Parking position identification method, parking position learning method, parking position identification system, parking position learning device, and non-transitory recording medium for recording program

A parking space learning method in which a computer learns a first parking space of a target vehicle, the parking space learning method comprising:

• acquiring training data, in which a first annotation as information which indicates a second parking space for one vehicle in a parking region is added to an image which indicates the parking region which has a width in which parking of at least one vehicle is feasible, and in which a second annotation as information which indicates whether or not a first parking obstacle that is an obstacle to parking is present in the second parking space for the one vehicle is further added to the image which indicates the parking region;
• and
• constructing a learning model by learning the second parking space for the one vehicle in the parking region by using the training data, and by further learning whether or not the first parking obstacle is present in the second parking space for the one vehicle by using the training data,
• wherein the learning model identifies the first parking space of the target vehicle without identifying lines that serve as marks for parking vehicles in the parking region, and
• the learning model identifies whether or not a second parking obstacle is present in the first parking space of the target vehicle.

A parking space identification device that identifies a first parking space of a target vehicle, the parking space identification device comprising:

• a processor;
• and
• a memory that stores at least one program that, when executed by the processor, causes the processor to perform operations including:
• acquiring input data of an image which is generated by photographing a parking area by a camera which is installed in the target vehicle;
• and
• identifying the first parking space of the target vehicle in the photographed parking area by inputting the input data to a learning model which indicates a relationship between a parking region which has a width in which parking of at least one vehicle is feasible and a second parking space for one vehicle in the parking region,
• wherein the learning model identifies the first parking space of the target vehicle in the photographed parking area without identifying lines that serve as marks for parking vehicles in the parking region, and
• the learning model is constructed by:
• acquiring training data, in which a first annotation as information which indicates the second parking space for the one vehicle in the parking region is added to a second image which indicates the parking region which has the width in which parking of the at least one vehicle is feasible, and in which a second annotation as information which indicates whether or not a first parking obstacle that is an obstacle to parking is present in the second parking space for the one vehicle is further added to the second image which indicates the parking region;
• and
• learning the second parking space for the one vehicle in the parking region by using the training data, and by further learning whether or not the first parking obstacle is present in the second parking space for the one vehicle by using the training data, and
• in the identifying of the first parking space of the target vehicle, the input data are input to the learning model to identify whether or not a second parking obstacle is present in the first parking space of the target vehicle.

A parking space identification system that identifies a first parking space of a target vehicle, the parking space identification system comprising:

• a processor;
• and
• a memory that stores at least one program that, when executed by the processor, causes the processor to perform operations including:
• acquiring training data, in which a first annotation as information which indicates a second parking space for one vehicle in a parking region is added to a second image which indicates the parking region which has a width in which parking of at least one vehicle is feasible, and in which a second annotation as information which indicates whether or not a first parking obstacle that is an obstacle to parking is present in the second parking space for the one vehicle is further added to the second image which indicates the parking region;
• constructing a learning model by learning the second parking space for the one vehicle in the parking region by using the training data, and by further learning whether or not the first parking obstacle is present in the second parking space for the one vehicle by using the training data;
• acquiring input data of a first image which is generated by photographing a parking area by a camera which is installed in the target vehicle;
• and
• identifying the first parking space of the target vehicle in the photographed parking area by inputting the input data to the learning model,
• wherein the learning model identifies the first parking space of the target vehicle in the photographed parking area without identifying lines that serve as marks for parking vehicles in the parking region, and
• in the identifying of the first parking space of the target vehicle, the input data are input to the learning model to identify whether or not a second parking obstacle is present in the first parking space of the target vehicle.

A parking space learning device that learns a first parking space of a target vehicle, the parking space learning device comprising:

• a processor;
• and
• a memory that stores at least one program that, when executed by the processor, causes the processor to perform operations including:
• acquiring training data, in which a first annotation as information which indicates a second parking space for one vehicle in a parking region is added to an image which indicates the parking region which has a width in which parking of at least one vehicle is feasible, and in which a second annotation as information which indicates whether or not a first parking obstacle that is an obstacle to parking is present in the second parking space for the one vehicle is further added to the image which indicates the parking region;
• and
• constructing a learning model by learning the second parking space for the one vehicle in the parking region by using the training data, and by further learning whether or not the first parking obstacle is present in the second parking space for the one vehicle by using the training data,
• wherein the learning model identifies the first parking space of the target vehicle without identifying lines that serve as marks for parking vehicles in the parking region, and
• the learning model identifies whether or not a second parking obstacle is present in the first parking space of the target vehicle.

A computer-readable non-transitory recording medium that records a program of learning a first parking space of a target vehicle,

• the program being executable by a computer and causing the computer to execute a method, the method comprising:
• acquiring training data, in which a first annotation as information which indicates a second parking space for one vehicle in a parking region is added to an image which indicates the parking region which has a width in which parking of at least one vehicle is feasible, and in which a second annotation as information which indicates whether or not a first parking obstacle that is an obstacle to parking is present in the second parking space for the one vehicle is further added to the image which indicates the parking region;
• and
• constructing a learning model by learning the second parking space for the one vehicle in the parking region by using the training data, and by further learning whether or not the first parking obstacle is present in the second parking space for the one vehicle by using the training data,
• wherein the learning model identifies the first parking space of the target vehicle without identifying lines that serve as marks for parking vehicles in the parking region, and
• the learning model identifies whether or not a second parking obstacle is present in the first parking space of the target vehicle.