which the contrast with respect to the background of the target object in the image (particularly, the color tone away from the intermediate value, for example, the background of white waves, etc.) is enhanced. A processed image (9a ′) is generated, and the subtracted processed image 9a ′ is sent to the feature detection unit 10 (step S4).

In the luminance calculation processing unit 9, the selection of whether to perform the above-described division processing or subtraction processing on the frame data may employ the calculation that increases the luminance difference with respect to the target target background. Although desirable, it may be a setting for selecting a process in which the contrast with respect to the background of the target object is further improved by comparing the processed images obtained in the final result.

Next, the luminance calculation processing unit 9 inputs the generated division processing image 9a or subtraction processing image 9a ′ to the feature detection unit 10.
Here, the luminance calculation processing unit 9 displays the generated division processing image 9a or subtraction processing image 9a ′ as a through image when there is a certain luminance difference between the background image and the target object. 5 may be set to be sent out.
In this case, the division processing image 9a or the subtraction processing image 9a ′ is displayed on the image display unit 5 as the arithmetic processing image.

Here, the content of the process of enhancing the contrast of the target object in the image by the luminance calculation processing unit 9 will be described. Here, a description will be given based on FIG. 3A and FIG. 6A schematically showing the luminance shift image 7a of the infrared image.

3B is a luminance profile in the X direction passing through the target object 3a in FIG. 3A, and 602 in FIG. 3B is an inversion obtained by inverting the luminance of the luminance shift image 7a. A luminance profile passing through the target object 3a in the shift image 8a is shown.
Further, similarly in FIG. 6B, reference numeral 801 in FIG. 6B denotes a luminance profile in the X direction passing through the target object 3a in FIG. 6A, and reference numeral 602 in FIG. The brightness | luminance profile which passes the target object 3a in the reverse shift image 8a which reversed the brightness | luminance of the shift image 7a is shown.

For example, when the luminance calculation processing unit 9 performs the subtraction process as described above, the background region 3b occupying the majority of the luminance frequencies in the images of the luminance shift image 7a and the inverted shift image 8a (FIG. 6A). ), The luminance difference between the luminance shift image 7a and the inverted shift image 8a is in the vicinity of 0 with an extremely small luminance level according to [Equation 2].

On the other hand, the luminance difference with respect to the background region in the vicinity of the target object in the processed image, which is the processing result of the subtraction process, is enlarged as shown in the luminance profile 701 in FIG. The luminance level of the background area in this processed image becomes a luminance level almost the same as the luminance level in the luminance shifted image 7a before the subtraction processing by adding gain K (for example, K = 128), but near the target object. Since the difference in luminance level from the background is enlarged, the contrast is improved (emphasized), and the visibility of the target object is greatly improved.

Here, the image which imaged the ship which navigates the sea surface in FIG. 11 by the non-cooling infrared sensor part 3 is shown. FIG. 12 shows a processed image when the subtraction process is performed on the infrared image of FIG. 11 as described above.
In particular, in FIG. 12, which is a processed image of the subtraction process, the contrast of the ship, which is the target (target target), with respect to the background of the brightness level near the values (0, 255) on both ends in the display gradation number is effectively enhanced. In addition, the visibility is significantly improved as compared with the suspended matter between waves in FIG.

Next, a case where the above-described division process is executed in the luminance calculation processing unit 9 will be described.
Here, in the background region 3b (FIG. 3A) that occupies the majority of the appearance frequency of the brightness level in the brightness shift image 7a and the inverted shift image 8a, the appearing brightness level is the brightness shift image 7a and the inversion. Since both of the shift images 8a are close to the intermediate value of 128 (when the number of display gradations is 256 gradations), when the luminance calculation processing unit 9 executes the luminance division processing for each pixel, the quotient that is the division result is 1. It is a nearby value.

For this reason, in this division processing, even in a region where the background has a large number of pixels (3b: FIG. 3A), the luminance difference is extremely small, and even when an operation of multiplying the gain K (for example, K = 128) is performed (formula 1) The luminance level does not change greatly, and this is noticeable in the luminance value close to the gradation intermediate value (128) or in the region where the luminance difference is small in the luminance shift image 7a and the inverted shift image 8a.

On the other hand, the quotient that is the result of the calculation (division) processing of the luminance value of the luminance shift image 7a near the target object and the luminance value of the pixel position at the corresponding address position in the inverted shift image 8a is 1 or more, and the gain K By performing the operation of multiplying by (128), the luminance difference is further expanded.
For this reason, as shown in the luminance profile 901 of FIG. 6C, in the processed image (division processed image 9a), the luminance variation in the background region is suppressed, and the luminance difference between the target object and the background region becomes large.
Thereby, since the contrast of the target object with respect to the background region in the image is enlarged, the target object can be easily extracted.

Here, the image (infrared image) which imaged the ship which navigates the sea surface and the floating matter on the sea surface with the uncooled infrared sensor part 3 in FIG. 7 is shown. FIG. 8 shows a luminance histogram of the infrared image of FIG. 7 (corresponding to FIG. 4).
Further, FIG. 9 shows a processed image (division processed image 9a) when the division processing is performed on the infrared image of FIG. 7 as described above. In this division processed image 9a, in particular, the contrast of the target object with respect to the background having a luminance level close to the intermediate value (128) in the number of display gradations is effectively enhanced as compared with the subtraction processed image 9a ′. FIG. 10 shows a luminance histogram of the processed image of FIG.

In the present embodiment, the luminance level of the background region (sea surface) that occupies the highest frequency on the luminance histogram (FIG. 7) is set to the intermediate value (128) of the number of gradations, and other luminance levels (pixels other than the back region) are set. (Brightness level) is linearly converted based on the intermediate value.
Thereby, the range width of the luminance distribution of the infrared image can be effectively expanded, and thus the visibility of the target object in the image is improved.

In the present embodiment, the case where the luminance calculation processing unit 9 executes one of the division process or the subtraction process for the luminance shift image 7a and the inverted shift image 8a of the infrared image is described. In addition, the subtraction process has the effect of eliminating white waves on the sea surface, and the division process has the effect of increasing the contrast between the background in the vicinity of the intermediate color tone and the target object, so a process combining the division process and the subtraction process is performed. Accordingly, it may be set to perform processing for further expanding the distribution width of the luminance distribution.
Thereby, the contrast between the background and the target object in the processed image is enlarged, and the visibility of the sailing ship on the sea surface that is the target object can be further enhanced.

Next, the feature detection unit 10 applies a luminance region that is less than a preset luminance frequency in the image to the division processing image 9a (or the subtraction processing image 9a ′) sent from the luminance calculation processing unit 9 in step S4. A filtering image 10a is generated by performing filtering processing to remove (pixels), and is transmitted to the contrast extension unit 11 (step S5).
Thereby, the luminance level of the mode value in the image (filtering image 10a) is relatively increased.

The contrast extension unit 11 selects a region (referred to as “luminance region”) that is equal to or higher than a preset luminance value of the input filtered image 10a, and based on the selected luminance mode value, this mode value. A contrast expansion image 11a having a contrast expanded by linearly expanding the luminance range of the luminance region so as to maximize the vicinity is generated and sent to the binarization processing unit 12 (step S6).

Next, the binarization processing unit 12 generates a binarized image 12a by performing preset threshold processing on the luminance of each pixel of the contrast expansion image 11a (binarization processing), and The binarized image 12a is output to the image display unit 5 (step S7).

Further, the binarization processing unit 12 may be configured to extract a target object from the binarized image 12a output to the image display unit 5 and to perform highlight display of the target object.
Here, the binarization processing unit 12 displays, for example, a mark set corresponding to the extracted target object so as to be superimposed on the target object in the binarized image 12a as the highlighting method. To do.

As described above, in the image identification support device 100 according to the present embodiment, the luminance of the target object with respect to the luminance of the background region is suppressed while suppressing the luminance variation of the background region that occupies most of the luminance distribution of the captured infrared image. Processing for enlarging contrast (luminance difference) can be performed.
In addition, by the division (subtraction) processing on the luminance shift image and the inverted shift image, it is possible to remove noise that is very small in the image and extremely small in luminance difference, and further, an effect of increasing the luminance difference can be obtained.
This eliminates the need for an expensive and complex optical system, and further eliminates the need for processing with a high computational load such as histogram quantization and individual noise removal processing, and enables rapid and effective image processing. Can be performed. In addition, it is possible to extract a target object in which the temperature difference between the target and the background is minute and it is difficult to ensure the contrast with the background.
Furthermore, it is possible to effectively extract floating substances between waves included in the infrared image and eliminate waves such as white waves.

As mentioned above, although this invention was demonstrated with reference to embodiment (and an Example), this invention is not limited to the said embodiment (and Example). Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

This application claims priority based on Japanese Patent Application No. 2009-196926 filed on Aug. 27, 2009, the entire disclosure of which is incorporated herein.

The present invention can be usefully applied to an infrared image processing system that automatically detects a target such as a ship using an infrared image.

DESCRIPTION OF SYMBOLS 1 Condensing dome 2 Optical part 3 Uncooled infrared sensor part 4 Image processing part 5 Image display part 6 Histogram calculating part 7 Brightness shift calculating part 8 Reverse image processing part (luminance calculating process part)
9 brightness calculation processing unit 10 feature detection unit 11 contrast expansion unit 12 binarization processing unit

Claims

An optical unit that collects infrared light from an imaging target including a target object and acquires a specific wavelength band from the collected infrared light, and the imaging target based on the wavelength acquired by the optical unit An image target identification device comprising: an uncooled infrared image sensor that generates an infrared image corresponding to the temperature of the image; and an image processing unit that generates an arithmetic processing image in which the target object is highlighted based on the infrared image. And
The image processing unit
A luminance histogram calculator that generates a luminance histogram indicating the frequency of appearance of luminance values of pixels included in the infrared image;
The luminance value having the highest appearance frequency in the luminance histogram is set to the intermediate value of the luminance range width in the infrared image, and the contrast relationship with the luminance value set to the intermediate value is maintained, and other values in the infrared image are maintained. A brightness level adjustment unit that generates a brightness shift image obtained by linearly shifting each brightness value;
An inverted shift image in which the brightness level of the luminance shift image is inverted is generated, and an arithmetic processing based on a difference in luminance value between corresponding pixels of the inverted shift image and the luminance shift image is performed to generate the arithmetic processing image. An image target identification device comprising: a luminance calculation processing unit.
In the image target identification device according to claim 1,
The luminance calculation processing unit calculates an average luminance level value of each of the luminance shift image and the inverted shift image, and calculates one image of the luminance shift image and the inverted shift image based on the average luminance level value as a high brightness side image. And brightness difference image setting means for setting the other image as a low brightness side image,
Subtraction processing means for generating the arithmetic processing image by executing, as the arithmetic processing, processing for subtracting the frequency value of each pixel at the corresponding position of the low lightness side image from the luminance value of each pixel in the high brightness side image. An image target identification device comprising:
In the image target identification device according to claim 2,
Division processing means for executing the division process of dividing the luminance value of each pixel in the high brightness side image by the frequency value of each pixel at the corresponding position of the low brightness side image as the calculation process and generating the calculation process image. An image target identification device comprising:
In the image target identification device according to claim 1,
The luminance calculation processing unit calculates an average luminance level value of each of the luminance shift image and the inverted shift image, and calculates one image of the luminance shift image and the inverted shift image based on the average luminance level value as a high brightness side image. And brightness difference image setting means for setting the other image as a low brightness side image,
Division processing means for executing the division process of dividing the luminance value of each pixel in the high brightness side image by the frequency value of each pixel at the corresponding position of the low brightness side image as the calculation process and generating the calculation process image. An image target identification device comprising:
In the image target identification device according to claim 1, 2, 3, or 4,
A filtering processing unit that generates a filtering image by removing luminance in pixels in a luminance region less than a preset luminance frequency in the arithmetic processing image;
A contrast expansion unit that performs a process of expanding the contrast of the filtered image based on the luminance value included in the filtered image;
An image target identification apparatus comprising: a binarization processing unit that performs a process of binarizing pixels having a luminance level equal to or higher than a certain value in the arithmetic processing image.
An image target that collects infrared light from an imaging target including the target object and generates an arithmetic processing image in which the target object is highlighted based on the temperature of the imaging target from the condensed infrared light. An identification method,
Generating a luminance histogram indicating the frequency of appearance of luminance values of pixels included in the infrared image;
The luminance value having the highest appearance frequency in the luminance histogram is set to the intermediate value of the luminance range width in the infrared image, and the contrast value with the luminance value set in the intermediate value is maintained, and the other values in the infrared image are maintained. A brightness shift image in which each brightness value is linearly shifted is generated,
Generating an inverted shift image by inverting the luminance level of the luminance shift image;
An image target identification method, wherein the arithmetic processing image is generated by performing arithmetic processing based on a difference in luminance values at corresponding positions of the luminance shift image and the inverted shift image.
An operation in which infrared light from an imaging target including the target object is condensed and the target object is highlighted based on an infrared image generated from the collected infrared light based on the temperature of the imaging target An image target identification program for generating a processed image,
A luminance histogram calculation function for generating a luminance histogram indicating the frequency of appearance of luminance values of pixels included in the infrared image;
The luminance value having the highest appearance frequency in the luminance histogram is set to the intermediate value of the luminance range width in the infrared image, and the contrast value with the luminance value set in the intermediate value is maintained, and the other values in the infrared image are maintained. A brightness level adjustment function for generating a brightness shift image obtained by linearly shifting each brightness value of
An inverted shift image generation function for generating an inverted shift image obtained by inverting the luminance level of the luminance shift image;
A luminance calculation processing function for generating the calculation processing image by performing calculation processing based on a difference in luminance value at corresponding positions of the luminance shift image and the inverted shift image;
An image target identification program for causing a computer to execute.