JP2021163222A - Object detection apparatus - Google Patents

Abstract

To accurately identify an object, such as a pedestrian, from a background in image data.SOLUTION: An image pre-processing unit 22 generates inverted image data obtained by applying grayscale conversion on captured image data obtained by a far-infrared camera 10. An object discriminator 24 performs object identification processing based on a result obtained by learning each of the captured image data and the inverted image data using learning data in advance, to output an identification result including information on object type identification in each of the captured image data and the inverted image data and accuracy information indicating reliability of the identification in each of the captured image data and the inverted image data. An image post-processing unit 26 finally identifies the object in the captured image data on the basis of the identification result obtained by the object discriminator 24.SELECTED DRAWING: Figure 1

Description

The present invention relates to an object detection device.

Patent Document 1 describes the generation of a shift image by level-shifting the brightness value of a far-infrared image or the like, the generation of an inverted image of the image, and the brightness value of each pixel of the shifted image as the brightness value of the corresponding pixel of the inverted image. Disclosed is a technique for enlarging the contrast of an image by subtracting or dividing with, and facilitating the distinction between an object such as a pedestrian and the background.

Patent Document 2 utilizes the fact that when a far-infrared image is used as a detection target image, the brightness value of the region of a person having a high temperature becomes large, and the brightness of a peripheral region having a low temperature becomes smaller than that of a person. The invention of a person detection device for detecting a person or the like from a far-infrared image is disclosed.

Japanese Patent No. 5664551 Japanese Patent No. 6627680

However, in the far-infrared image, the brightness of the pedestrian and the background of the road or the like is reversed between summer and winter. Therefore, in order to ensure the accuracy of discrimination between the pedestrian and the background from the far-infrared image by the technique disclosed in Patent Document 2, the learning data according to the season is prepared and the device is trained by the learning data. However, there is a problem that a great deal of cost and labor are required to prepare the training data and learn the device.

Further, the technique disclosed in Patent Document 1 is effective if the background is an image with few textures such as the sea or the sky that falls within a certain brightness range, but the road surface and road during traveling are effective. When surrounding buildings or forests coexist as a background, there is a problem that it is difficult to effectively emphasize the contrast between the background and an object.

The present invention has been made in consideration of the above facts, and an object of the present invention is to obtain an object detection device capable of accurately identifying an object such as a pedestrian from the background in image data.

The object detection device according to the first aspect is an image processing unit that acquires far-infrared image data of the surrounding environment and an image process that generates converted image data obtained by gradation-converting the far-infrared image data acquired by the photographing unit. The far-infrared image data and the converted image data are obtained by performing object identification processing based on the results previously learned using the training data for each of the unit, the far-infrared image data, and the converted image data. Object identification that executes an identification process that outputs an identification result including information on identification of the type of object in each of the above and accuracy information indicating the reliability of the identification in each of the far-infrared image data and the converted image data. A unit and an object determination unit that finally identifies an object in the far-infrared image data based on the identification result by the object identification unit are included.

In the far-infrared image, the brightness of an object such as a pedestrian and the background (mainly a road) is reversed between summer and winter, but in the first aspect, the captured image data and the captured image data are gradation-converted. By performing object identification based on the same learning data for each of the converted image data obtained in the above, a good object identification result can be obtained in either the data or the converted image data.

In the second aspect, in the first aspect, the image processing unit generates the converted image data by inverting the brightness value of the far infrared image data acquired by the photographing unit.

As a result, in summer and winter, it is possible to identify an object corresponding to the characteristics of far-infrared image data in which the light and darkness of the object and the background are reversed.

In the third aspect, in the first aspect, the image processing unit uses the far-infrared image based on the correspondence between the predetermined gradation of the far-infrared image data and the gradation of the converted image data. The converted image data is generated from the external image data.

As a result, it is possible to generate inverted image data that is optimal for object identification by changing the characteristic of inverting the brightness of the gradation of the far infrared image data.

In the fourth aspect, in any one of the first to third aspects, the object identification unit identifies the position of the object in each of the far-infrared image data and the converted image data. The identification result including the position information of the identified object is output, and the object determination unit includes the position information of the object and the accuracy information in each of the far-infrared image data and the converted image data included in the identification result. The final object identification is performed based on.

This makes it easy to identify the position of the object and the object in the far-infrared image data.

In the fifth aspect, in any one of the first to third aspects, the object identification unit identifies the position of the object in each of the far-infrared image data and the converted image data. , A mapping image in which the accuracy information in each of the far-infrared image data and the converted image data is associated with the position of the object in each of the identified far-infrared image data and the converted image data is generated as the identification result. Then, the object determination unit superimposes the mapping images generated from each of the far-infrared image data and the converted image data, thereby superimposing the identification result in the far-infrared image data and the identification result in the converted image data. And are integrated to perform final object identification.

By integrating the identification result in the far-infrared image data and the identification result in the converted image data, the accuracy of object identification is improved.

A sixth aspect is, in any one of the first to fifth aspects, the learning data is the far-infrared image data for learning, the converted image data for learning, and the learning data. At least one of the visible light image data.

In the seventh aspect, in any one of the first to third aspects, the photographing unit further acquires visible light image data of the surrounding environment, and the object identification unit obtains the far infrared data. Information on identification of the type of object in each of the image data, the converted image data, and the visible light image data, and the reliability of the identification in each of the far infrared image data, the converted image data, and the visible light image data. The identification process for outputting the identification result including the accuracy information representing the above is executed, and the object determination unit finally identifies the object in the far-infrared image data based on the identification result by the object identification unit. ..

As a result, the identification results of the captured image data and the converted image data of the far-infrared image are integrated with the identification results in the visible light image data, and the reliability of the final object identification is determined, thereby determining the visible light image. Even the object identification unit learned from the data can perform object identification with high accuracy.

In the eighth aspect, in the seventh aspect, the learning data is the visible light image data for learning.

It has the effect of being able to accurately identify an object such as a pedestrian from the background in the image data.

It is a block diagram which showed an example of the object detection apparatus which concerns on 1st Embodiment of this invention. It is a flowchart which showed an example of the processing of the object detection apparatus which concerns on 1st Embodiment of this invention. (A) shows an example of object identification in captured image data by an object classifier, (B) shows an example of object identification in inverted image data by an object classifier, and (C) shows an example of object identification in captured image data. An example of a mapping image of captured image data created by extracting the reliability is shown, and (D) shows an example of a mapping image of the inverted image data created by extracting the identification reliability in the inverted image data. , (E) are schematic views showing an example of integration of identification results in each of the captured image data and the inverted image data. It is explanatory drawing of the inversion of the gradation of the photographed image data in the 2nd Embodiment of this invention. It is a block diagram which showed an example of the object detection apparatus which concerns on 3rd Embodiment of this invention. It is a functional block diagram which showed an example of the object detection apparatus which concerns on 4th Embodiment of this invention.

[First Embodiment]
Hereinafter, an example of the first embodiment of the present invention will be described in detail with reference to the drawings.

As shown in FIG. 1, the object detection device 100 according to the present embodiment includes a far-infrared camera 10 and a computer 20. The far-infrared camera 10 is mounted on a vehicle, has a detection range around the vehicle, acquires far-infrared image data within the detection range according to set measurement parameters and processing parameters, and obtains the acquired far-infrared image data. It is configured to be output to the computer 20 as a video signal, and the measurement parameter and processing parameter settings can be changed from the outside. The far-infrared camera 10 includes a single photodiode or the like corresponding to far-infrared rays, in addition to a general far-infrared camera equipped with an optical system such as a lens and an imaging element having a photosensitive characteristic specialized for far-infrared rays. It may be an arrayed device or the like. Therefore, the far-infrared camera 10 can acquire one-dimensional and two-dimensional far-infrared image data capable of detecting and classifying an object, and has measurement parameters and measurement parameters that affect the acquisition of the far-infrared image data. Includes all devices whose processing parameters can be changed.

As an example, the far-infrared camera 10 has a sensor device including an optical system such as a lens and an image sensor that outputs a signal according to the intensity of light reception, and a so-called image that converts the signal output by the sensor device into an image file. Includes a pretreatment device that is an engine. The measurement parameters of the far-infrared camera 10 include shutter speed (exposure time), gain (sensitivity), angle of view, aperture (F number), focus (focus), frame rate (measurement cycle), and the like with respect to the sensor device.

Processing parameters for the preprocessing device include white balance, contrast, spatial resolution (number of pixels), and the like.

When the sensor device and the preprocessing device are integrally configured, it is synonymous with performing processing such as white balance on the sensor device. Further, since it may be possible to change the frame rate by time-lapse, which is the thinning of frames in the sensor device, it becomes difficult to distinguish between the measurement parameter and the processing parameter. Therefore, the measurement parameter and the processing parameter may be treated without any particular distinction.

The computer 20 includes a CPU, a memory, a non-volatile storage unit that stores programs for executing various processes, a network interface, and the like. Functionally, the computer 20 receives far-infrared image data (hereinafter referred to as “captured image data”) acquired by the far-infrared camera 10 and generates inverted image data of the far-infrared image data. The image preprocessing unit 22 that performs image processing, the object classifier 24 that performs object identification that estimates the type of object in the captured image data and the inverted image data, and the object identification results in each of the captured image data and the inverted image data. It includes an image post-processing unit 26 that outputs an integrated identification result.

The image preprocessing unit 22 is provided with an image memory which is a storage device, stores the captured image data as it is in the image memory, and creates inverted image data in which the gradation of the captured image data is inverted to be stored in the image memory. Remember.

The object classifier 24 identifies an object such as a pedestrian from a background such as a road surface for each of the captured image data and the inverted image data stored in the image memory of the image preprocessing unit 22.

The object classifier 24 has been trained in advance using the training data, and as an example, performs signal processing based on a DNN (Deep Neural Network). Further, as learning data for learning the object classifier 24, photographed image data and inverted image data in which the positions of objects such as pedestrians are assigned in advance are used. Visible light image data may be used as the learning data for learning the object classifier 24.

The image post-processing unit 26 integrates the object identification results for each of the captured image data and the inverted image data of the object classifier 24 and outputs the object identification results. As described later, the integration of the identification results is performed by superimposing the identification result of the captured image data and the identification result of the inverted image data and adding the reliability of the identification associated with each identification result. Based on the numerical values obtained above, the reliability of object identification is determined, and final object identification is performed.

FIG. 2 is a flowchart showing an example of processing of the object detection device 100 according to the present embodiment. It is assumed that the object classifier 24 has been trained in advance by using the photographed image data and the inverted image data in which the positions of objects such as pedestrians are given in advance as the training data.

In step 200, far-infrared image data around the vehicle is acquired by, for example, photographing the traveling path in front of the vehicle with the far-infrared camera 10. The acquired far-infrared image data is stored in the image memory of the image preprocessing unit 22 as captured image data.

In step 204, the image preprocessing unit 22 creates inverted image data in which the contrast of the gradation of the captured image data is inverted and stores it in the image memory.

In step 206, each of the captured image data and the inverted image data stored in the image memory of the image preprocessing unit 22 is input to the object classifier 24, and the object classifier 24 outputs the identification result for each image data. ..

In step 208, the image post-processing unit 26 integrates the identification results of the captured image data and the inverted image data to obtain the final identification result. Then, in step 210, the final identification result obtained in step 208 is output to end the process.

FIG. 3 is an explanatory diagram showing an example of object identification in the object classifier 24 and an example of integration of identification results in the image post-processing unit 26. FIG. 3A shows an example of object identification in the captured image data by the object classifier 24, and FIG. 3B shows an example of object identification in the inverted image data by the object classifier 24. As shown in FIG. 3A, in the object identification in the photographed image data, the car is identified with the identification reliability of 70%, and the person is identified with the identification reliability of 40%. Further, as shown in FIG. 3B, in the object identification in the inverted image data, the car is identified with the identification reliability of 50%, and the person is identified with the identification reliability of 80%.

FIG. 3C is an example of a mapping image of the captured image data created by extracting the identification reliability in the captured image data, and FIG. 3D is created by extracting the identification reliability in the inverted image data. It is the schematic which showed an example of the mapping image of the inverted image data to be done, respectively. The mapping images shown in FIGS. 3 (C) and 3 (D) are created by, for example, the object classifier 24, and the identification reliability is determined in the region corresponding to the position of the object identified in each of the photographed image and the inverted image. Is associated with.

As shown in FIG. 3C, in the object identification in the photographed image data, although the car can be identified with high reliability, the identification reliability of the person is low. Further, as shown in FIG. 3D, in the object identification in the inverted image data, although the person can be identified with high reliability, the identification reliability of the car is low. In the present embodiment, as described above, by integrating the identification results of each of the photographed image data and the inverted image data to obtain the final identification result, an object such as a car or a person is accurately identified from the background.

FIG. 3 (E) is a schematic view showing an example of integration of identification results in each of the captured image data and the inverted image data. In the present embodiment, as an example, by superimposing the mapping images generated from each of the captured image data and the inverted image data, the identification result of the captured image data and the identification result of the inverted image data are superimposed, and each of them is superimposed. The reliability of object identification is determined based on the numerical value obtained by adding the reliability of identification associated with the identification result of. In FIG. 3 (E), the overlapped portion is shown with hatching. In FIG. 3E, the identification reliability of the region shown by hatching is the total of the identification reliability of each of the photographed image data and the inverted image data. The total identification reliability is 120% for cars and 120% for people, indicating that the reliability of object identification is fairly high, and accurately identifies objects such as pedestrians from the background of far-infrared images. can do.

As described above, in the present embodiment, by adding the inverted image data in which the brightness value is inverted in addition to the brightness image data of the normal far-infrared image, the object is only the normal brightness image data. Even when detection is not possible, it is possible to detect an object.

As described above, as a feature of far-infrared images, the brightness of pedestrians and the background (mainly roads) is reversed between summer and winter. Therefore, when the device is trained using the far-infrared image data, it is required to prepare the training data corresponding to such a wide range of temperature environments. However, it takes a huge amount of time, cost, and labor to prepare learning data corresponding to a wide range of temperature environments, and it is considered to be practically difficult.

In the present embodiment, the inverted image data of the captured image data is generated, and both the captured image data and the inverted image data are input to the object classifier 24. The object classifier 24 has been trained with training data regardless of the season, and performs object identification based on the same training data for each of the captured image data and the inverted image data. As described above, in the far-infrared image, the brightness of the pedestrian and the background is reversed in summer and winter, but the captured image data and the inverted image data are each identified as an object based on the same learning data. If there is, there is a possibility that good object identification results can be expected in either the captured image data or the inverted image data. Therefore, in the present embodiment, it is not necessary to relearn the object classifier 24 using different learning data depending on the season.

[Second Embodiment]
Subsequently, the second embodiment of the present invention will be described. FIG. 4 is an explanatory diagram of inversion of the gradation of the captured image data in the present embodiment. This embodiment is different from the first embodiment in the generation of inverted image data from the captured image data by the image preprocessing unit 22, but is the same as the first embodiment in other respects, so that the same configuration has the same reference numerals. A detailed description will be omitted.

In the first embodiment, the image preprocessing unit 22 simply inverts the contrast of the gradation of the captured image data to generate the inverted image data. The straight line shown by the inclination = -1 in FIG. 4 shows the gradation of the captured image data when the contrast of the gradation of the captured image data is simply inverted to generate the inverted image data, and the gradation of the inverted image data. Indicates the correspondence of. When the inclination = -1, for example, the brightest pixel of the captured image data is the darkest pixel of the inverted image data, and the darkest pixel of the captured image data is the brightest pixel of the inverted image data.

In the present embodiment, the inverted image data is generated from the captured image data with characteristics such as a straight line shown by the inclination ≤ -1 or the inclination ≥ -1. When the inverted image data is generated from the captured image data with characteristics such as a straight line indicated by tilt ≤ -1 or tilt ≥ -1, the contrast of the pixels of the inverted image data corresponding to the dark pixels of the captured image data is determined by the captured image. It can be improved as compared with the case where the contrast of the gradation of the data is simply inverted to generate the inverted image data. For example, when the inverted image data is generated from the captured image data with the characteristic of inclination ≤ -1 as shown in FIG. 4, the captured image data is obtained from the luminance value indicated by the intersection of the solid line showing the characteristic and the horizontal axis of FIG. The brightness value of each pixel showing the maximum brightness value is converted into the pixel having the smallest brightness value, that is, the darkest pixel in the inverted image data. As a result, the generated inverted image data shows an image in which the dark part is more remarkable than the inverted image data generated with the characteristic of inclination = -1. Further, when the inverted image data is generated from the captured image data with the characteristic of inclination ≧ -1, the generated inverted image data is an image in which the bright part is more remarkable than the inverted image data generated with the characteristic of inclination = -1. Is shown.

The characteristic of reversing the lightness and darkness of the gradation of the captured image data may be an upwardly convex curve or a downwardly convex curve instead of the straight line shown in FIG. In the present embodiment, the captured image data is converted into the inverted image data based on the correspondence between the predetermined gradation of the captured image data and the gradation of the inverted image data regardless of whether it is linear or curved. You may.

As described above, the identification results of the photographed image and the inverted image may differ depending on the difference between objects such as a car and a person. As shown in FIG. 4, it is possible to improve the accuracy of object identification by changing the characteristic of reversing the brightness of the gradation of the captured image data.

In order to improve the accuracy of object identification, a method such as improving the learning data of the object detection device 100 can be considered, but preparation of new learning data and learning of the object detection device 100 using the image data are complicated and costly. It takes. However, it is relatively easy to change the characteristic of reversing the brightness of the gradation of the captured image data as shown in FIG. Therefore, according to the present embodiment, the accuracy of object identification can be easily improved as compared with the case where the device is relearned with new learning data.

[Third Embodiment]
Subsequently, a third embodiment of the present invention will be described. FIG. 5 is a block diagram showing an example of the object detection device 200 according to the present embodiment. The present embodiment differs from the first embodiment in that it has a visible light camera 12 that acquires a visible light image and a computer 120 that includes an object classifier 124 that has been trained using the visible light image. Is the same as that of the first embodiment, so that the same configuration is designated by the same reference numerals and detailed description thereof will be omitted.

The image preprocessing unit 122 generates inverted image data from the captured image data acquired by the far-infrared camera 10, and the generated inverted image data and captured image data are input to the object classifier 124. It is the same as the embodiment. In the present embodiment, not only the captured image data acquired by the far-infrared camera and the inverted image data generated from the captured image data, but also the visible light image data acquired by the visible light camera 12 is input to the object classifier 124. Will be done.

The object classifier 124 describes each of the captured image data of the far-infrared camera 10 input from the image preprocessing unit 122, the inverted image data of the far-infrared image, and the visible light image data input from the visible light camera 12. Perform object identification.

The object classifier 124 has been trained in advance using the training data, and as an example, performs signal processing based on DNN. Further, as the learning data for learning the object classifier 124, visible light image data to which the position of an object such as a pedestrian is assigned in advance is used. Further, as the learning data, photographed image data and inverted image data may be used.

Then, the image post-processing unit 126 superimposes the identification result of the captured image data, the identification result of the inverted image data, and the identification result of the visible light image data, and the reliability of the identification associated with each identification result. The reliability of object identification is determined based on the numerical value obtained by adding the degrees.

Generally, in a recognition system for automatic driving of a vehicle, an object classifier is learned using visible light image data. If the object classifier learned using visible light image data can be used for recognition of far-infrared image data, there is no need to prepare a new object recognizer specialized for far-infrared image data, and the cost is high. It is advantageous. With the object recognizer learned from the visible light image data, it is not possible to predict which of the far-infrared image captured image data and the inverted image data is more suitable for recognition, but the far-infrared image captured image data and the inverted image data By integrating each identification result with the identification result in the visible light image data and determining the reliability of the final object identification, object identification can be performed even with an object classifier learned from the visible light image data. It becomes possible to do it with high accuracy. Further, according to the present embodiment, by object identification using far-infrared image data, an object such as a pedestrian at night or an object such as a pedestrian existing in a distant place can be identified by object identification in visible light image data. Can be identified more clearly.

[Fourth Embodiment]
Subsequently, a fourth embodiment of the present invention will be described. FIG. 6 is a functional block diagram showing an example of the object detection device according to the present embodiment. In the present embodiment, in addition to the captured image data and the inverted image data, the inverted image data, for example, the image data of a rectangular region having a predetermined number of pixels cut out from the center of the image is input to the object classifier 224 to identify the object. ..

There is a high possibility that an object far from the far-infrared camera 10 or an object having a small size exists in the central portion of the captured image data and the inverted image data. In the present embodiment, the image data cut out from the rectangular region near the center of the image of the inverted image data as shown in FIG. 8 is also used for object identification, so that an object existing far from the far infrared camera 10 or an object or an object existing far from the far infrared camera 10 or It is possible to accurately identify small objects such as pedestrians.

In the present embodiment, each of the identification result (1) based on the captured image data, the identification result (2) based on the inverted image data, and the identification result (3) based on the image data in the rectangular region of the inverted image data. , The classification of the object, the position of the object, and the likelihood of identification of the object are calculated, and the identification results (1), (2), and (3) are integrated to perform the final object identification.

Each of the captured image data, the inverted image data, and the image data of the rectangular region of the inverted image data includes an overlapping region. The accuracy of object identification can be ensured by comprehensively determining the identification result based on each image in the overlapping region.

In the present embodiment, a pedestrian existing on the roadway is detected by detecting an object from the image data cut out from the rectangular area near the center of the image of the inverted image data as the image data of the area corresponding to the traveling direction of the own vehicle. Etc. can be detected with higher accuracy.

As described above, in the first embodiment, the brightness value of the captured image data is inverted to generate the anti-conversion image data, and in the second embodiment, the gradation of the captured image data and the inverted image are predetermined. Inverted image data was generated from the captured image data based on the correspondence with the gradation of the data. However, the present invention is not limited to these. For example, preprocessing such as contrast enhancement and contour extraction may be performed on each of the captured image data and the inverted image data, or in addition to such preprocessing, other known image processing methods may be combined. Further, the resolutions of the captured image data and the inverted image data may be enlarged by super-resolution processing or the like.

In the above, the mode in which the neural network model as an example of the model is trained by deep learning has been described, but the present invention is not limited to this. For example, another model different from the neural network model may be trained by another learning method different from deep learning.

10 Far-infrared camera 12 Visible light camera 20 Computer 22 Image pre-processing unit 24 Object classifier 26 Image post-processing unit 100 Object detection device 120 Computer 122 Image pre-processing unit 124 Object classifier 126 Image post-processing unit 200 Object detection device

Claims (8)

An imaging unit that acquires far-infrared image data of the surrounding environment,
An image processing unit that generates converted image data obtained by gradation-converting the far-infrared image data acquired by the photographing unit, and an image processing unit.
By performing object identification processing on each of the far-infrared image data and the converted image data based on the result learned in advance using the training data, the far-infrared image data and the converted image data are each subjected to the object identification process. An object identification unit that executes an identification process that outputs an identification result including information on identification of an object type and accuracy information indicating the reliability of the identification in each of the far-infrared image data and the converted image data.
An object determination unit that finally identifies an object in the far-infrared image data based on the identification result by the object identification unit, and an object determination unit.
Object detector including. The object detection device according to claim 1, wherein the image processing unit inverts the brightness value of the far-infrared image data acquired by the photographing unit to generate the converted image data. The image processing unit is requested to generate the converted image data from the far-infrared image data based on a predetermined correspondence between the gradation of the far-infrared image data and the gradation of the converted image data. Item 2. The object detection device according to item 1. The object identification unit identifies the position of the object in each of the far-infrared image data and the converted image data, and outputs the identification result including the position information of the identified object, and the object determination unit outputs the identification result including the position information of the identified object. The invention according to any one of claims 1 to 3, wherein the final object identification is performed based on the position information of the object and the accuracy information in each of the far-infrared image data and the converted image data included in the identification result. Object detector. The object identification unit identifies the position of the object in each of the far-infrared image data and the converted image data, and identifies the accuracy information in each of the far-infrared image data and the converted image data. A mapping image associated with the position of the object in each of the infrared image data and the converted image data is generated as the identification result.
The object determination unit superimposes mapping images generated from each of the far-infrared image data and the converted image data to obtain an identification result in the far-infrared image data and an identification result in the converted image data. The object detection device according to any one of claims 1 to 3, wherein the object detection device is integrated to perform final object identification. The learning data according to any one of claims 1 to 5, which is at least one of the far-infrared image data for learning, the converted image data for learning, and visible light image data for learning. Object detector. The photographing unit further acquires visible light image data of the surrounding environment, and obtains the visible light image data.
The object identification unit includes information on identifying the type of object in each of the far-infrared image data, the converted image data, and the visible light image data, and the far-infrared image data, the converted image data, and the visible light. An identification process is executed to output an identification result including accuracy information indicating the reliability of the identification in each of the image data.
The object detection device according to any one of claims 1 to 3, wherein the object determination unit finally identifies an object in the far-infrared image data based on the identification result by the object identification unit. The object detection device according to claim 7, wherein the learning data is the visible light image data for learning.

Images