JP2012088785A - Object identification device and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce erroneous determination by properly selecting an image suitable for identification of an object.SOLUTION: An imaging device 12 acquires a plurality of captured images taken under different exposure conditions. A window image extraction unit 22 sets a determination frame in each captured image, and extracts window images by scanning the determination frames. A window image selection unit 24 calculates dispersion of luminance distribution of each window image from respective window images of the same size that is extracted from the same position on the captured images, and selects window images whose dispersion of luminance distribution is equal to or more than a predetermined value. An identification unit 28 identifies whether the selected window image represents an object based on the selected window image and an identification model stored in an identification model storage unit 26.

Description

  The present invention relates to an object identification device and a program, and more particularly to an object identification device and a program for identifying an object from a captured image.

  In recent years, an increasing number of vehicles are equipped with an object identification device that performs image processing on a video around a vehicle imaged by an in-vehicle camera, identifies a specific object, and presents an identification result to a driver. In such an object identification device, it is a problem that an image that can easily identify the object is obtained.

  Therefore, in an in-vehicle image recognition device, problems such as overexposure and underexposure of an image caused by a lack of the dynamic range of the camera are taken, and a plurality of images are captured with different exposure times, and the same partial image of the plurality of images A vehicle-mounted image recognition device that compares the contrasts of the images and generates a composite image in which a partial image having the best contrast is selected and performs image recognition has been proposed (see, for example, Patent Document 1).

  In addition, an authentication device that performs bracket shooting has been proposed in order to prevent a reduction in recognition performance caused by improper camera exposure control (see, for example, Patent Document 2). In the authentication apparatus described in Patent Document 2, in order to prevent an increase in processing time due to an increase in the number of bracket shootings, a feature image is detected from the captured image, an exposure correction value is generated, and the first bracket shooting is performed. The first exposure control value for shooting is calculated, and the second exposure control value for second shooting is calculated based on the first exposure control value and the exposure correction value.

JP 2009-177250 A JP 2008-5081 A

  However, in the technique of Patent Document 1, since partial images having different exposure times are combined into one image, there is a problem in that a boundary line that is a discontinuous point of gradation occurs at the boundary between the partial image and the partial image. is there. This can be solved by appropriately setting the cutout position of the partial image, but it can be applied when it is unknown where and in what size the object is imaged in the image. Can not. Moreover, in the technique of Patent Document 1, a process of making the boundary line inconspicuous is performed by obtaining a composite image by weighting according to the contrast of the partial image, but this process reduces the contrast of the composite image, There is a problem that the image is not suitable for identification of the object.

  Moreover, since the technique of patent document 2 is an apparatus which authenticates the face image imaged at the center of the screen, the position and size of the target object on the image are specified, but these are not specified Then, since the exposure correction value cannot be calculated, there is a problem that the second exposure control value cannot be calculated. Further, there is a problem that it is not clear how to determine the second exposure control value even when there are a plurality of objects on the image and the optimum exposure control values are different from each other.

  The present invention has been made to solve the above-described problems, and provides an object identification device and a program capable of appropriately selecting an image suitable for identification of an object and reducing erroneous determination. For the purpose.

  In order to achieve the above object, an object identification device according to a first aspect of the present invention is an acquisition unit that acquires a plurality of captured images captured under different exposure conditions, and each of a plurality of captured images acquired by the acquisition unit. An extraction unit that sets a determination frame to the image and extracts an image in the determination frame as a window image while changing the position and size of the determination frame; and the same size extracted from the same position of each of the plurality of captured images A window image selecting means for selecting a window image from which each object is easy to identify based on a luminance distribution of each of the window images; a window image selected by the window image selecting means; and the object. Identification means for identifying whether the window image is an image representing the object based on an identification model generated in advance for identifying It is configured to include a.

  According to the object identification device of the first invention, the acquisition unit acquires a plurality of captured images captured under different exposure conditions, and the extraction unit determines a determination frame for each of the plurality of captured images acquired by the acquisition unit. And the image within the determination frame is extracted as a window image while changing the position and size of the determination frame. By changing the position and size of the determination frame, it is possible to detect objects of various sizes existing at various positions on the captured image.

  Then, the window image selection means selects a window image from which each object is easily identified based on the luminance distribution of each window image from each of the same size window images extracted from the same position of each of the plurality of captured images. select. Since an image that can easily identify an object is an image that does not cause blackout or whiteout, a window image that is free from blackout or whiteout is selected based on the luminance distribution of the window image. Then, the identification unit identifies whether the window image is an image representing the object based on the window image selected by the window image selection unit and the identification model generated in advance for identifying the object.

  In this way, a window image having the same size is extracted from the same position of a plurality of captured images captured under different exposure conditions, and a window image that can easily identify an object is selected based on the luminance distribution of each window image. Therefore, it is possible to appropriately select an image suitable for identifying an object and reduce erroneous determination.

  In the first invention, the acquisition unit acquires a plurality of captured images captured under different exposure conditions during light projection and a plurality of captured images captured under the different exposure conditions during non-projection. And generating means for generating a difference image between the picked-up image at the time of projection and the non-light-projected image obtained under the same exposure condition acquired by the acquiring means, wherein the extracting means is obtained from each of the difference images. A window image can be extracted. As a result, when the object is an object that does not emit light such as a pedestrian, it is possible to improve the identification accuracy of the object by excluding illumination components from other light emitters such as street lights and oncoming light. .

  In addition, the object identification device of the second invention includes an acquisition unit that acquires a plurality of captured images captured under different exposure conditions, and a size of a determination frame from each of the plurality of captured images acquired by the acquisition unit. A picked-up image selection means for selecting a picked-up image that is brighter as the value is smaller, and setting the judgment frame to the picked-up image selected by the picked-up image selection means, and changing the position of the judgment frame to display an image in the judgment frame Whether or not the window image is an image representing the object based on extraction means for extracting as an image, and a window image extracted by the extraction means and an identification model generated in advance for identifying the object And identification means for identifying.

  According to the object identification device of the second invention, the acquisition unit acquires a plurality of captured images captured under different exposure conditions, and the captured image selection unit determines from each of the plurality of captured images acquired by the acquisition unit. As the size of the determination frame is smaller, a brighter captured image is selected. This is based on the property that the farther object appears darker and the farther object appears smaller on the captured image.

  Then, the extraction unit sets a determination frame for the captured image selected by the captured image selection unit, extracts the image in the determination frame as a window image while changing the position of the determination frame, and the identification unit uses the extraction unit. Based on the extracted window image and an identification model generated in advance to identify the object, it is identified whether the window image is an image representing the object.

  Thus, since the brighter image is selected as the size of the determination frame is smaller, that is, the object is farther away, it is possible to appropriately select an image suitable for identifying the object and reduce erroneous determination. .

  In the second invention, the acquisition unit acquires a plurality of captured images captured under different exposure conditions during light projection and a plurality of captured images captured under the different exposure conditions during non-projection. Generating means for generating a difference image between a captured image obtained during projection and a captured image obtained during non-projection under the same exposure condition acquired by the acquisition means, and the captured image selection means includes: A difference image that makes it easy to identify an object can be selected from each. As a result, when the object is an object that does not emit light such as a pedestrian, it is possible to improve the identification accuracy of the object by excluding illumination components from other light emitters such as street lights and oncoming light. .

  In addition, the exposure condition can be varied by varying the light projection intensity, or any one of the aperture, exposure time, and gain can be varied, and the light projection intensity can be varied. . As a result, it is possible to identify an object located farther away, and it is possible to vary the brightness of the captured image more finely.

  The object identification program according to the third aspect of the present invention provides a computer with an acquisition unit for acquiring a plurality of captured images captured under different exposure conditions, and a determination frame for each of the plurality of captured images acquired by the acquisition unit. An extraction means for extracting and extracting an image in the determination frame as a window image while changing the position and size of the determination frame, from each of the window images of the same size extracted from the same position of each of the plurality of captured images A window image selecting means for selecting a window image that is easy to identify an object based on a luminance distribution of each of the window images, and a window image selected by the window image selecting means for identifying the object Identifying means for identifying whether the window image is an image representing the object based on an identification model generated in advance; Is a program for making the function Te.

  The object identification program according to the fourth aspect of the present invention is directed to an acquisition unit that acquires a plurality of captured images captured under different exposure conditions, and a determination frame from each of the plurality of captured images acquired by the acquisition unit. A picked-up image selection unit that selects a brighter picked-up image as the size of the image is smaller, sets the determination frame to the picked-up image selected by the picked-up image selection unit, and changes the position of the determination frame to change the image in the determination frame Extraction means for extracting as a window image, and whether the window image is an image representing the object based on the window image extracted by the extraction means and an identification model generated in advance for identifying the object It is a program for functioning as an identification means for identifying the above.

  The storage medium for storing the program of the present invention is not particularly limited, and may be a hard disk or a ROM. Further, it may be a CD-ROM, a DVD disk, a magneto-optical disk or an IC card. Furthermore, the program may be downloaded from a server or the like connected to the network.

  As described above, according to the object identification device and the program of the present invention, the object is based on the luminance distribution of the window image extracted from the determination frame set in the captured image or based on the size of the determination frame. Since an image that can be easily identified is selected, it is possible to appropriately select an image suitable for identifying an object and reduce erroneous determination.

It is a block diagram which shows schematic structure of the target object identification apparatus which concerns on 1st Embodiment. It is an image figure which shows an example of three types of captured images from which exposure conditions differ. It is an image figure for demonstrating the setting of a determination frame. It is an image figure for demonstrating selection of the window image based on luminance distribution. It is a flowchart which shows the content of the target object identification process routine performed in the computer of the target object identification apparatus which concerns on 1st Embodiment. It is a block diagram which shows schematic structure of the target object identification apparatus which concerns on 2nd Embodiment. It is an image figure which shows an example of the captured image at the time of light projection for every exposure condition, the captured image at the time of non-light projection, and a difference image. It is an image figure for demonstrating the relationship between the magnitude | size of a target object and exposure conditions. It is a flowchart which shows the content of the target object identification process routine performed in the computer of the target object identification apparatus which concerns on 2nd Embodiment.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the present embodiment, a case will be described in which the present invention is applied to an object identification device that is mounted on a vehicle and identifies a pedestrian as an object.

  As shown in FIG. 1, an object identification device 10 according to the first embodiment is attached to a vehicle (not shown), and captures an image of the front of the vehicle to generate an image, and an imaging The computer 14 which performs the process which identifies a pedestrian from the captured image imaged with the apparatus 12, and the display apparatus 16 which displays the identification result of the computer 14 are provided.

  The imaging device 12 includes an imaging unit (not illustrated) that generates an image signal of an image captured in front of the vehicle, and an A / D conversion unit (not illustrated) that performs A / D conversion on the image signal generated by the imaging unit. And an image memory (not shown) for temporarily storing the A / D converted image signal.

  The computer 14 includes a CPU, a RAM, and a ROM that stores a program for executing an identification processing routine to be described later, and is functionally configured as follows. The computer 14 includes an exposure condition setting unit 20 that sets an exposure condition of the imaging device 12, a window image extraction unit 22 that extracts a window image from each of a plurality of captured images captured by the imaging device 12 under different exposure conditions, Based on the luminance distribution of each window image, a window image selection unit 24 that selects a window image for identifying an object, and an identification model storage unit 26 that stores an identification model generated in advance by learning are selected. The identification unit 28 for identifying whether the window image is an image representing a pedestrian based on the window image and the identification model, and the identification result in the identification unit 28 is superimposed on the captured image on the display device 16. And a display control unit 30 that controls the display.

  The exposure condition setting unit 20 sets at least one of an aperture, an exposure time, and a gain for the imaging device 12 as exposure conditions in accordance with the illumination state of the object scene. In the present embodiment, in one cycle imaging, in particular, standard exposure with an exposure time as a standard exposure time, overexposure with an exposure time longer than the standard, and underexposure with an exposure time shorter than the standard. Imaging is performed under the three types of exposure conditions. The exposure condition setting unit 20 sequentially switches the exposure conditions within one cycle of the imaging period so that three captured images corresponding to the three types of exposure conditions are obtained. Thereby, for example, three captured images as shown in FIG. 2 are obtained.

  The window image extraction unit 22 sets a determination frame for each of the three captured images captured by the imaging device 12, and extracts an image within the determination frame as a window image. As shown in FIG. 3, the determination frames are prepared in various sizes, scanned on the captured image, and window images of various positions and sizes are extracted.

  The window image selection unit 24 uses the window image extraction unit 22 to calculate the luminance distribution of each of the window images of the same size extracted from the same position of each of the three captured images, and identify the object. Select the most suitable window image for. FIG. 4 shows an example of the window image extracted from the determination frame set in the captured image under the exposure conditions of underexposure, standard exposure, and overexposure, and the luminance distribution of each window image. In the luminance distribution in which a high peak appears on the low luminance side as in the case of underexposure or standard exposure, the image is blacked out and cannot be said to be an image suitable for identification. Further, in the luminance distribution in which a high peak appears on the high luminance side, the image is overexposed, and in this case, it cannot be said that the image is suitable for identification. On the other hand, when the luminance distribution is gently distributed in a predetermined range from low luminance to high luminance, as in the case of overexposure, there is no blackout or overexposure, and the image is suitable for identification. You can say that. Therefore, the window image selection unit 24 selects a window image whose luminance distribution is greater than or equal to a predetermined value as a window image used for object identification.

  The identification model storage unit 26 stores, for example, an identification model for identifying an object learned by a technique such as SVM (Support Vector Machine).

  Based on the identification model stored in the identification model storage unit 26 and the window image selected by the window image selection unit 24, the identification unit 28 uses, for example, a conventionally known SVM classifier to walk the window image. Whether or not the image represents the person.

  Next, an object identification processing routine executed by the computer 14 of the object identification apparatus 10 according to the first embodiment will be described with reference to FIG.

  In step 100, the imaging device 12 is controlled so that three captured images captured under three types of exposure conditions of standard exposure, overexposure, and underexposure are obtained, and the captured captured image is acquired.

  Next, in step 102, for each of the three captured images acquired in step 100, a determination frame of the same size is set to a predetermined position (for example, the upper left corner) on the image, and the image in the determination frame is selected. Extract as a window image.

  Next, in step 104, the variance of the luminance distribution of each of the window images extracted from the three captured images in step 102 is obtained, and a window image having a luminance distribution variance equal to or greater than a predetermined value is determined as an object. Is selected as a window image used for identification.

  Next, based on the identification model stored in the identification model storage unit 26 in step 106 and the window image selected in step 104, for example, a window image is walked using a conventionally known SVM classifier. Whether or not the image represents the person. When the window image is identified as an image indicating a pedestrian, the position and size of the determination frame when the window image is extracted are stored in a predetermined storage area.

  Next, in step 108, it is determined whether or not the process of scanning the determination frame for the entire range of the captured image and extracting the window image is completed. If it is determined that the process has not been completed for the entire range of the captured image, the process proceeds to step 110, the determination frame is moved by one step (predetermined pixels), and the process returns to step 102. On the other hand, if it is determined that the process has been completed for the entire range of the captured image, the process proceeds to step 112 to determine whether the process has been completed for all sizes of the determination frame. If it is determined that the processing has not been completed for all the sizes of the determination frames, the process proceeds to step 114, the determination frame of the next size is selected, and the process returns to step 102. If it is determined that the processing has been completed for all the determination frame sizes, the process proceeds to step 116, and based on the identification result stored in the predetermined storage area in step 106, for example, a pedestrian of the captured image A portion identified as an image representing the image is displayed on the display device 16 by surrounding it with a rectangular frame.

  As described above, according to the object identification device according to the first embodiment, a window image from a determination frame of the same size set at the same position of each of a plurality of captured images captured under different exposure conditions. In order to identify whether or not the image shows the target object by selecting a window image in which the variance of the luminance distribution of each window image is equal to or greater than a predetermined value, an image suitable for identifying the target object is appropriately selected. Thus, erroneous determination can be reduced.

  Further, it is not necessary to synthesize a plurality of captured images before performing the identification process or to integrate the identification results after the identification process.

  In the first embodiment, the case where an appropriate window image is selected based on whether or not the distribution of the luminance distribution of the window image is equal to or greater than a predetermined value has been described. However, the present invention is not limited to this. It is only necessary to eliminate an image that is not suitable for identification such as blackout or whiteout. Therefore, it is sufficient that the luminance distribution is within a predetermined range. For example, it is determined whether the average value, median value, deviation, mode value, etc. of the luminance value of the window image is within the predetermined range, or the maximum value of the luminance value. Is determined to be less than or equal to a predetermined value, whether or not the minimum luminance value is equal to or greater than a predetermined value, or whether or not the difference between the maximum and minimum luminance values is within a predetermined range. Or you may. Further, it may be determined whether the image is suitable for identification based on the shape of the distribution curve indicating the luminance distribution.

  Next, a second embodiment will be described. In addition, about the structure similar to the target object identification apparatus 10 of 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  As illustrated in FIG. 6, the object identification device 210 according to the second embodiment includes an imaging device 12, a light projecting device 18 that projects light onto an imaging target region by the imaging device 12, and the imaging device 12. The computer 214 which performs the process which identifies a pedestrian from the imaged captured image, and the display apparatus 16 are provided.

  Functionally, the computer 214 has a plurality of exposure condition setting units 20, a light projection control unit 34 that controls light projection and non-light projection of the light projecting device 18, and a plurality of images captured by the imaging device 12 under different exposure conditions. For each of the captured images, a difference image generation unit 36 that generates a difference image of light projection and non-light projection, and a difference image for identifying an object based on the size of the determination frame is selected from the generated difference image A difference image selection unit 38, a window image extraction unit 222 that extracts a window image from the selected difference image, an identification model storage unit 26, an identification unit 28, and a display control unit 30. be able to.

  The light projection control unit 34 captures three captured images captured during light projection under three types of exposure conditions of standard exposure, overexposure, and underexposure, and standard exposure, overexposure, and underexposure. The light projection device 18 is switched on and off so that three captured images captured at the time of non-projection can be obtained under the three types of exposure conditions.

  The difference image generation unit 36 generates a difference image between a captured image captured at the same exposure condition and an image captured at the time of non-projection. FIG. 7 shows an example of a captured image at the time of projection, a captured image at the time of non-projection, and a difference image for each exposure condition. Thus, by generating the difference image, it is possible to exclude illumination components from other light emitters such as street lights and oncoming vehicle lights.

  The difference image selection unit 38 selects the most suitable window image for identifying the target object from the generated difference image according to the size of the determination frame when the window image extraction unit 222 in the subsequent stage extracts the window image. select. Since the luminance value of the difference image attenuates in inverse proportion to the square of the distance from the light projecting device 18, there is a property that a farther object is captured darker. Further, when identifying a distant object, a window image is extracted from a small determination frame, and when identifying a near object, a window image is extracted from a large determination frame. Therefore, it can be said that it is more suitable to use a bright window image as an image used for identification as the size of the determination frame is smaller. Therefore, as shown in FIG. 8, when the size of the determination frame is large, that is, when the target is a nearby pedestrian, the difference image selection unit 38 selects a difference image of the underexposed captured image, and determines When the size of the frame is small, that is, when a distant pedestrian is targeted, an overexposed difference image is selected.

  Next, an object identification processing routine executed by the computer 214 of the object identification apparatus 210 according to the second embodiment will be described with reference to FIG. In addition, about the process similar to the target object identification process of 1st Embodiment, the same code | symbol is attached | subjected and detailed description is abbreviate | omitted.

  In step 200, the light projecting device 18 is controlled to project light onto the imaging target area, so that three captured images captured under three types of exposure conditions of standard exposure, overexposure, and underexposure are obtained. Then, the imaging device 12 is controlled to acquire a captured image.

  Next, in step 202, the light projecting device 18 is controlled so as to turn off the light projection to the imaging target area, and three captured images captured under three types of exposure conditions of standard exposure, overexposure, and underexposure. Then, the imaging device 12 is controlled to obtain a captured image.

  Next, in step 204, from each of the captured images acquired in step 200 and step 202, a difference image between the captured image at the time of projection and the captured image at the time of non-projection captured under the same exposure conditions is generated. To do. As a result, a difference image for standard exposure, a difference image for overexposure, and a difference image for underexposure are generated.

  Next, in step 206, the size of the determination frame is selected. For example, three sizes of determination frames of large (20 × 40 pixels), medium (16 × 32 pixels), and small (12 × 24 pixels) are prepared, and these sizes are sequentially selected.

  Next, in step 208, the most suitable window image for identifying the object is selected from the difference image generated in step 204 according to the size of the determination frame selected in step 206. For example, if the size of the judgment frame selected in the above step is “Large”, select an underexposed difference image, if “Medium”, select a standard exposure difference image, and select “Low” an overexposed difference image. can do.

  Next, in step 212, the determination frame of the size selected in step 206 is set to a predetermined position (for example, the upper left corner) on the image with respect to the difference image selected in step 208, and the image within the determination frame is set. Whether or not the window image is an image representing a pedestrian based on the identification model stored in the identification model storage unit 26 and the extracted window image in step 106. Identify

  Next, in step 108, it is determined whether or not the process has been completed for the entire range of the difference image. If it is determined that the process has not been completed, the process proceeds to step 110 and the determination frame is set to one step. Move (predetermined pixels) and return to step 212. On the other hand, if it is determined that the process has been completed for the entire range of the difference image, the process proceeds to step 112, where it is determined whether the process has been completed for all sizes of the determination frame, and the process has been completed. If it is determined that there is not, the process proceeds to step 114, the determination frame of the next size is selected, and the process returns to step 208. When it is determined that the processing has been completed for all the determination frame sizes, the process proceeds to step 116, and the identification result is displayed.

  As described above, according to the object identification device according to the second embodiment, a brighter image is selected as the size of the determination frame is smaller, that is, the object is farther away, and a window is selected from the selected image. Since the image is extracted to identify whether or not the image indicates the object, it is possible to appropriately select an image suitable for identifying the object and reduce erroneous determination. In addition, by using the difference image at the time of projection-non-projection, it is possible to exclude illumination components due to other light emitters such as street lights and oncoming light, and also the object on the image with respect to the exposure conditions It is possible to clarify the difference in luminance value.

  In the second embodiment, the case of identifying an object using a difference image between a captured image at the time of projection and a captured image at the time of non-projection has been described. An image used for identification may be selected from a plurality of captured images with different exposure conditions according to the size of the determination frame. Even in the captured image captured without performing the light projection control, the distant object has the property of being captured darker, so the present embodiment can be applied without using the difference image. Can do.

  Also in the first embodiment, light projection control may be performed to identify an object using a difference image between a captured image during light projection and a captured image during non-light projection. In this case, instead of step 100 of the object identification process (FIG. 5) in the first embodiment, steps 200 to 204 of the object identification process (FIG. 9) in the second embodiment are executed. In 102, a window image may be extracted from each difference image of each exposure condition.

  In the above embodiment, the case where the exposure conditions are three types has been described. However, the present invention is not limited to this. Further, the present invention is not limited to the case where the exposure conditions are varied depending on the aperture, the exposure time, and the gain. Instead of or in addition to the change of the exposure conditions based on the aperture, the exposure time, and the gain, the light projection intensity of the light projecting device 18 is changed. By making them different, a plurality of captured images with different brightness may be captured. By performing the light projection, it is possible to identify an object that exists farther away. Further, the brightness of the captured image can be varied more finely by combining the change of the exposure condition by the imaging device and the change of the light projection intensity.

  In the above embodiment, a case has been described in which a plurality of captured images with different exposure conditions at the time of light projection are acquired, and then a plurality of captured images with different exposure conditions at the time of non-light projection are acquired. After acquiring the captured image at the time of projection and the captured image at the time of non-projection under the conditions, the captured image at the time of projection and the captured image at the time of non-projection may be acquired under different exposure conditions.

  In the above-described embodiment, the case where the image identification process is performed using the SVM classifier has been described as an example. However, the present invention is not limited to this, and other conventionally known identification methods (for example, nearest neighbors) The image identification processing may be performed using identification, linear discrimination, or the like.

  In the above embodiment, the case where the window is extracted by scanning the entire range of the captured image or the difference image as the determination frame has been described. However, for example, the determination frame is detected by performing pre-processing such as detecting a road surface area. The scanning range may be limited.

  Moreover, although the case where this invention was applied to the target object identification apparatus which identifies a pedestrian as a target object in the said embodiment was demonstrated to the example, a motor vehicle, a two-wheeled vehicle, etc. may be sufficient as a target object.

  In the specification of the present application, the embodiment has been described in which the program is installed in advance. However, the program may be provided by being stored in a storage medium such as a CDROM.

DESCRIPTION OF SYMBOLS 10,210 Object identification device 12 Imaging device 14, 214 Computer 16 Display device 18 Projection device 20 Exposure condition setting unit 22, 222 Window image extraction unit 24 Window image selection unit 26 Identification model storage unit 28 Identification unit 30 Display control unit 34 Light projection control unit 36 Difference image generation unit 38 Difference image selection unit

Claims (9)

Acquisition means for acquiring a plurality of captured images each captured under different exposure conditions;
An extraction unit that sets a determination frame for each of the plurality of captured images acquired by the acquisition unit, and extracts an image in the determination frame as a window image while changing the position and size of the determination frame;
Window image selection means for selecting, from each of window images of the same size extracted from the same position of each of the plurality of captured images, a window image that can easily identify an object based on the luminance distribution of each of the window images. When,
Identification means for identifying whether or not the window image is an image representing the object based on a window image selected by the window image selection means and an identification model generated in advance to identify the object; ,
An object identification device including: The acquisition means acquires a plurality of captured images captured under different exposure conditions during light projection and a plurality of captured images captured under different exposure conditions during non-light projection,
Including generation means for generating a difference image between the captured image at the time of projection and the captured image at the time of non-projection of the same exposure condition acquired by the acquisition means,
The object identifying apparatus according to claim 1, wherein the extraction unit extracts a window image from each of the difference images. Acquisition means for acquiring a plurality of captured images each captured under different exposure conditions;
From each of the plurality of captured images acquired by the acquisition unit, a captured image selection unit that selects a brighter captured image as the size of the determination frame is smaller;
An extraction unit that sets the determination frame on the captured image selected by the captured image selection unit and extracts an image in the determination frame as a window image while changing the position of the determination frame;
Identification means for identifying whether the window image is an image representing the object based on a window image extracted by the extraction means and an identification model generated in advance to identify the object;
An object identification device including: The acquisition means acquires a plurality of captured images captured under different exposure conditions during light projection and a plurality of captured images captured under different exposure conditions during non-light projection,
Including generation means for generating a difference image between the captured image at the time of projection and the captured image at the time of non-projection of the same exposure condition acquired by the acquisition means,
The object identification device according to claim 3, wherein the captured image selection unit selects a difference image from which the object is easily identified from each of the difference images.   The object identification device according to any one of claims 1 to 4, wherein the exposure condition is varied by varying a light projection intensity.   The object identification according to any one of claims 1 to 5, wherein the exposure condition is changed by changing any one of an aperture, an exposure time, and a gain, and by changing a light projection intensity. apparatus. Computer
Acquisition means for acquiring a plurality of captured images each captured under different exposure conditions;
An extraction unit that sets a determination frame for each of the plurality of captured images acquired by the acquisition unit and extracts an image in the determination frame as a window image while changing the position and size of the determination frame;
Window image selection means for selecting, from each of window images of the same size extracted from the same position of each of the plurality of captured images, a window image that can easily identify an object based on the luminance distribution of each of the window images. And identifying whether or not the window image is an image representing the object based on a window image selected by the window image selecting means and an identification model generated in advance for identifying the object. An object identification program for functioning as a means. Computer
Acquisition means for acquiring a plurality of captured images each captured under different exposure conditions;
A captured image selection unit that selects a brighter captured image as the size of the determination frame is smaller from each of the plurality of captured images acquired by the acquisition unit;
An extraction unit that sets the determination frame on the captured image selected by the captured image selection unit, extracts an image in the determination frame as a window image while changing the position of the determination frame, and an image extracted by the extraction unit An object identification program for functioning as identification means for identifying whether or not the window image is an image representing the object based on a window image and an identification model generated in advance for identifying the object.   An object identification program for causing a computer to function as each means constituting the object identification device according to any one of claims 1 to 6.