JP2019215755A - Image recognition device, image recognition method, machine learning model providing device, machine learning model providing method, machine learning model generating method, and machine learning model device - Google Patents

Abstract

To provide an image recognition device and an image recognition method which enable accurate image recognition of a recognition object irrespective of a relative speed between an imaging apparatus that images the recognition object and the recognition object.SOLUTION: An image recognition device includes an acquisition part, an identification part, a storage part and a determination part. The acquisition part acquires from an imaging apparatus, an image including an imaged recognition object. The identification part identifies a relative speed between the imaging apparatus and the recognition object. The storage part stores a machine learning model that derives from the image including the imaged recognition object, recognition result that a subject in the image is the recognition object, for each of speed ranges. The determination part adopts a machine learning model in a speed range according to the relative speed identified by the identification part, out of the plurality of machine learning models to determine the subject in the image.SELECTED DRAWING: Figure 1

Description

  The disclosed embodiments relate to an image recognition device, an image recognition method, a machine learning model providing device, a machine learning model providing method, a machine learning model generation method, and a machine learning model device.

  Conventionally, image data used for image recognition processing is collected from a plurality of image recognition devices, machine learning about image recognition processing is performed using the stored image data as learning data, and parameters used for image recognition processing are updated. There is a system for providing the information to an image recognition device.

JP-A-2013-135552

  However, the conventional image recognition device may not be able to accurately recognize the recognition target object depending on the relative speed between the imaging device that captures the recognition target object and the recognition target object.

  An aspect of the embodiment is made in view of the above, and can accurately recognize an image of a recognition target object regardless of the relative speed between the imaging device that captures the recognition target object and the recognition target object. It is an object to provide an image recognition device, an image recognition method, a machine learning model providing device, a machine learning model providing method, a machine learning model generating method, and a machine learning model device.

  An image recognition device according to an aspect of an embodiment includes an acquisition unit, a determination unit, a storage unit, and a determination unit. The acquisition unit acquires an image of the recognition target object captured from the imaging device. The determining unit determines a relative speed between the imaging device and the recognition target. The storage unit stores, for each speed range, a machine learning model that derives a recognition result indicating that a subject in the image is the recognition target object from an image of the recognition target object. The determining unit determines a subject in the image by using the machine learning model in the speed range corresponding to the relative speed determined by the determining unit from the plurality of machine learning models.

  An image recognition device, an image recognition method, a machine learning model providing device, a machine learning model providing method, a machine learning model generation method, and a machine learning model device according to one embodiment of the present invention are provided with an image capturing device that captures an object to be recognized and a recognition device. The image of the recognition target can be accurately recognized regardless of the relative speed with respect to the target.

FIG. 1 is an explanatory diagram illustrating an outline of an image recognition method according to the embodiment. FIG. 2 is a block diagram illustrating an example of a configuration of the image recognition device according to the embodiment. FIG. 3A is an explanatory diagram illustrating a procedure for generating a machine learning model according to the embodiment. FIG. 3B is an explanatory diagram illustrating a procedure for generating a machine learning model according to the embodiment. FIG. 3C is an explanatory diagram illustrating a procedure for generating a machine learning model according to the embodiment. FIG. 4 is an explanatory diagram illustrating an example of an image recognition process according to the embodiment. FIG. 5 is a flowchart illustrating an example of a process executed by the control unit of the image recognition device according to the embodiment.

  Hereinafter, embodiments of an image recognition device, an image recognition method, a machine learning model providing device, a machine learning model providing method, a machine learning model generating method, and a machine learning model device will be described in detail with reference to the accompanying drawings. In addition, this invention is not limited by embodiment shown below. Hereinafter, an image recognition device and an image recognition method for recognizing a subject appearing in a captured image of the periphery of the vehicle, which is acquired from an imaging device mounted on the vehicle, will be described as an example.

  FIG. 1 is an explanatory diagram illustrating an outline of an image recognition method according to the embodiment. The image recognition device 1 according to the embodiment acquires an image in which the periphery of a vehicle is imaged from an imaging device mounted on the vehicle. Here, for example, as shown in FIG. 1A, the image recognition device 1 is configured to travel around a vehicle equipped with the imaging device and the image recognition device 1 (hereinafter referred to as “own vehicle”). The case where the image 101 of the vehicle 100 is obtained will be described as an example.

  When an image of a recognition target that moves like another vehicle 100 is captured, the image 101 may show another vehicle 100 having a shape slightly different from the shape of the other vehicle 100 as seen by a human. For example, when an image of another vehicle 100 moving at a high speed is captured by an imaging device having a relatively slow shutter speed, blurring occurs in the image of the other vehicle 100 in the image 101 as shown in FIG. There is.

  Further, when an image capturing apparatus having a relatively slow shutter speed captures an image of a subject moving at a high speed in the lateral direction, the subject may appear in the image as a shape extending in the horizontal direction more than the actual shape.

  Specifically, the imaging device sequentially drives (exposes) the imaging elements arranged in a row above the imaging region in which a plurality of imaging elements are arranged in a matrix to capture an image of one frame. For this reason, the subject imaged in the upper part of the imaging region has already moved in the horizontal direction when the image is captured in the lower part of the imaging region.

  As a result, a so-called rolling shutter phenomenon (hereinafter, referred to as a rolling phenomenon) occurs in the image, in which a subject having a shape extending in a lateral direction more than the actual shape is photographed. The degree of the rolling phenomenon and the magnitude of the shake vary depending on the difference in the relative speed between the imaging device and the subject.

  For this reason, for example, in order to generate a machine learning model that derives an image recognition result that the subject is another vehicle 100 from an image 101 in which another vehicle 100 is captured, the degree of the rolling phenomenon and the magnitude of blurring must be reduced. Machine learning using a huge number of different images as teaching materials is required.

  Even if a machine learning model is generated by machine learning using an enormous number of images as teaching materials, an accurate image recognition result that the subject is another vehicle 100 is obtained from the image 101 in which the other vehicle 100 is captured. In some cases, it cannot be derived.

  For example, when recognizing a vehicle in an image in which a rolling phenomenon or blurring has occurred using a machine learning model, even if an image similar to the image is included in the learning material for the machine learning, the learning material includes the rolling phenomenon or the like. Many images with different degrees of blur are also included.

  For this reason, in the determination using the machine learning model, the likelihood of the determination result that the subject in the image is a vehicle is low, the subject in the image is not determined to be a vehicle, and an accurate image recognition result cannot be derived. There are cases.

  Therefore, when acquiring the image 101, the image recognition device 1 according to the embodiment first determines the relative speed between the imaging device and the recognition target (the other vehicle 100 in this case). Then, for example, the image recognition device 1 determines which of the first speed range, the second speed range, and the third speed range the relative speed is included in.

  Here, it is assumed that the first speed range is the slowest speed range, the second speed range is faster than the first speed range, and is slower than the third speed range. The image recognition device 1 determines, for example, a moving speed of a subject in an image as a relative speed. Note that the method of determining the moving speed is not limited to this.

  At this time, when the image recognition device 1 determines that the relative speed is included in the second speed range, for example, as illustrated in FIG. 1B, the image recognition device 1 performs the image recognition process illustrated in FIG. Specifically, the image recognition device 1 includes, for example, a first speed range machine learning model 1-1, a second speed range machine learning model 1-2, and a third speed range machine learning model 1-3. It has three machine learning models.

  The first speed range machine learning model 1-1 is, for example, a machine learning model specialized for image recognition of a vehicle whose relative speed to a subject is included in the first speed range. The second speed range machine learning model 1-2 is, for example, a machine learning model specialized for image recognition of a vehicle whose relative speed to the subject is included in the second speed range.

  The third speed range machine learning model 1-3 is, for example, a machine learning model specialized for image recognition of a vehicle whose relative speed to the subject is included in the third speed range. Note that the number of machine learning models provided in the image recognition device 1 is not limited to three as long as it is plural.

  Further, hereinafter, when an unspecified machine learning model among three machine learning models is indicated, it is simply described as a machine learning model. An example of a procedure for generating these three machine learning models will be described later with reference to FIGS. 3A, 3B, and 3C.

  As described above, when the image recognition device 1 determines that the relative speed with respect to the subject is included in the second speed range, the image recognition device 1 sends the image to the second speed range machine learning model 1-2 among the three machine learning models. Enter 101. Thus, the image recognition device 1 can derive an accurate image recognition result 102 indicating that the subject is another vehicle 100.

  Next, an example of a configuration of the image recognition device 1 according to the embodiment will be described with reference to FIG. FIG. 2 is a block diagram illustrating an example of a configuration of the image recognition device 1 according to the embodiment. As shown in FIG. 2, the image recognition device 1 is connected to an imaging device 41, a radar 42, a steering angle sensor 43, and a vehicle control device 44.

  Note that the image collection by relative speed 110 is an image collection including a plurality of images in which the recognition target object is captured. Each image included in the relative speed-specific image collection 110 is provided with information indicating the type of the subject shown in the image and the relative speed between the subject and the imaging device at the time of imaging.

  The image collection 110 for each relative speed is used by the image recognition apparatus 1 as a teaching material of supervised machine learning of the image recognition processing. An example of the relative speed-specific image collection 110 will be described later with reference to FIGS. 3A, 3B, and 3C, together with a description of a procedure for generating a machine learning model.

  The imaging device 41 is, for example, a vehicle-mounted camera that is installed at the front of a vehicle on which the image recognition device 1 is mounted and captures an image in front of the vehicle. Note that the imaging device 41 may be a vehicle-mounted camera that is provided on a side portion or a rear portion of the vehicle and captures an image around the vehicle. The imaging device 41 outputs the captured image to the image recognition device 1.

  The radar 42 receives, for example, a reflected wave of the transmission wave reflected on the target by radiating a millimeter wave transmission wave to the surroundings of the vehicle, and based on the transmission wave and the reception wave, a distance from the vehicle to the target, This is a millimeter wave radar that detects the angle at which the target exists and the relative speed of the target with respect to the vehicle. The radar 42 outputs the detected relative speed to the image recognition device 1. The steering angle sensor 43 is a sensor that detects a steering angle of the vehicle. The steering angle sensor 43 outputs the detected steering angle to the image recognition device 1.

  The vehicle control device 44 is, for example, an ECU (Electronic Control Unit) that performs overall control of the entire vehicle. The vehicle control device 44 performs speed control, steering control, and braking control of the vehicle, for example, when the image recognition device 1 recognizes an image of a subject that interferes with the traveling of the vehicle, thereby causing an obstacle to the vehicle. Avoid objects.

  Further, the image recognition device 1 includes a control unit 2 and a storage unit 3. The storage unit 3 is, for example, an information storage device such as a data flash. The storage unit 3 stores the first speed range machine learning model 1-1, the second speed range machine learning model 1-2,... The nth speed range machine learning model 1-n (n is a natural number of 3 or more) ) Are stored.

  The control unit 2 includes a microcomputer having a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and various circuits.

  The control unit 2 includes a learning unit 21, an acquisition unit 22, a determination unit 23, and a determination unit 24 that function by executing a program stored in the ROM by the CPU using the RAM as a work area. The control unit 2 may be configured by hardware such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

  The learning unit 21, the acquisition unit 22, the determination unit 23, and the determination unit 24 included in the control unit 2 each realize or execute an operation of information processing described below. The internal configuration of the control unit 2 is not limited to the configuration illustrated in FIG. 2 and may be another configuration as long as the configuration performs information processing described later.

  The learning unit 21 acquires a plurality (for example, thousands) of images for each relative speed range and for each type of subject from the relative speed-specific image collection 110. Then, the learning unit 21 uses the acquired image as a teaching material for supervised machine learning, and uses a machine learning model (for example, a first speed range for image recognition of a vehicle) for each relative speed range and for each type of subject. A machine learning model 1-1) is generated and stored in the storage unit 3.

  Here, an example of a procedure for generating a machine learning model according to the embodiment will be described with reference to FIGS. 3A, 3B, and 3C. FIG. 3A, FIG. 3B, and FIG. 3C are explanatory diagrams illustrating a procedure for generating a machine learning model according to the embodiment.

  Here, a case will be described in which a machine learning model for recognizing a vehicle included in an image is generated. The learning unit 21 can also generate a machine learning model for image recognition of a motorcycle, a bicycle, a pedestrian, and the like, in addition to the vehicle.

  As illustrated in FIG. 3A, when the learning unit 21 generates the first speed range machine learning model 1-1 in which the relative speed between the subject and the imaging device 41 is the slowest, the learning unit 21 determines the first speed from the image collection 110 for each relative speed. The images 11-1 and 11-2 of the range image collection 10-1 are acquired.

  Since the images included in the first speed range image collection 10-1 have the slowest relative speed between the subject and the imaging device 41, the rolling phenomenon and blurring occur as in the images 11-1 and 11-2. Absent.

  The learning unit 21 machine-learns the features of these images, so that when an image of a vehicle whose relative speed with respect to the imaging device falls within the first speed range is input, the image that the subject is a vehicle is input. A first speed range machine learning model 1-1 for outputting a recognition result is generated.

  Further, as illustrated in FIG. 3B, the learning unit 21 generates the second speed range machine learning model 1-2 in which the relative speed between the subject and the imaging device 41 is the next highest after the first interval range. The images 12-1 and 12-2 of the second speed range image collection 10-2 are acquired from the speed-specific image collection 110.

  The images included in the second speed range image collection 10-2 are slightly different, such as the images 12-1, 12-2, etc., because the relative speed between the subject and the imaging device 41 is faster than the first speed range. However, rolling phenomenon and blurring have occurred.

  The learning unit 21 machine-learns the features of these images, so that when an image of a vehicle whose relative speed to the imaging device is included in the second speed range is input, the image that the subject is a vehicle is input. A second speed range machine learning model 1-2 that outputs a recognition result is generated.

  Further, as illustrated in FIG. 3C, when the learning unit 21 generates the n-th speed range machine learning model 1-n in which the relative speed between the subject and the imaging device 41 is the fastest, the learning unit 21 uses the The images 13-1 and 13-2 of the n-speed range image collection 10-n are acquired.

  Since the images included in the n-th speed range image collection 10-n have the highest relative speed between the subject and the imaging device 41, large blurring and a remarkable rolling phenomenon as in the images 13-1 and 13-2 and the like occur. It has occurred.

  The learning unit 21 machine-learns the features of these images, so that when an image of a vehicle whose relative speed with respect to the imaging device is included in the n-th speed range is input, the image that the subject is a vehicle is obtained. An n-th speed range machine learning model 1-n for outputting a recognition result is generated.

  As described above, the learning unit 21 performs, for each of the speed ranges, an authentication target of a known type that moves at a relative speed within each of the speed ranges input for each of the speed ranges. Generate a machine learning model. As a result, each speed range machine learning model does not reflect the influence of the image of the authentication target moving at a relative speed that is not included in each speed range.

  The learning unit 21 does not need to identify and learn the speed of the subject from the image serving as the teaching material by machine learning the image for each speed range, so that the number of images used as the teaching material can be significantly reduced. it can.

  The above-described procedure for generating the machine learning model has been described assuming that the frame rate (shutter speed) is fixed. However, in the case of the imaging device 41 having a variable shutter speed, each speed in the case of performing machine learning for each speed range is described. Normalize the range according to the shutter speed.

  For example, a certain frame rate is set as a reference frame rate, and when the frame rate of the imaging device 41 is the reference frame rate, a speed range obtained by multiplying a reference speed range corresponding to the reference frame rate by 1 as a coefficient is set. .

  If the frame rate is other than the reference frame rate, the speed range is obtained by multiplying the reference speed range by a coefficient (reference frame rate / imaging frame rate). Accordingly, the learning unit 21 can generate a machine learning model of an image for each appropriate speed range according to the frame rate set in the imaging device 41.

  In an image used as a learning material for machine learning, the moving speed of a recognition target that moves in the image varies depending on the zoom magnification. For this reason, it is desirable to normalize the zoom magnification of the image used for the teaching material so that the distance from the host vehicle to the recognition target in the image becomes the reference distance.

  Similarly, for an image actually recognized by the image recognition device 1 (the determination unit 24 described later), the zoom magnification is normalized such that the distance from the vehicle to the recognition target in the image becomes the reference distance. It is desirable to keep. The distance to the recognition target used when normalizing the zoom of the image can be acquired from the radar 42.

  Returning to FIG. 2, the description of the processing units other than the learning unit 21 included in the control unit 2 proceeds. The acquisition unit 22 sequentially acquires images of the surroundings of the vehicle from the imaging device 41 and outputs the images to the determination unit 23. The determination unit 23 determines the relative speed between the subject of the image input from the acquisition unit 22 and the imaging device 41.

  The determination unit 23 determines the moving speed of the subject in the image as the relative speed between the subject and the imaging device 41. Accordingly, the determination unit 23 can determine the relative speed that directly affects the blurring and the rolling phenomenon of the subject, rather than the relative speed based on the physical positional relationship between the subject and the imaging device 41.

  The determination unit 23 determines, for example, the moving speed of the position of the subject between the images of a plurality of frames continuously captured by the imaging device 41 as the relative speed between the subject and the imaging device 41. Thereby, the determination unit 23 can determine an accurate relative speed between the subject and the imaging device 41 in the image.

  At this time, in the image input from the imaging device 41, the moving speed of the recognition target moving in the image varies depending on the zoom magnification. For this reason, it is desirable that the determination unit 23 normalizes the zoom magnification of the image input from the imaging device 41 so that the distance from the host vehicle to the recognition target in the image becomes the reference distance. The determination unit 23 can acquire the distance to the recognition target used when normalizing the zoom of the image from the radar 42.

  Further, the position where the recognition target appears between consecutive frames also changes depending on the frame rate (shutter speed) of the imaging device 41. Therefore, when the frame rate of the imaging device 41 is variable, it is preferable that the determination unit 23 normalizes the moving speed of the authentication target according to the frame rate set in the imaging device 41.

  The determination unit 23 can also determine, for example, the relative speed between the vehicle and the subject input from the radar 42 as the relative speed between the subject and the imaging device 41. Accordingly, the determination unit 23 can determine the relative speed between the subject and the imaging device 41 without performing the process of determining the relative speed.

  However, if the determination unit 23 determines the relative speed input from the radar 42 as the relative speed between the subject and the imaging device 41, the determination unit 23 may not be able to determine the accurate moving speed of the subject in the image. For example, when the vehicle makes a right or left turn at an intersection, the imaging direction of the imaging device 41 changes suddenly, so that the subject may move in the image at a moving speed higher than the relative speed between the subject and the imaging device 41.

  Therefore, when determining the relative speed input from the radar 42 as the relative speed between the subject and the imaging device 41, the determination unit 23 receives the input from the radar 42 based on the speed of change in the imaging direction by the imaging device 41. Correct the relative speed.

  For example, the determination unit 23 estimates a change speed in the traveling direction of the vehicle, that is, a change speed in the imaging direction based on the steering angle of the vehicle input from the steering angle sensor 43, and determines the change speed in the imaging direction. Based on this, the relative speed input from the radar 42 is corrected.

  Accordingly, the determination unit 23 can determine the moving speed close to the accurate moving speed of the subject in the image as the relative speed between the subject and the imaging device 41. Then, the determination unit 23 outputs the determined relative speed between the subject and the imaging device 41 and the image input from the imaging device 41 to the determination unit 24.

  When recognizing the subject in the image input from the determination unit 23, the determination unit 24 includes a speed range including the relative speed input from the determination unit 23 among the plurality of machine learning models stored in the storage unit 3. The subject is determined by employing a machine learning model for use.

  Here, an example of the image recognition processing by the determination unit 24 will be described with reference to FIG. FIG. 4 is an explanatory diagram illustrating an example of an image recognition process according to the embodiment. As illustrated in FIG. 4A, for example, when the image 104 including the image 103 of the blurred vehicle is input from the determination unit 23, the determination unit 24 divides the image 104 into a plurality of divided regions. .

  Then, when the relative speed input from the determination unit 23 is included in the second speed range, the determination unit 24 generates the second speed range machine learning model 1-2 as shown in FIG. adopt. The second speed range machine learning model 1-2 conceptually includes a processing layer having a three-layer structure of an input layer 2a, an intermediate layer 2b, and an output layer 2c, as shown in FIG. 4B. Although FIG. 4A shows a case where the intermediate layer 2b is a single layer, a plurality of intermediate layers 2b may be provided.

  The input layer 2a, the intermediate layer 2b, and the output layer 2c include a plurality of nodes 2z that determine in parallel whether or not the image of each divided region is a part of the vehicle. Each node 2z of the input layer 2a is connected to all nodes 2z of the intermediate layer 2b. Each node 2z of the intermediate layer 2b is connected to all nodes 2z of the output layer 2c. Thus, the second speed range machine learning model 1-2 has a neural network structure in which the nodes 2z are connected to each other.

  The determination unit 24 inputs the image of each divided region of the image 104 to each node 2z of the input layer 2a. Each node 2z of the input layer 2a determines whether or not the sequentially input image of each divided region is a part of the vehicle, and intermediately determines the weighted determination result as the possibility that the image is a part of the vehicle is higher. Output to the node 2z of the layer 2b.

  Similarly, each node 2z of the intermediate layer 2b also determines whether or not the image of each divided region is a part of the vehicle, and outputs a higher weighted determination result as the possibility of being a part of the vehicle is higher than the output layer 2c. To the node 2z.

  Each node 2z of the output layer 2c determines whether or not the image of each divided region is a part of the vehicle, and outputs a determination result 2d for each divided region. FIG. 4B shows a determination result 2d of the divided region in the fourth row from the bottom in the image 104.

  In the example illustrated in FIG. 4B, the second speed range machine learning model 1-2 uses the vehicle (one of the vehicles) for the divided region on the fourth line from the bottom in the image 104 other than the images of the divided regions on the left and right ends. Part).

  The second speed range machine learning model 1-2 determines whether or not the image of all the divided regions in the image 104 is a part of the vehicle. The determination unit 24 derives an image recognition result 102 indicating that the subject is a vehicle, as shown in FIG. 4C, based on the determination result 2d for each divided region.

  At this time, since the second speed range machine learning model 1-2 is generated by machine learning using only an image of a subject whose relative speed with respect to the imaging device 41 is within the second speed range as a learning material, Can derive an accurate image recognition result indicating that the vehicle is a vehicle.

  Thereafter, the determination unit 24 determines whether the vehicle whose image has been recognized interferes with the traveling of the own vehicle. The determining unit 24 determines whether or not the traveling of the own vehicle is hindered based on the position of the vehicle with respect to the own vehicle whose image has been recognized.

  The determination unit 24 determines whether or not the subject interferes with the traveling of the own vehicle when the subject such as a pedestrian or a falling object on a road surface is image-recognized in addition to the vehicle. Then, when the determination unit 24 determines that the image-recognized object interferes with the traveling of the own vehicle, the determination unit 24 outputs information indicating the position of the image-recognized object to the vehicle control device 44.

  When the information indicating the position of the image-recognized one is input from the determination unit 24, the vehicle control device 44 performs the speed control, the steering control, and the braking control of the vehicle to make the own vehicle avoid danger.

  Next, an example of a process performed by the control unit 2 of the image recognition device 1 according to the embodiment will be described with reference to FIG. FIG. 5 is a flowchart illustrating an example of a process executed by the control unit 2 of the image recognition device 1 according to the embodiment.

  The control unit 2 executes the process shown in FIG. 5 every time an image around the vehicle is captured by the imaging device 41. Specifically, when an image around the vehicle is captured by the imaging device 41, the control unit 2 first acquires an image from the imaging device 41 (step S101).

  Subsequently, the control unit 2 determines a relative speed between the imaging device 41 and the recognition target (Step S102). Thereafter, the control unit 2 selects a machine learning model in a speed range including the determined relative speed (Step S103).

  Then, the control unit 2 inputs an image to the selected machine learning model (step S104), and determines a subject based on the output of the machine learning model (step S105). Subsequently, the control unit 2 determines whether or not the notification to the vehicle control device 44 is necessary (Step S106).

  At this time, based on the position of the subject with respect to the vehicle, the type of the subject, and the like, the control unit 2 determines that the notification is necessary when it is determined that the subject interferes with the traveling of the own vehicle, and determines that there is no obstacle. When determining, it is determined that notification is not necessary.

  Then, when the control unit 2 determines that the notification is not necessary (Step S106, No), the process ends. When it is determined that the notification is necessary (Step S106, Yes), the control unit 2 notifies the vehicle control device 44 of the position of the subject with respect to the vehicle (Step S107), and ends the process.

  In the above-described embodiment, the image recognition apparatus and the image recognition method for recognizing a subject appearing in a captured image of the periphery of the vehicle acquired from the imaging device mounted on the vehicle have been described as examples. Is an example.

  For example, the functions of the control unit 2 and the storage unit 3 included in the image recognition device 1 described above may be provided in an information providing device provided in a center that can wirelessly communicate with a plurality of vehicles. In such a case, the information providing apparatus acquires images for each relative speed from a plurality of vehicles and performs machine learning, and creates and stores the first to n-th speed range machine learning models 1-1 to 1-n.

  Note that the identification data of each speed range machine learning model is added to the first to n-th speed range machine learning models 1-1 to 1-n by, for example, a method of adding the data learning section to a data header section, and image recognition is performed. It is used for selecting a machine learning model for each speed range in image recognition processing by a processing device or the like.

  The information providing apparatus, when receiving an image for determination captured from each vehicle, by the same method as the image recognition apparatus 1 described above, uses the first to n-th speed range machine learning models 1-1 to 1-1- The recognition target is determined using the machine learning model selected from n. Then, the information providing device returns the determination result of the recognition target to the vehicle that is the transmission source of the image.

  Further, the information providing device may be configured to include the functions of the learning unit 21 and the storage unit 3 included in the image recognition device 1 described above. In the case of such a configuration, the information providing apparatus obtains images according to relative speeds from a plurality of vehicles and performs machine learning, creates and stores machine learning models 1-1 to 1-n for the first to n-th speed ranges. I do.

  The information providing apparatus is configured to provide a machine learning model having a speed range corresponding to the speed information designated by the image recognition device provided in the vehicle and determining a subject in the image to the first to n-th speed range machine learning models 1. -1 to 1-n to provide to the vehicle image recognition device. According to such an information providing device, the processing load on the vehicle can be significantly reduced.

  In addition, the image recognition device and the image recognition method according to the embodiment are, for example, captured in an image captured at an arbitrary location by an arbitrary imaging device, such as a camera or a security camera that captures an image of a product conveyed by a belt conveyor. It can be applied to image recognition of a subject.

  Further, in the above-described embodiment, the case where the image recognition device 1 includes the learning unit 21 has been described. However, the image recognition device according to the embodiment does not necessarily need to include the learning unit 21. The image recognition device includes a first speed range machine learning model 1-1, a second speed range machine learning model 1-2,... An nth speed range machine learning model 1 generated by another machine learning device. By storing n, the above-described image recognition processing can be performed.

  Further effects and modifications can be easily derived by those skilled in the art. Thus, the broader aspects of the present invention are not limited to the specific details and representative embodiments shown and described above. Accordingly, various modifications can be made without departing from the spirit or scope of the general inventive concept as defined by the appended claims and their equivalents.

DESCRIPTION OF SYMBOLS 1 Image recognition apparatus 2 Control part 21 Learning part 22 Acquisition part 23 Judgment part 24 Judgment part 3 Storage part 1-1 Machine learning model for 1st speed range 1-2 Machine learning model for 2nd speed range 1-3 Third speed Range machine learning model 41 Imaging device 42 Radar 43 Steering angle sensor 44 Vehicle control device 110 Image collection by relative speed

Claims (10)

An acquisition unit configured to acquire an image in which the recognition target is imaged from the imaging device;
A determining unit that determines a relative speed between the imaging device and the recognition target object,
A storage unit that stores, for each speed range, a machine learning model that derives a recognition result that a subject in the image is the recognition target object from an image in which the recognition target object is captured,
A determination unit that determines a subject in the image by using the machine learning model in the speed range according to the relative speed determined by the determination unit from the plurality of machine learning models. Image recognition device. The determination unit includes:
The image recognition device according to claim 1, wherein the relative speed is determined based on a moving speed of the subject in the image. The determination unit includes:
The image recognition device according to claim 2, wherein the relative speed is determined based on a moving speed of the position of the subject between a plurality of frames of images continuously captured by the imaging device. The determination unit includes:
The image recognition device according to claim 1, wherein the relative speed is corrected based on a change speed of an imaging direction by the imaging device. The machine learning model for each speed range based on a plurality of images in which the type of recognition target moving at the relative speed within each speed range input for each relative speed range is imaged. The image recognition device according to claim 1, further comprising: a learning unit configured to generate the image. An acquisition step of acquiring an image in which the recognition target is imaged from the imaging device,
A determining step of determining a relative speed between the imaging device and the recognition target;
A storage step of storing, for each speed range, a machine learning model that derives a recognition result that a subject in the image is the recognition target object from an image in which the recognition target object is captured,
Determining a subject in the image by employing the machine learning model in the speed range according to the relative speed determined in the determining step from the plurality of machine learning models. Image recognition method. A storage unit that stores, for each speed range, a machine learning model that derives a recognition result that a subject in the image is the recognition target object from an image in which the recognition target object is captured by the imaging device.
A providing unit that selects the machine learning model in the speed range according to the speed information designated by the image recognition device that determines a subject in an image from a plurality of the machine learning models and provides the selected machine learning model to the image recognition device. An apparatus for providing a machine learning model, comprising: A storage step of storing, for each speed range, a machine learning model that derives a recognition result that a subject in the image is the recognition target from an image of the recognition target captured by the imaging device.
Providing the selected machine learning model in the speed range according to the speed information specified by the image recognition device that determines the subject in the image from a plurality of the machine learning models and providing the selected machine learning model to the image recognition device. A method for providing a machine learning model, the method comprising: A method for generating a machine learning model used by an image recognition device for image recognition processing,
A model generation step of generating a machine learning model for each speed range by learning an image of a moving object moving at a speed of each speed range in a plurality of speed ranges,
An identification data adding step of adding identification data for selecting and using the machine learning model for each speed range,
A machine learning model generation method, comprising: An image recognition device is a machine learning model device used for image recognition processing,
Machine learning model unit for each of a plurality of speed ranges generated by learning the image of the moving body moving at the speed of each speed range in the plurality of speed ranges,
An identification data unit added to the speed range machine learning model unit to select and use the speed range machine learning model unit;
A machine learning model device comprising: