JP2021068350A - Vehicle estimation device - Google Patents

Abstract

To provide a vehicle determination device capable of determining a vehicle with high flexibility.SOLUTION: An image pickup unit 2 captures an image of an automobile and outputs the captured image. Vehicle recognition means 4 recognizes and outputs a vehicle included in the captured image. Vehicle image extracting means 6 extracts an image (vehicle image) of a vehicle portion based on the above recognition. Part extraction means 8 extracts a part region such as front and rear tires, a front surface, a side surface, and an upper surface in the vehicle image. Size estimation means 10 estimates whether the vehicle is a large vehicle or a small vehicle based on a relationship between the extracted plurality of part regions. This size estimation means 10 can be constructed by inference means in which the relationship between a plurality of part regions is specified and learned for the large vehicle and the small vehicle. As described above, the large vehicle and the small vehicle can be estimated.SELECTED DRAWING: Figure 1

Description

The present invention relates to a device for estimating a large vehicle such as an automobile or a small vehicle from an captured image.

It is necessary to determine whether a vehicle is a large vehicle or a small vehicle for traffic volume surveys and facility management. For example, in a traffic volume survey, an investigator visually classifies passing vehicles, large and small vehicles, and the like at intersections, and counts the traffic volume for each classification.

In recent years, software (for example, YOLO) that extracts a part of an automobile from a captured image and estimates the type of the automobile (passenger car, truck, etc.) has been put into practical use.

JP-A-2017-142587

However, when the above software is applied to traffic volume surveys, etc., even if the type of automobile can be estimated, it cannot be estimated whether it is a large car or a small car. This is because automobiles of the same type have similar shapes regardless of whether they are large or small.

In addition, a system has been proposed in which characters on the license plate of an automobile are read and the type or the like is estimated from the characters (Patent Document 1). According to this system, it is possible to obtain not only the type of automobile but also the distinction between a large car and a small car appearing on the letters on the license plate.

By estimating by the letters of the license plate like this, it is possible to distinguish not only the type but also the large size and the small size.

However, it may not be possible to read the characters on the license plate depending on the imaging distance, position, weather, and so on. For this reason, there are restrictions on its use for traffic surveys and the like.

An object of the present invention is to solve the above-mentioned problems and to provide a vehicle determination device capable of performing highly flexible vehicle determination.

The independently applicable features of the present invention are listed below.

(1) (2) The vehicle estimation device according to the present invention is provided by an imaging unit that images a vehicle and outputs an captured image, a vehicle recognition means that recognizes and outputs a vehicle included in the captured image, and a vehicle recognition means. A vehicle image extraction means for extracting a vehicle image of a recognized vehicle, a part extraction means for extracting a part region including front and rear tires and side surfaces of the vehicle in the extracted vehicle image, and the above-mentioned extraction. It is provided with a size estimation means for estimating whether the vehicle is a large vehicle or a small vehicle based on the relationship between a plurality of site regions.

Therefore, even if the vehicles have similar outer shapes, it is possible to distinguish between a large vehicle and a small vehicle.

(3) In the vehicle estimation device according to the present invention, the imaging unit continuously images a traveling vehicle and outputs a plurality of captured images, and the size estimation means is a large vehicle for a plurality of captured images. It is characterized by integrating the judgment of whether it is a small car or a small car and making the final judgment of whether it is a large car or a small car.

Therefore, the estimation can be performed more accurately.

(4) The vehicle estimation device according to the present invention is characterized in that the part extraction means extracts a part region including a windshield, a rear glass or a license plate in place of or in addition to the front and rear tires and side surfaces of the vehicle. There is.

Therefore, large and small estimates can be made based on these regions.

(5) The vehicle estimation device according to the present invention is characterized in that the magnitude estimation means also uses the extracted vehicle image as a basis for estimation.

Therefore, the estimation can be performed more accurately.

(6) The vehicle estimation device according to the present invention is characterized in that the size estimation means also uses the size of the frame when the frame of the extracted vehicle image is normalized according to the distance from the camera as the basis for estimation. It is supposed to be.

Therefore, the estimation can be performed more accurately.

(7) In the vehicle estimation device according to the present invention, the area occupied by the vehicle in the vehicle image when the size estimation means normalizes the extracted vehicle image according to the distance from the camera is also used as the basis for estimation. It is characterized by that.

Therefore, the estimation can be performed more accurately.

(8) In the vehicle estimation device according to the present invention, the vehicle recognition means determines at least the vehicle types of passenger cars, buses, and trucks, and the size estimation means separately performs size estimation for each vehicle type. It is said.

Therefore, the estimation can be performed more accurately.

(9) (10) The admission management device according to the present invention has an imaging unit that captures an image of a vehicle trying to enter a facility and outputs an image, and a vehicle recognition that recognizes and outputs a vehicle included in the captured image. The means, the vehicle image extracting means for extracting the vehicle image of the vehicle recognized by the vehicle recognition means, and the extracted vehicle image are received, and the part region including the front and rear tires and the side surface of the vehicle in the image is extracted. A facility based on the extraction means, the size estimation means for estimating whether the vehicle is a large vehicle or a small vehicle based on the relationship between the extracted plurality of site regions, and at least whether the estimation result is a large vehicle or a small vehicle. It is equipped with a gate control means for controlling the opening and closing of the gate for entering the vehicle.

Therefore, it is possible to estimate whether the large vehicle is a small vehicle and control the opening and closing of the gate correspondingly.

(11) (12) The charge management device according to the present invention has an imaging unit that captures an image of a vehicle trying to enter a facility and outputs an captured image, and a vehicle recognition that recognizes and outputs a vehicle included in the captured image. The means, the vehicle image extracting means for extracting the vehicle image of the vehicle recognized by the vehicle recognition means, and the extracted vehicle image are received, and the part region including the front and rear tires and the side surface of the vehicle in the image is extracted. A facility based on the extraction means, the size estimation means for estimating whether the vehicle is a large vehicle or a small vehicle based on the relationship between the extracted plurality of site regions, and at least whether the estimation result is a large vehicle or a small vehicle. It is equipped with a charge calculation means for calculating usage charges.

Therefore, it is possible to estimate whether the large vehicle is a small vehicle and calculate the charge corresponding to this.

(13) In the apparatus according to the present invention, the imaging unit continuously images a traveling vehicle and outputs a plurality of captured images, and the magnitude estimation means is a large vehicle or a small vehicle for the plurality of captured images. It is characterized by integrating the judgments and making the final judgment as to whether it is a large vehicle or a small vehicle.

Therefore, the estimation can be performed more accurately.

(14) The apparatus according to the present invention is characterized in that the site extraction means extracts a site region including a windshield, a rear glass or a license plate in place of or in addition to the front and rear tires and side surfaces of the vehicle.

Therefore, large and small estimates can be made based on these regions.

(15) The apparatus according to the present invention is characterized in that the magnitude estimation means also uses the extracted vehicle image as a basis for estimation.

Therefore, the estimation can be performed more accurately.

(16) The apparatus according to the present invention is characterized in that the magnitude estimation means also uses the size of the frame when the frame of the extracted vehicle image is normalized according to the distance from the camera as the basis for estimation. ..

Therefore, the estimation can be performed more accurately.

(17) The apparatus according to the present invention uses the area occupied by the vehicle in the vehicle image when the size estimation means normalizes the extracted vehicle image according to the distance from the camera as the basis for estimation. It is a feature.

Therefore, the estimation can be performed more accurately.

(18) In the device according to the present invention, the vehicle recognition means determines at least the vehicle type of a passenger car, a bus, or a truck, and the size estimation means is characterized in that the size is estimated separately for each of the vehicle types. ..

Therefore, the estimation can be performed more accurately.

In the embodiment, step S4 corresponds to the "vehicle recognition means".

In the embodiment, step S8 corresponds to the “vehicle image extraction means”.

In the embodiment, step S10 corresponds to the “site extraction means”.

The "large / small estimation means" is a concept that at least distinguishes between a large vehicle and a small vehicle, and includes a vehicle type determination. In the embodiment, step S12 corresponds to this.

The "program" is a concept including not only a program that can be directly executed by the CPU, but also a source format program, a compressed program, an encrypted program, and the like.

It is a functional block diagram of the vehicle estimation device by one Embodiment of this invention. This is the hardware configuration of the car estimation device. It is a flowchart of a car estimation program. This is an example of a captured image, a vehicle recognition image, and a part estimation image. It is a flowchart of learning process for vehicle recognition. It is a figure which shows the data for learning. This is an example of an image of a vehicle cut out. It is a flowchart of the learning process for site estimation. It is a figure which shows the learning data for site estimation. It is a flowchart of learning process for magnitude estimation. It is a figure which shows the data for learning. This is a table for modifying the vehicle type. It is a functional configuration of the entrance management device according to the second embodiment. It is a figure which shows the appearance of the gate. It is a figure which shows the hardware configuration. It is a flowchart of a control program.

1. 1. First Embodiment
1.1 Functional configuration Figure 1 shows the functional configuration of the vehicle estimation device according to the embodiment of the present invention. The image pickup unit 2 takes an image of the automobile and outputs the captured image. The vehicle recognition means 4 recognizes and outputs the vehicle included in the captured image. For example, the vehicle is enclosed in a boundary box and output. The vehicle recognition means 4 can be constructed by, for example, an inference means learned from an image of a vehicle.

The vehicle image extracting means 6 extracts an image (vehicle image) of the vehicle portion based on the above recognition. The part extraction means 8 extracts part regions such as front and rear tires, front surface, side surface, and upper surface in the vehicle image. The site extraction means 8 can be constructed by an inference means in which a site is specified and learned.

The size estimation means 10 estimates whether the vehicle is a large vehicle or a small vehicle based on the relationship between the extracted plurality of region regions. The size estimation means 10 can be constructed by an inference means in which the relationship between a plurality of site regions is specified and learned for a large vehicle and a small vehicle.

As described above, large vehicles and small vehicles can be estimated.

1.2 Hardware configuration Figure 2 shows the hardware configuration of the vehicle estimation device. A memory 32, a display 34, a camera 2, a hard disk 36, a DVD-ROM drive 38, a keyboard / mouse 40, and a communication circuit 42 are connected to the CPU 30.

The communication circuit 42 is for connecting to the Internet. The operating system 44 and the vehicle estimation program 46 are recorded on the hard disk 36. The vehicle estimation program 46 exerts its function in cooperation with the operating system 44.

These programs are those recorded on the DVD-ROM 48 and installed on the hard disk 36 via the DVD-ROM drive 38.

1.3 Vehicle estimation processing Figure 3 shows a flowchart of vehicle estimation processing. The CPU 30 acquires an image captured by the camera 2 (step S2). In this embodiment, imaging is performed as a moving image. FIG. 4A shows an example of a captured image. In this embodiment, a camera 2 is provided on a utility pole or a street, and an image is taken from an angle from above.

The CPU 30 recognizes the vehicle included in the captured image and estimates the vehicle type (step S4). In this embodiment, a trained inference model (for example, a model of machine learning by deep learning) is used for such estimation processing.

In this embodiment, the learning process is performed using the computer of FIG. Of course, other computers may be used.

First, a large number of image data including a vehicle as shown in FIG. 6A are prepared and recorded on the hard disk 36. If possible, it is preferable that there are many images of various vehicle types and backgrounds.

At this time, it is preferable that the angle at which the vehicle is imaged is as close as possible to the angle when actually operating the vehicle. For example, if an image is taken at an angle as shown in FIG. 4A for operation, an angle close to this is preferable.

Each image data (original image data) is displayed on the display 34. The user operates the keyboard / mouse 40 while looking at the image, encloses the vehicle portion with a rectangle (boundary box) as shown in FIG. 6B, and inputs the vehicle type of the vehicle. As described above, the analysis data in which the upper left coordinate, the lower right coordinate (coordinates in the two-dimensional image) and the vehicle type of the boundary box are recorded are associated with the original image data and recorded on the hard disk 36. In this embodiment, passenger cars, freight cars, and buses are distinguished as vehicle types. The above process is performed on all original images. As a result, a large amount of learning data will be recorded on the hard disk 36.

FIG. 5 shows a flowchart of a process of learning an estimation model (deep learning) for vehicle recognition and vehicle type estimation.

The CPU 30 acquires the original image from the hard disk 36 (step S52). For example, the original image as shown in FIG. 6A is read out. Next, the CPU 30 performs vehicle recognition and vehicle type estimation on the original image by a trained estimation model (deep learning) (steps S54 and S56).

Next, the CPU 30 reads out the analysis data (FIG. 6C) recorded in association with the original image (step S58). Subsequently, the CPU 30 uses the analysis data of FIG. 6C as teacher data, and learns the parameters for estimation in steps S54 and S56 based on the results of vehicle recognition and vehicle estimation (step S60).

As described above, the trained inference model is generated. Initially, it is an unlearned inference model or an under-learned inference model, but by repeating the above processing, a sufficiently learned inference model can be obtained.

In this embodiment, not only the photograph from the front of the vehicle but also the photograph from the rear or the side is used as the learning data. Therefore, it is possible to generate an inference model that can make inferences regardless of the vehicle imaged in any direction. If it is only necessary to determine the vehicle from a specific direction such as the front during operation, the learning data may be created based on the photograph from only the specific direction such as the front.

A program such as YOLO may be used as the learned inference model. In addition, a program such as YOLO may be additionally learned and used.

In step S4 of FIG. 3, the CPU 30 uses the inference model generated in this way to enclose the vehicle projected in the captured image (FIG. 4A) with a boundary box and estimate the vehicle type.

FIG. 4B shows an image in which the vehicle is recognized and surrounded by a boundary box. Further, although not shown, the estimated vehicle type is recorded in association with each boundary box (recorded in the format as shown in FIG. 6B).

Subsequently, the CPU 30 cuts out an image of each vehicle based on the boundary box (step S8). An example of the cut out vehicle image is shown in FIG.

The CPU 30 estimates the part of the vehicle based on the vehicle image (step S10). The size of the cut-out image differs depending on the distance from the vehicle. Therefore, in this embodiment, vehicle images having different sizes depending on the distance are converted (normalized) into a unified size before estimation. For example, the entire vehicle image is uniformly enlarged or reduced so that the width of the vehicle image (width of the boundary box) becomes a predetermined width.

In this embodiment, a trained inference model (for example, a model of machine learning by deep learning) is used for such estimation processing.

In this embodiment, the learning process is performed using the computer of FIG. Of course, other computers may be used.

FIG. 8 shows a flowchart of the learning process. In this embodiment, SegNet (available at https://qiita.com/uni-3/items/a62daa5a03a02f5fa46d), which performs estimation processing by semantic segmentation, was used as an engine for deep learning.

First, a large number of image data including an automobile as shown in FIG. 9A are prepared and recorded on the hard disk 36. If possible, it is preferable that there are many images with various vehicle types and various backgrounds.

Each image data (original image data) is displayed on the display 34, and the operator operates the mouse 40 while viewing the image to display the front license plate, the rear license plate, the front tire, the rear tire, the left tire, and the right. Label the tires, front, front glass, left side, right side, back, rear glass, top and background (non-automobile parts) with different colors. An example of a label image is shown in FIG. 9B. The generated label image data is recorded on the hard disk 36 in association with the original image data.

As described above, a large number of original image data and corresponding label image data are recorded on the hard disk 36.

First, the CPU 30 acquires the original image data from the hard disk 36 (step S62). For example, the original image as shown in FIG. 9A is read out. Next, the CPU 30 uses the trained semantic segmentation on the original image to perform front license plate, rear license plate, front tire, rear tire, left tire, right tire, front, windshield, left side, right side, and back. , Rear glass, and upper surface are estimated (step S64). An image of the estimation result is shown in FIG. 9C. As shown in FIG. 9C, the images are colored differently for each part.

In this embodiment, the original image is repeatedly pooled and convolved, and then upsampling and convolution are repeated to obtain an image of the estimation result.

Next, the CPU 30 reads out the label image recorded on the hard disk 36 corresponding to the original image (step S66). For example, a label image as shown in FIG. 9B is read out.

Subsequently, the CPU 30 uses the label image of FIG. 9B as teacher data, and learns the parameters for estimation in step S64 based on the estimation result image of FIG. 9C (step S68).

When learning is performed based on all the original images / label images, the CPU 30 ends the learning process (steps S60 and S70). Initially, it is an unlearned inference model or an under-learned inference model, but by repeating the above processing, a sufficiently learned inference model can be obtained.

In the above embodiment, learning is performed based on an image of each vehicle. However, learning may be performed based on images captured by a plurality of vehicles.

In step S10 of FIG. 3, the CPU 30 estimates a part region of one vehicle using the estimation model learned as described above (step S10). As a result, as shown in FIG. 4C, each part of the automobile is estimated. In this embodiment, a part estimation image (distinguished by color, density, etc.) divided into areas for each part is output.

FIG. 4C shows an example of a site estimation image. In the case of this example, the front surface, the front license plate, the windshield, the left side surface, the upper surface, the left front tire, and the left rear tire are recognized.

Next, the CPU 30 estimates a large vehicle / a small vehicle based on the above-mentioned part estimation image and the vehicle type estimated in step S4 (step S12).

As described above, in this embodiment, a passenger car, a freight car, and a bus are obtained in step S3. However, freight vehicles include small freight vehicles (light trucks) and ordinary freight vehicles (ordinary trucks), and in order to distinguish between them, large and small estimates are made. The former is small and the latter is large.

In addition, some passenger cars, such as wagons and minibuses, may be mistakenly recognized as large buses because of their shape. Therefore, in this embodiment, in order to distinguish between them, large size and small size are estimated. The former is small and the latter is large.

As described above, by distinguishing between large and small vehicles, passenger cars (small cars), small freight cars (small cars), buses (large cars), and ordinary freight cars (small cars), which are the classifications used for road traffic census, etc. You can get a large car).

In this embodiment, a trained inference model (for example, a machine learning model by deep learning) is used for large-sized and small-sized estimation processing.

In this embodiment, the learning process is performed using the computer of FIG. Of course, other computers may be used.

FIG. 10 shows a flowchart of the learning process. First, a large number of site estimation image data as shown in FIG. 11A are prepared and recorded on the hard disk 36. If possible, it is preferable that there are many images of various vehicle types.

Each part estimation image is displayed on the display 34, and the operator operates the keyboard / mouse 40 while viewing the image to input the distinction between a large car and a small car. In addition, since it may be difficult to distinguish only from the part estimation image, the original image is also displayed. As described above, a passenger car (including a minibus and a wagon car) and a small freight car are referred to as a small car, and a bus and an ordinary freight car are referred to as a large car. The distinction between the large vehicle and the small vehicle is recorded on the hard disk 36 in association with the part estimation image.

As described above, the part estimation image and the corresponding distinction between the large vehicle and the small vehicle are recorded on the large number of hard disks 36.

First, the CPU 30 acquires a site estimation image from the hard disk 36 (step S82). For example, a site estimation image as shown in FIG. 11A is read out. Next, the CPU 30 estimates the size and size of the part estimation image by the trained estimation engine (step S84).

Next, the CPU 30 reads out the distinction between large and small size recorded on the hard disk 36 corresponding to the site estimation image (step S66).

Subsequently, the CPU 30 uses the read large-sized / small-sized distinction as teacher data, and learns the parameters for estimation in step S84 based on the estimation result in step S84 (step S88).

When learning is performed based on all the part estimation images, the CPU 30 ends the learning process (steps S80 and S90). Initially, it is an unlearned inference model or an under-learned inference model, but by repeating the above processing, a sufficiently learned inference model can be obtained.

It is considered that the reason why this estimation works is that the ratio of the windshield and the distance between the front and rear tires are different depending on the large car and the small car, even if the outer shapes are similar.

In the above embodiment, learning is performed based on an image of each vehicle. However, learning may be performed based on images captured by a plurality of vehicles.

In step S12 of FIG. 3, the CPU 30 uses the estimation model learned as described above to perform large-sized and small-sized estimation based on the site estimation image. Next, the CPU 30 modifies the vehicle model estimation in step S4 based on the large / small estimation results (step S14).

In step S4, the vehicle types of passenger cars, freight cars, and buses were estimated. In this embodiment, in step S14, as shown in FIG. 12, the vehicle type estimation is corrected to be accurate based on the large / small estimation.

Next, the CPU 30 determines whether or not the vehicle first appears in the moving image of the vehicle being continuously imaged. This can be determined by whether a vehicle of similar shape is recognized for the first time. In addition, it is possible to narrow down the target vehicles for determining whether or not they have appeared for the first time in relation to the traveling direction of the road. For example, in the image of FIG. 4A, if the vehicle is approaching, it can be determined only for the vehicle above the line 200. The same applies to vehicles moving in other directions.

If this is the first vehicle imaged, a new ID is generated for the vehicle, and the estimated result corrected in association with the vehicle is recorded (step S20). If the vehicle has already been recognized, the estimated result corrected in association with the vehicle is added and recorded (step S18).

The CPU 30 repeats the above estimation for the number of vehicles (steps S6 and S22). As a result, it is possible to estimate the vehicle type for each vehicle projected in the image captured by the camera.

Next, the CPU 30 captures the next captured image (step S2), estimates the vehicle type of each vehicle in the same manner as described above, and records the results in association with each other (steps S4 to S22).

If the above is repeated, some objects will not be imaged due to the movement of the vehicle. For the vehicle, the recorded modified vehicle type estimation results (s) are integrated to obtain the final vehicle type estimation result, and the final vehicle type estimation result is associated with the vehicle (vehicle ID). And record (step S24). For example, among a plurality of vehicle model estimation results, the vehicle model estimation result having the largest number is used as the final vehicle model estimation result.

As described above, it is possible to accurately determine the vehicle type while performing real-time imaging. For example, it can be used for determining a vehicle type in a traffic volume survey or the like.

1.4 Other
(1) In the above embodiment, in step S14, large / small size estimation is performed based on the site estimation image. However, the vehicle image extracted in step S8 (preferably normalized) may also be used for estimation.

(2) In the above embodiment, the size of the vehicle image is normalized and then the estimation is performed. However, the estimation may be performed without normalization.

(3) In the above embodiment, first, the vehicle type is estimated, then the large / small size is estimated, and the final vehicle type is obtained. However, in step S4, the final vehicle type may be estimated at once by using the vehicle part estimation image.

(4) In the above embodiment, the vehicle type is estimated in real time. However, the vehicle type may be estimated by performing the above processing based on the recorded captured image. In this case, the image captured by the camera 2 may be recorded on a portable recording medium, and the image may be read into the hard disk 36 for processing.

(5) In the above embodiment, as the part estimation image, an image in which the part areas such as the front surface, the front license plate, the side surface, the tire, and the windshield are clarified is used. However, in large cars and small cars, it is possible to use a part estimation image including a part where the overall size can be known and a part where the size of the windshield, tire spacing and license plate can be known. For example, a site estimation image including at least the windshield and any side surface may be used. Further, at least, a part estimation image including any front tire and any rear tire and any side surface may be used.

(6) In the above embodiment, a common learned inference model for all vehicles is constructed to estimate large vehicles and small vehicles. However, a learned inference model may be constructed for each vehicle type estimated in step S4, and a large vehicle / a small vehicle may be estimated for each vehicle type. In addition, the vehicle type may be determined.

(7) In the above embodiment, the inference of a large vehicle / small vehicle is performed based on the part estimation image. However, instead of or in addition to this, the size of the boundary box frame when the vehicle image is normalized, the area occupied by the vehicle, and the like may be used as the basis for inference.

(8) In the above embodiment, a vehicle moving in various directions is imaged. However, when the camera is installed, if it is set so that only a vehicle moving to only one side (such as an oncoming vehicle) is imaged, the estimation process becomes easy and the accuracy is improved.

(9) In the above embodiment, the size is determined based on a plurality of vehicle images for the same vehicle. However, this may be done on the basis of a single image.

(10) The above-described embodiment and its modifications can be implemented in combination with other embodiments as long as it does not contradict its essence.

2. Second embodiment
2.1 Functional configuration Figure 13 shows a functional block diagram of an admission management system according to an embodiment of the present invention. The imaging unit 2 is provided at the entrance of a parking lot such as a facility, and images a vehicle approaching the entrance gate. The vehicle type estimation means 100 estimates the vehicle type based on the captured image by, for example, the method described in the first embodiment. The gate control means 150 controls the opening and closing of the gate based on the estimated vehicle type. For example, a large car and a small car control whether to open a gate to a parking lot for a large car and a gate to a parking lot for a small car.

2.2 System configuration and operation Figure 14 shows the appearance of the entrance gate of the entrance management system. A gate 120 for a small car and a gate 140 for a large car are provided. The camera 2 images a vehicle trying to enter these gates.

FIG. 15 shows the hardware configuration. The computer 160 is similar to the configuration of FIG. 2 in the first embodiment. However, the CPU 30 can give a command to the gate control unit 180 that controls the opening and closing of the gates 120 and 140.

FIG. 16 shows a flowchart of the control program. In step S110, the CPU 30 acquires an image captured by the camera 2. If the vehicle is not recognized in the captured image, step S110 is repeated.

When the vehicle is recognized, the CPU 30 estimates the vehicle type (large / small) by the process described in the first embodiment (step S114). When it is estimated that the size is large, the CPU 30 instructs the gate control unit 180 to open the large gate 140 (step S116). Further, when it is estimated that the size is small, the CPU 30 instructs the gate control unit 180 to open the small gate 120 (step S116).

After opening the gate, the CPU 30 closes the gate when it receives an output from a passage detection sensor (not shown) (step S118).

As described above, it is possible to automatically determine whether the size is large or small, select the destination gate, and open / close the gate.

2.3 Others
(1) In the above embodiment, the parking lot has been described, but the same can be applied to other facilities such as a drive-through safari. Further, in the above, the gate is selectively opened and closed based on the judgment of large size and small size, but the gate may be selectively opened and closed according to a finer vehicle type classification.

In addition, it may be determined whether or not to open the gate according to a predetermined vehicle type (or depending on whether it is large or small). For example, in the case of a bridge dedicated to small cars, it is possible to control the entrance of the bridge so that the gate cannot be opened for large cars.

(2) In the above embodiment, the case of controlling the opening and closing of the gate has been described. However, the parking fee and the facility usage fee may be calculated according to the vehicle type estimation (including the estimation of large and small size). It can be used at entrances such as highways.

(3) The above-described embodiment and its modified examples can be implemented in combination with other embodiments and modified examples as long as the essence is not contrary to the essence.

Claims (18)

An imaging unit that captures the vehicle and outputs the captured image,
A vehicle recognition means that recognizes and outputs the vehicle included in the captured image,
A vehicle image extraction means for extracting a vehicle image of a vehicle recognized by the vehicle recognition means, and a vehicle image extraction means.
A part extraction means that receives the extracted vehicle image and extracts a part region including the front and rear tires and side surfaces of the vehicle in the image, and a part extraction means.
A size estimation means for estimating whether a large vehicle or a small vehicle is used based on the relationship between the extracted plurality of site regions, and
Car estimation device equipped with. It is a vehicle type estimation program for realizing a vehicle estimation device by a computer.
Vehicle recognition means that recognizes and outputs the vehicle included in the captured image of the vehicle,
A vehicle image extraction means for extracting a vehicle image of a vehicle recognized by the vehicle recognition means, and a vehicle image extraction means.
A part extraction means that receives the extracted vehicle image and extracts a part region including the front and rear tires and side surfaces of the vehicle in the image, and a part extraction means.
A vehicle estimation program for functioning as a size estimation means for estimating whether a large vehicle or a small vehicle is based on the relationship between the extracted plurality of region regions. In the apparatus of claim 1 or the program of claim 2.
The imaging unit continuously images a traveling vehicle and outputs a plurality of captured images.
The magnitude estimation means is a device or program that integrates the determination of a large vehicle or a small vehicle from a plurality of captured images to make a final determination of whether the vehicle is a large vehicle or a small vehicle. In any of the devices or programs of claims 1-3
A device or program, wherein the site extraction means extracts a site region including a windshield, a rear glass or a license plate in place of or in addition to the front and rear tires and sides of the vehicle. In any of the devices or programs of claims 1 to 4,
The magnitude estimation means is a device or program that also uses the extracted vehicle image as a basis for estimation. In any of the devices or programs of claims 1-5
The magnitude estimation means is an apparatus or program characterized in that the size of the frame when the frame of the extracted vehicle image is normalized according to the distance from the camera is also used as the basis for estimation. In the device or program according to any one of claims 1 to 6.
The magnitude estimation means is a device or program that also uses the area occupied by the vehicle in the vehicle image when the extracted vehicle image is normalized according to the distance from the camera as the basis for estimation. In any of the devices or programs of claims 1-7
The vehicle recognition means determines at least the vehicle type of a passenger car, a bus, or a truck.
The magnitude estimation means is an apparatus or program characterized in that the magnitude is estimated separately for each vehicle type. An imaging unit that captures the vehicle trying to enter the facility and outputs the captured image,
A vehicle recognition means that recognizes and outputs a vehicle included in the captured image, and
A vehicle image extraction means for extracting a vehicle image of a vehicle recognized by the vehicle recognition means, and a vehicle image extraction means.
An extraction means that receives the extracted vehicle image and extracts a site region including the front and rear tires and side surfaces of the vehicle in the image, and
A size estimation means for estimating whether a large vehicle or a small vehicle is used based on the relationship between the extracted plurality of site regions, and
A gate control means for controlling the opening and closing of the gate for entering the facility, at least based on whether the estimation result is a large vehicle or a small vehicle.
Admission management device equipped with. An admission management program for realizing an admission management device with a computer,
Vehicle recognition means that recognizes and outputs the vehicle included in the captured image of the vehicle that is about to enter the facility,
A vehicle image extraction means for extracting a vehicle image of a vehicle recognized by the vehicle recognition means, and a vehicle image extraction means.
An extraction means that receives the extracted vehicle image and extracts a site region including the front and rear tires and side surfaces of the vehicle in the image, and
A size estimation means for estimating whether a large vehicle or a small vehicle is used based on the relationship between the extracted plurality of site regions, and
An entrance management program for functioning as a gate control means for controlling the opening and closing of a gate for entering a facility, at least based on whether the estimation result is a large vehicle or a small vehicle. An imaging unit that captures the vehicle trying to enter the facility and outputs the captured image,
A vehicle recognition means that recognizes and outputs a vehicle included in the captured image, and
A vehicle image extraction means for extracting a vehicle image of a vehicle recognized by the vehicle recognition means, and a vehicle image extraction means.
An extraction means that receives the extracted vehicle image and extracts a site region including the front and rear tires and side surfaces of the vehicle in the image, and
A size estimation means for estimating whether a large vehicle or a small vehicle is used based on the relationship between the extracted plurality of site regions, and
A fee calculation means for calculating a facility usage fee based on at least whether the estimation result is a large vehicle or a small vehicle.
Charge management device equipped with. It is a charge management program for realizing a charge management device by a computer.
Vehicle recognition means that recognizes and outputs the vehicle included in the captured image of the vehicle that is about to enter the facility,
A vehicle image extraction means for extracting a vehicle image of a vehicle recognized by the vehicle recognition means, and a vehicle image extraction means.
An extraction means that receives the extracted vehicle image and extracts a site region including the front and rear tires and side surfaces of the vehicle in the image, and
A size estimation means for estimating whether a large vehicle or a small vehicle is used based on the relationship between the extracted plurality of site regions, and
A charge management program for functioning as a gate control means for controlling the opening and closing of a gate for entering a facility, depending on whether the estimation result is a large vehicle or a small vehicle. In any of the programs of claims 9-12
The imaging unit continuously images a traveling vehicle and outputs a plurality of captured images.
The magnitude estimation means is a device or program that integrates the determination of a large vehicle or a small vehicle from a plurality of captured images to make a final determination of whether the vehicle is a large vehicle or a small vehicle. In any of the devices or programs of claims 11-13.
The site extracting means is a device or program characterized in that the site extracting means extracts a site region including a windshield, a rear glass or a license plate in place of or in addition to the front and rear tires and side surfaces of a vehicle. In any of the devices or programs of claims 11-14.
The magnitude estimation means is a device or program that also uses the extracted vehicle image as a basis for estimation. In any of the devices or programs of claims 11-15.
The magnitude estimation means is an apparatus or program characterized in that the size of the frame of the vehicle image when the frame of the extracted vehicle image is normalized according to the distance from the camera is also used as the basis for estimation. In any of the devices or programs of claims 11-16,
The magnitude estimation means is a device or program that also uses the area occupied by the vehicle in the vehicle image when the extracted vehicle image is normalized according to the distance from the camera as the basis for estimation. In any of the devices or programs of claims 11-17,
The vehicle recognition means determines at least the vehicle type of a passenger car, a bus, or a truck.
The magnitude estimation means is an apparatus or program characterized in that the magnitude is estimated separately for each vehicle type.

Images