JP7412326B2 - Parking lot management device - Google Patents

Description

The present application relates to a parking lot management device .

Machine learning (deep learning) of AI (Artificial Intelligence) is used to extract predetermined objects from two-dimensional images. For example, it has been proposed to use AI to determine whether a vehicle is parked in a parking slot based on images of a parking lot captured by a surveillance camera.

As a system that detects the usage status of parking spaces in parking lots, it detects images of parking spaces from images captured by parking lot surveillance video cameras, and determines whether the images of parking spaces satisfy the characteristics of empty parking spaces. It has been proposed to detect vacant parking spaces by doing so (Patent Document 1).

Japanese Patent Application Publication No. 2017-84364

As shown in Patent Document 1, in order to predetermine a target object (empty space) and determine the state of the target object from the image, images taken by a camera corresponding to each parking space are required. It is. Therefore, it is necessary to install the camera at a high position or to provide a large number of cameras.

An object of the present application is to provide a parking lot management device that requires cameras but is less affected by restrictions on the location and number of cameras .

The parking lot management device of the present application includes an imaging device that acquires image information, a first storage device that stores information on characteristic parts of a vehicle, a second storage device that stores the image information acquired by the imaging device, The image information in the second storage device is divided into a plurality of regions, and the image information in each of the divided regions and the information on the characteristic part of the vehicle stored in the first storage device are used. An object identification device equipped with an information processing device that determines that the vehicle is included in the image information is provided as a parking lot management device, and the presence of a vehicle blocked by a shielding object is provided in an identification area by the object identification device. is unknown, the image information stored in the second storage device is retrospectively checked to confirm whether or not the vehicle is behind the shielding object.

According to the parking lot management device of the present application, it is possible to obtain a parking lot management device that is less subject to restrictions on the position and number of cameras.

1 is a schematic configuration diagram showing an object identification device according to a first embodiment; FIG. FIG. 2 is a hardware configuration diagram that implements a block diagram according to an embodiment. FIG. 7 is an explanatory diagram showing a flow of Embodiment 2; It is an explanatory view showing the relationship between a shielding object and an identification target object. FIG. 3 is a flowchart showing a procedure for counting the number of vehicles when the presence of vehicles is unknown due to an obstruction. FIG. 7 is an explanatory diagram showing a flow of the parking lot management device according to the third embodiment. FIG. 7 is an explanatory diagram showing a flow of the parking lot management device according to the fourth embodiment. FIG. 7 is an explanatory diagram showing a flow of the parking lot management device according to the fifth embodiment. FIG. 12 is an explanatory diagram showing a flow of the parking lot management device according to the seventh embodiment. It is an explanatory diagram showing a flow of a parking lot management device of Embodiment 8. FIG. 12 is an explanatory diagram showing a flow of the parking lot management device according to the ninth embodiment. FIG. 12 is an explanatory diagram showing a flow of the parking lot management device according to Embodiment 10; FIG. 12 is an explanatory diagram showing a flow of the parking lot management device according to the eleventh embodiment. FIG. 12 is an explanatory diagram showing the flow of the parking lot management device according to the twelfth embodiment. FIG. 12 is an explanatory diagram showing a flow of a vehicle management device according to a thirteenth embodiment.

Embodiments of an object identification device, a parking lot management device using the same, and a vehicle management device according to the present application will be described below with reference to the drawings. Note that the same or corresponding parts in each figure are designated by the same reference numerals, and redundant explanation will be omitted.

Embodiment 1.
FIG. 1 is a schematic configuration diagram showing an object identification device according to Embodiment 1 of the present application.
As shown in FIG. 1, an object identification device 100 according to the first embodiment includes an imaging device 1 that acquires image information, a first storage device 2 that stores information on characteristic parts of an object to be identified, and an object identification device 1 that acquires image information using the imaging device 1. A second storage device 3 that stores the image information that has been created, and a second storage device 3 that divides the image information in the second storage device 3 into a plurality of areas and stores the image information of each divided area and the first storage device 2. The information processing apparatus 4 is equipped with an information processing device 4 that determines whether the image information includes an identification target object based on information on the characteristic portion of the identification target object.

Based on the determination result by the information processing device 4, the identification device 5 identifies whether the subject acquired by the imaging device 1 is an object to be identified. The identification result of the identification device 5 is displayed on the display device 6.
The identification target object is set in the identification target object setting section 7. The characteristic portion of the object to be identified stored in the first storage device 2 is generated by image recognition using a deep learning technique using training data of the object to be identified. That is, the information on the characteristic part stored in the first storage device 2 is obtained by learning the comparison between the teacher data and the similar data, and is based on the identification object set in advance in the identification object setting section 7. Regarding objects, a teacher data creating section 8 creates training data used for deep learning, and a similar object data creating section 9 creates similar object data.

The teacher data is information about the external shape of the object to be identified, which is obtained by changing the drawing parameters of the front view, side view, and image of the three-dimensional structure of the object. The similar object data is information about the external shape of a different structure that is similar to the external shape of the identification target object. Then, using the teacher data created in the teacher data creation section 8 and the similar data created in the similar object data creation section 9, it is determined in what ways the object to be identified is similar to other similar objects. The learning processing device 10 repeatedly learns what features are distinguishable and what features are distinguishable. The information on the characteristic parts of the object to be identified that has been learned in the learning processing device 10 is stored in the first storage device 2 .

That is, the learning processing means 11 includes an identification target object setting section 7, a teacher data creation section 8, a similar data creation section 9, and a learning processing device 10. The object identification means 12 includes a first storage device 2 , a second storage device 3 , an information processing device 4 , and an identification device 5 .

Note that the learning processing means 11 and the object identification means 12 are constituted by a processor 200 and a storage device 201, as an example of hardware is shown in FIG. Although the storage device is not shown, it includes a volatile storage device such as a random access memory, and a nonvolatile auxiliary storage device such as a flash memory. Further, an auxiliary storage device such as a hard disk may be provided instead of the flash memory. Processor 200 executes a program input from storage device 201. In this case, the program is input from the auxiliary storage device to the processor 200 via the volatile storage device. Further, the processor 200 may output data such as calculation results to a volatile storage device of the storage device 201, or may store data in an auxiliary storage device via the volatile storage device.

Here, the object to be identified means an object to be identified from an image acquired by the imaging device 1 such as a camera. The object to be identified is not particularly limited, and can be set as appropriate depending on the purpose. For example, human visual perception includes small vehicles, large vehicles, specific types of vehicles, license plates with specific area names, people entering the facility, men and women, adults, children, animals, etc. included in images (videos) of parking lots. Examples include objects that can be detected by

Training data is a pair of "input data" and "correct label" used in supervised deep learning. Here, deep learning learning is performed by inputting input data to a neural network having a large number of parameters, and the frequency of correct answers is increased by repeatedly updating the difference between the inference label and the correct answer label. Furthermore, here, the presence of an identification target in an image is identified by recognizing a part of the appearance of the identification target, rather than making a determination based on the entire image of the identification target.

For this purpose, the image information acquired by the imaging device 1 is divided into a plurality of regions, and it is determined whether the image information of each of the plurality of divided regions includes a characteristic part of the object to be identified. For example, when a manager visually checks the number of small vehicles in a parking lot, the object to be identified is a small vehicle and only a small part of the vehicle body is visible. It performs the same function as counting using image recognition.
Image recognition is a technology that analyzes the image content of image data to recognize its shape and appearance. Here, image recognition is performed by extracting the outline of an object to be identified from image data, separating it from the background of the image, and then analyzing what is assumed to be the object to be identified.

Next, detection of an identification target using deep learning will be described.
What is used here is an algorithm that uses a one-step object detection method. The steps for detecting an object by this algorithm are as follows. First, an image acquired by the imaging device 1 is divided into grids of a predetermined size (for example, grid areas of equal size). A preset number of bounding boxes with a predefined shape centered at the center of the grid are predicted in the divided grid area. Each prediction is associated with class establishment and object confidence (whether the region contains an object or only the background).
Next, select bounding boxes that are associated with high object confidence and class probabilities. The object class with the highest class probability becomes the category (class) of that object. Then, only default boxes with high scores are extracted. Next, the overlap rate of default boxes is calculated for each class, and if the overlap rate is high, default boxes with low scores are removed. Then, a bounding box representing the position of the object is output.

Embodiment 2.
Next, a case will be described in which the object identification device is applied to a parking lot management device.
FIG. 3 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device.
As shown in FIG. 3, when used in the parking lot management device 201, multiple types of vehicles are set as objects to be identified. For example, ordinary cars, trucks, buses, motorcycles, bicycles, etc. are set as objects to be identified. Then, the imaging device 1 (camera) photographs the parking lot at predetermined time intervals (step S200). The edge computer 202 performs the functions of the learning processing means 11 and the object identification means 12 shown in FIG.

That is, a camera image is acquired (step S210), and a bounding box is predicted for object detection (step S220). Image recognition is performed and objects (regular cars, trucks, buses, motorcycles, bicycles) with high identification reliability are selected (step S230). Objects whose reliability is higher than a preset threshold are targeted for processing (step S240). Then, the number of identification objects to be processed is counted (step S250). As described in Embodiment 1, the counting of the number of vehicles here is carried out by the identification by the object identification device, that is, by dividing the image information stored in the second storage device into a plurality of regions. It is determined whether or not the image information includes the object to be identified based on the image information of each of the areas and the information on the characteristic part of the object stored in the first storage device.

It is determined whether the parking lot is full, empty, or crowded (step S260). The determination result is displayed on the display panel as full, empty, or crowded (step S270). Further, the cloud server displays on the home page whether the vehicle is full, vacant, or crowded (step S280). Furthermore, the results of counting the number of vehicles are aggregated by vehicle type and time zone (step S290).
In addition, in step S240, it is shown that those with high identification reliability are to be processed. This is similar to the situation where the parking lot manager stands at a predetermined location (monitoring location) and visually checks the parking lot, but there is another vehicle between the parking lot and the object. As a result, the object may not be visible at all. That is, due to the presence of a truck between the object (vehicle) and the imaging device, there may be cases where the object is not counted even though the object is present.

When a shield exists in this way, if a person (caretaker) is viewing the object, he or she will not be able to see the hidden object until the shield moves. However, in the parking lot management device of Embodiment 2, the imaging device 1 is set to acquire images of a predetermined area of the parking lot. If this happens, special management will be applied.

That is, FIG. 4 is an explanatory diagram showing the relationship between the shielding object and the identification target object. FIG. 4A shows a state in which the vehicle 30 to be identified is completely hidden by the truck 20. FIG. 4B shows a state in which the truck 20 serving as a shield is about to park after the vehicle 30 to be identified is parked. FIG. 4C shows a state in which when the truck 20 serving as a shield is parked, no vehicle 30 is parked in the shielded area. FIG. 4(d) shows a state in which the vehicle 30 is about to be parked in an area that is blocked by the truck 20 that is a blocking object.

As shown in FIG. 4(a), there are cases where even a portion of the vehicle 30 that is the object to be identified cannot be confirmed. At the time of checking the parking lot, since the truck 20 serving as a shield is parked, it is unclear whether the vehicle 30 already exists in the shielded area or whether the vehicle 30 does not exist.

As shown in FIG. 4A, if there is an area where the existence of the object to be identified is determined to be unknown due to a blocking object, processing is performed as shown in FIG. That is, FIG. 5 is a flowchart showing a processing procedure for counting the number of vehicles when the presence of a vehicle that is an object to be identified is unknown due to a blocking object.
In parking lot management, the state of vehicles entering and exiting is stored as image information in the second storage device 3, so this stored image information is used to identify the presence of vehicles as objects to be identified. confirm.

First, as shown in FIG. 5, it is determined whether there is a possibility that the vehicle is completely shielded (step S10). If a vehicle parked in a parking lot can be confirmed in some way (for example, if a part of its outline is confirmed, such as partially protruding), it is assumed that there is no occluded area. , the process is completed by counting the number of confirmed vehicles (step S11). If there is a possibility that the vehicle is completely shielded, image information before the vehicle is hidden by the shielding object is checked (step S12). Then, it is checked whether the vehicle was parked before the obstruction existed (step S13).

To ascertain whether there is any information that a vehicle that was already present left after the shield entered. If no vehicle is parked before the shield exists, check whether there is a vehicle parked behind the shield (in the area covered by the shield) after the shield exists (step S14). In short, if a vehicle is parked in the shielded area, the number of vehicles is added (step S15).
If there continues to be no vehicle behind the shield, it is determined that there is no increase or decrease in the number of vehicles (step 16).

Furthermore, if a vehicle was present in the shielded area before the existence of the shielding object, the number of hidden vehicles is added (step S17). Furthermore, it is confirmed whether the vehicle that was present behind the shielding object has moved after the image information indicating that the vehicle was present (step S18). If the vehicle is moving after counting, the number of vehicles is subtracted (step S19). Further, it is checked in step S14 whether a new vehicle has been parked after the vehicle has moved.
As shown in the above procedure, even if there is an obstruction, the image information of the time when the parking lot was managed is stored in the second storage device, so the image information can be returned to the time. By reproducing the information, it is possible to accurately determine whether the vehicle is parked behind the shield.

In the parking lot management device of Embodiment 2, the presence of an identification target is estimated from partial information of a characteristic part. Furthermore, if there is a shielding object that blocks the entire object to be identified, the existence of the object to be identified can be confirmed by confirming the image information before and after the existence of the shielding object. In other words, by checking past image information for areas to be monitored in parking lots where there is a possibility that objects to be identified may exist, and by keeping a record of the checking status, it is possible to obtain management information that could not be managed by humans. This allows for more reliable management.

Furthermore, in the second embodiment, it is possible to detect a vehicle not only during the day but also at night and under bad weather conditions such as snow and rain. In other words, with vehicle detection technology using clustering, vehicle detection is stable during daytime hours when images are clear, but detection accuracy is unstable at night and in bad weather conditions such as snow and rain. In contrast, by using deep learning object detection technology and training with images from night, snow, and rain, it is possible to stably detect vehicles. By making it possible to stably detect vehicles at night and during bad weather such as snow and rain, it is possible to provide stable services.

Embodiment 3.
Next, a case will be described in which the object identification device is applied to a parking lot management device and the attributes of parking lot users are totaled.
FIG. 6 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device.
In the second embodiment (FIG. 3), the objects to be identified were vehicles, but in this third embodiment, the objects to be identified are people and pets, and a system for detecting whether a car is full or empty in a parking lot is used. By using the image information from the imaging device to understand the attributes of people entering the facility, it is used for marketing and to communicate information on how crowded the facility is. Here, we use object recognition technology based on deep learning to detect people and count the number of people entering the facility. It is also possible to detect gender (male/female), age, and pet. Furthermore, it is also possible to read emotions from people's facial expressions.

In this third embodiment, in step S330, a person is included as an object to be identified in step S230 of the second embodiment. Then, "person" is set for image cropping (step S350). Then, as image recognition using deep learning, discrimination of gender, age, and facial expression is performed (step S351), and the acquired data of gender, age, and facial expression are counted (step S352). In parallel, the number of dogs and cats is counted (step S360). Then, the cloud server aggregates the data (step S370), and displays the congestion status of the facility on the homepage (step S380). Furthermore, the information is classified by time of day, day of the week, public holidays, weather, etc., and the relationship between facility users and sales performance is analyzed (step S390). Based on this analysis result, it can be reflected in the product configuration to be sold (step S391).

As mentioned above, by classifying information by time of day, day of the week, holidays, weather, etc. and analyzing sales results, we can adjust the product lineup at the store and the menu at the cafeteria to suit the customers who visit, and increase sales. It can be used to improve. Furthermore, by displaying the congestion status of the shop, it is possible to communicate the congestion level, which can lead to improved user satisfaction in the service area.

Embodiment 4.
Next, a case will be described in which the object identification device is applied to a parking lot management device, and demand is predicted based on the region name and vehicle type of the license plate.
FIG. 7 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device.
In this fourth embodiment, the region names on the license plates of the vehicles of users of the facility are read and tallied using deep learning image recognition. In addition, it uses deep learning to recognize vehicle types (regular cars, trucks, buses), determines customer demographics based on area names and vehicle types, and predicts demand at commercial facilities with parking lots.

In the second embodiment (FIG. 3), the object to be identified is a vehicle, but in this fourth embodiment, the object to be identified is a license plate, and images are taken to detect whether a car is full or empty in a parking lot. By using the device's image information to aggregate the data (region name and vehicle type) of the license plates of vehicles entering the facility, we can predict the demand for commercial facilities and tailor the products and services we sell to meet the demand. This will lead to increased sales and improved services.

In this fourth embodiment, a license plate is included as an object to be identified in step S230 of the second embodiment as step S430. Then, "license plate" is set for image cutting (step S450). Then, as image recognition using deep learning, characters and numbers are discriminated (step S451), and the acquired license plate data is counted (step S452). Then, the data is aggregated in the cloud server (step S490).

Embodiment 5.
Next, a case will be described in which the object identification device is applied to a parking lot management device, where information on whether the parking lot is full or empty can be detected according to the layout of the parking lot (the layout of parking spaces for vehicles).
FIG. 8 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device.
In this fifth embodiment, when it is not possible to judge whether a parking lot is full or empty based on the number of vehicles alone due to differences in operation, such as parking multiple regular cars in a space reserved for large cars, object recognition using deep learning is used to detect regular cars, Efficient use of parking space by classifying large vehicles and motorcycles and determining that there is a vacant space even if a regular vehicle is parked in the parking space of a large vehicle, if there is space for another regular vehicle. This is what makes it possible.

Similarly, multiple motorcycles may be parked in a vehicle space, so if a motorcycle is parked in a parking space for a large vehicle or a regular vehicle, the decision should be made based on whether there is an empty space to park a regular vehicle. In addition, if a vehicle is parked outside a parking space, the empty vehicle count should not be reduced, and by counting it as "parking outside a parking space", it is possible to prevent parking in an inappropriate parking space. be able to recognize it.
Furthermore, by creating classification boundaries that classify parking locations into multiple groups and allowing them to be changed arbitrarily, efficient use is made possible.

In the second embodiment (FIG. 3), the object to be identified is a vehicle, but in the fifth embodiment, the object to be identified is an ordinary car, a truck, a bus, a motorcycle, or a bicycle, and the object is an ordinary car, a large vehicle, or a two-wheeled vehicle. By making separate judgments, inappropriate parking can be recognized.

In this fifth embodiment, in step S530, in step S230 of the second embodiment, the objects to be identified are classified into ordinary cars, trucks, buses, motorcycles, and bicycles and are configured to be aggregated. There is. Then, it is determined whether the vehicle is full, vacant, or crowded, and it is determined whether the appropriate parking position for each vehicle type is being used, and vehicles parked in inappropriate areas are recognized (step S550). The results are then displayed on the display panel to draw attention to inappropriate parking (step S570). Additionally, the cloud server displays information on full, empty, and congestion information on the home page (step S580). Furthermore, various data are compiled based on the count of the number of vehicles (step S590).

Embodiment 6.
Next, we will explain the case where the object identification device is applied to a parking lot management device, and by recognizing vehicles by car type, the optimal vehicle allocation is calculated not only for parking slots in the parking area. .
This configuration and flow are the same as in the fifth embodiment. In this sixth embodiment, in a parking lot where the parking areas for large and ordinary/small vehicles are divided, the parking area for small/normal cars is crowded, but the parking area for large vehicles is empty. This section shows what to do when a small or regular vehicle is parked in an area reserved for large vehicles. That is, when a small vehicle is parked in a parking area for a large vehicle, two vehicles can be parked, but with conventional vehicle detection methods, it is difficult to determine whether a small vehicle is parked in the large vehicle area. Therefore, by performing vehicle detection using deep learning, it is possible to recognize large vehicles, regular vehicles, small vehicles, and two-wheeled vehicles separately.

This makes it possible to determine how many small vehicles are parked in the parking area for large vehicles and how many more can be parked. Furthermore, by analyzing how many large vehicles, regular vehicles, and small vehicles are parked where and how crowded they are, depending on the time of day, day of the week, and weather, we can calculate more efficient parking ratios. Can be done.
It is possible to detect and report occupancy violations by setting in advance the parking lot occupancy locations for each vehicle class, such as large, regular, small, and two-wheeled vehicles, and comparing them with the vehicle class that is actually parked. It is.
By classifying and determining large/small/normal/two-wheeled vehicles, it is possible to determine parking lot congestion with higher accuracy. Furthermore, by analyzing the usage status of parking lots by time of day, day of the week, and weather, it is effective in formulating parking lot improvement plans.

Embodiment 7.
Next, a case will be described in which the object identification device is applied to a parking lot management device, and it is possible to grasp the availability of parking zones for people with disabilities.
FIG. 9 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device. Embodiment 7 is intended to deal with the problem that the availability of parking zones for disabled people is not known until the driver arrives at the parking location.

The parking lot has a parking area for people with disabilities, but the parking lot occupancy monitoring system, which counts entry and exit, does not know where each vehicle is parked. Therefore, by performing vehicle detection using deep learning, it is possible to determine whether a parking space for the physically disabled is vacant or full. This makes it possible to notify users, security guards, and maintenance personnel by using a display board at the parking lot entrance or a distribution means such as the Web.

Convenience is improved because it can be determined in advance whether the vehicle can be parked in a parking space for the handicapped.
In the seventh embodiment, in determining whether the parking area is full, empty, or crowded, the parking area is displayed as full or empty by determining whether the parking area for the disabled is full or empty (step S351). is displayed (step S670).

Embodiment 8.
Next, a case will be described in which the object identification device is applied to a parking lot management device, and image processing is performed in real time in order to transmit information about the congestion level of the parking lot to the outside.
FIG. 10 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device.
In this eighth embodiment, a case will be described in which, when distributing information regarding parking lots to the outside, appropriate measures are taken to respect personal information and portrait rights contained therein.

In order to check how crowded a parking lot is, images are checked in real time using the Internet. However, since images captured by cameras clearly show images of vehicles, vehicle numbers, and people's faces, it is necessary to respect personal information and portrait rights.
Therefore, the video cannot be distributed as is. To address this, the system uses object detection using deep learning to detect vehicle license plates, people's faces, and other characteristic areas in the video, and then automatically mosaics or blurs only those areas to make them unclear.

Alternatively, by performing processing to recognize and remove humans from images in real time, the images can be processed into videos that can be published on the Internet.
Therefore, from the predicted bounding boxes, the object to be detected (human, license plate) is classified by selecting it with high reliability (step S730), and the object (human, license plate) that is to be detected is classified (step S730). A mosaic process is performed (step S750). Then, in the cloud server, the mosaic-processed image is published on the Internet in real time (step S780).

Embodiment 9.
Next, the object identification device is applied to a parking lot management device to prevent accidents by detecting parked vehicles at locations other than designated locations in the parking lot and vehicles driving in the wrong direction in the aisle.
FIG. 11 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device.
In this embodiment 9, an accident may occur in a parking lot due to a vehicle being stopped in a place other than a parking spot and a vehicle not traveling in a determined direction.

In this ninth embodiment, by detecting vehicles parked on traffic roads and taxiways using deep learning object detection technology, vehicles are prevented from illegally stopping on traffic roads and taxiways other than parking lots. It is possible to detect whether the Furthermore, a case will be described in which a wrong-way vehicle is detected by comparing the traveling direction of a preset path and guideway with the traveling direction of the vehicle.

As shown in FIG. 11, the edge computer 202 determines the driving direction of the vehicles in the parking lot and the vehicles entering and exiting the parking lot (step S850). A reverse phase is detected from the relationship between the vehicle's travel direction and the travel lane (step S851). Then, information about the wrong-way vehicle is notified to the target vehicle and surrounding vehicles on the display panel (step S870). Further, in the security room, information about the wrong-way vehicle is notified to the maintenance/security guard (step S880). This makes it possible to prevent accidents by notifying the target vehicle, surrounding vehicles, and maintenance/security personnel of information about illegally parked vehicles and vehicles driving the wrong way.

Embodiment 10.
Next, the object identification device is applied to a parking lot management device, in which a line of vehicles attempting to enter a parking lot is detected and information on a line of vehicles outside the parking lot is provided.
FIG. 12 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device.
In Embodiment 10, not only the inside of the parking lot but also the line of vehicles outside the parking lot is monitored by a camera, and when the inside of the parking lot is full, the image of the camera for monitoring the line of cars outside the parking lot is distributed to the outside, It is made available for viewing on the Internet.

As shown in FIG. 12, the imaging device 1 (camera) photographs a vehicle waiting to enter the warehouse (step S900). Inside the edge computer 202, it is determined whether the vehicle is full, empty, or crowded, and the presence or absence of the last vehicle in the warehouse is displayed (step S950). In addition, the number of vehicles waiting to enter the warehouse is counted (step S960). Then, the display panel displays whether the vehicle is full, vacant, or crowded, and whether or not there are waiting vehicles (step S970). In addition, the cloud server displays information on fullness, vacancy, congestion, and the presence or absence of waiting vehicles on the homepage (step S980). Data is also collected (step S980).
This allows vehicles attempting to enter the parking lot to be informed of the parking lot situation.

Embodiment 11.
Next, when applying the object identification device to a parking lot management device, in a parking lot where the parking lot is divided into multiple areas such as a multi-story parking lot, each of the multiple parking areas is full or empty. By notifying the extent of the situation, congestion can be prevented.
FIG. 13 is an explanatory diagram showing the configuration and flow of a parking lot management device using an object identification device.

In this Embodiment 11, the state of fullness, vacancy, and congestion for each of a plurality of parking areas (a plurality of floors) is determined (step S1050). The number of vehicles on each floor is counted (step S1060), and the state of fullness, vacancy, and congestion on each floor is displayed on the display panel (step S1070, the cloud server displays the state of fullness, vacancy, and congestion on each floor on the homepage. (Step S1080).Furthermore, data for each floor is aggregated (Step S1090).
As a result, by looking at the fullness rate on the display panel, it is possible to know which floors and parking lots have fewer vehicles compared to the parking spaces, so it is possible to eliminate congestion on one floor or parking lot.

Embodiment 12.
This embodiment applies an object identification device to a parking lot management device to achieve the purpose of obtaining information on the usage status of a parking lot.
In this twelfth embodiment, a parking frame used by a vehicle is determined based on image information of a vehicle parked above a parking frame in a parking lot. For this determination, the object identification device has a function of recognizing a vehicle as an object to be identified using deep learning, as described in the first embodiment. In addition, coordinates are set that subdivide the area of the parking slot in the parking lot, so that it can be determined whether the vehicle is parked within the parking slot or outside the parking slot. ing. This judgment can be learned by repeating learning based on the coordinate system representing the parking space and the vehicle appearance information, and the reliability of the judgment can be improved.

FIG. 14 is an explanatory diagram showing a configuration flow of a parking lot management device using an object recognition device.
As shown in FIG. 14, the imaging device 1 (camera) photographs a parking lot. Through this photographing, the parking slot in the parking lot and the vehicle being used are photographed (step S200). Inside the edge computer 202, the degree of confidence that the detected vehicle is parked in a predetermined parking space is predicted by overlapping predictions of parking space usage using deep learning (step S1000). When the reliability of prediction that a vehicle is using the parking slot is higher than a predetermined threshold, it is determined that a vehicle is parked in the parking slot (step S1010). In addition, the edge computer 202 calculates the optimal height and installation angle (depression angle) of the imaging device 1 (camera) in order to obtain data related to the parking slot (step S1020).
Then, the presence or absence of a stopped vehicle is displayed on the display panel (step S1170), and the presence or absence of a stopped vehicle is displayed on the cloud server's home page (step S1180). Additionally, information on vehicles stopped in the vehicle slots is notified to the parking lot manager (step S1030).

Embodiment 13.
When applying an object identification device to a vehicle management device, by processing images on the road to determine the presence or absence of stopped vehicles using deep learning object detection technology, it is possible to detect objects faster than humans can. This makes it possible to detect stopped vehicles.
FIG. 15 is an explanatory diagram showing the configuration and flow of a vehicle management device using an object identification device.
In this twelfth embodiment, the situation on the road is monitored with a camera, and vehicles are detected based on the appearance or part of the outline of the vehicle, and stopped vehicles are detected using the deep learning technology of the first embodiment. It is something.

As shown in FIG. 15, the imaging device 1 (camera) photographs the road (step S1100). Inside the edge computer 202, it is determined whether the vehicle is full, empty, or congested, and the direction and speed of vehicles traveling on the road are determined (step S1150). Furthermore, the presence or absence of a stopped vehicle is determined in consideration of the speed difference with surrounding vehicles (step S1160). Then, the presence or absence of a stopped vehicle is displayed on the display panel (step S1170), and the presence or absence of a stopped vehicle is displayed on the cloud server's home page (step S1180). Also, the information on the stopped vehicle is notified to the road administrator (step S1190).
As a result, the occurrence of rear-end collisions can be suppressed by shortening the time from the occurrence of a stopped vehicle to the arrival of a road administrator.

Although this application describes various exemplary embodiments and examples, various features, aspects, and functions described in one or more embodiments may be applicable to a particular embodiment. The present invention is not limited to, and can be applied to the embodiments alone or in various combinations.
Accordingly, countless variations not illustrated are envisioned within the scope of the technology disclosed herein. For example, this includes cases where at least one component is modified, added, or omitted, and cases where at least one component is extracted and combined with components of other embodiments.

1 Imaging device, 2 First storage device, 3 Second storage device, 4 Information processing device, 5 Identification device, 6 Display device, 7 Identification target object setting section, 8 Teacher data creation section, 9 Similar data creation section , 10 learning processing device, 11 learning processing means, 12 object identification means, 20 truck, 30 vehicle, 100 object identification device, 200 processor, 201 storage device, 202 edge computer

Claims (1)

an imaging device that acquires image information; a first storage device that stores information on characteristic parts of the vehicle; a second storage device that stores the image information acquired by the imaging device; The image information is divided into a plurality of regions, and the vehicle is included in the image information based on the image information of each of the divided regions and the information of the characteristic part of the vehicle stored in the first storage device. An object identification device equipped with an information processing device for determining whether a vehicle is present as a parking lot management device is provided as a parking lot management device, and when the presence of a vehicle is unknown because it is obscured by a blocking object in an identification area by the object identification device, 2. A parking lot management device characterized in that the image information stored in the storage device No. 2 is retrospectively checked to confirm whether or not the vehicle exists behind the shielding object.

Images