Management program and management device of parking lot

Abstract

To provide a management program and a management device of a parking lot for setting a parameter value of a parking fee to estimate sales.SOLUTION: At a step 1201, an operator sets parameter values P, Q, R, S of a parking fee for a future period V0 of a weekday. At a step 1202, sales W1 per one day are estimated and extracted by using a learning control model for the weekday, by using the parameter values P, Q, R, S (=PV0, QV0, RV0, SV0) of the parking fee set at the step 1201 as input data. At a step 1203, the optimum sales per one day, for example, the greatest sales per one day among the sales W1, W2, ... per one day of the future periods V0, V1, ... are selected. At a step 1204, the parameter value of the parking fee with respect to the sales per one day selected at the step 1203 is set as a new parking fee for the weekday.SELECTED DRAWING: Figure 12A

Description

[Technical field]

[0001] The present invention relates to a parking lot management program and management device for setting parameter values ​​for parking fees and estimating sales. [Background technology]

[0002] Generally, parking fees for a parking lot have two or more parameters, for example four parameters: maximum time, maximum fee, daytime fee, and nighttime fee, and the values ​​of each parameter are set, for example, 12 hours, 1,200 yen, 20 minutes / 100 yen, and 60 minutes / 100 yen.

[0003] Conventionally, parameter values ​​for parking fees at parking lots have been set based on the market price of land for the parking lot, the accessibility of the parking lot, and parameter values ​​for parking fees at surrounding parking lots.

[0004] Incidentally, a method for authenticating a vehicle's license plate number is publicly known (see Patent Document 1). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2022-54632 A (Patent No. 7277879) Summary of the Invention [Problem to be solved by the invention]

[0006] However, the parameter value setting for the parking fee of a parking lot has a large impact on revenue. In other words, if the parking fee based on the parameter value is too high, the number of customers using the parking lot will decrease, and the parking lot's sales will decline. On the other hand, if the parking fee based on the parameter value is too low, the number of customers using the parking lot will increase, but the parking lot's sales will not increase. Therefore, the parameter value setting for the above-mentioned conventional parking fee is done based on the operator's experience, but it is difficult to set the parameter value quickly, especially when the parking lot has a large capacity. As a result, there is a problem that the parking lot's sales cannot be predicted quickly and cannot be optimized. [Means for solving the problem]

[0007] In order to solve the above-mentioned problems, the parking lot management program of the present invention uses a learned learning control model with parameter values ​​of the parking lot's parking fees for multiple past periods as input teacher data and the parking lot's daily sales for each of the past periods as output teacher data, and causes a computer to execute a sales estimation procedure that uses parameter values ​​of the parking lot's parking fees for a future period as input data to estimate the parking lot's daily sales for the future period.

[0008] In addition, the parking lot management device of the present invention is equipped with a sales estimation means that uses a learned learning control model with parameter values ​​of the parking lot's parking fees for multiple past periods as input teacher data and the parking lot's daily sales for each of the past periods as output teacher data, and estimates the parking lot's daily sales for a future period using parameter values ​​of the parking lot's parking fees for a future period as input data. Effect of the Invention

[0009] According to the present invention, parameter values ​​for future parking fees can be quickly set and revenue can be quickly estimated, which contributes to optimal revenue for parking lots. [Brief description of the drawings]

[0010] [Figure 1]FIG. 1 is a diagram for explaining the principle of the present invention, in which (A) shows an example of parameter values ​​for parking fees, (B) is a graph showing an example of sales distribution against usage time, and (C) is a graph showing an example of sales distribution against entry time. [Diagram 2] 1 shows a parking lot including an embodiment of a management device according to the present invention; [Diagram 3] 3 is a detailed block circuit diagram of the management device of FIG. 2. [Figure 4] 4 is a flowchart for explaining a parameter value change routine for parking fees in the central processing unit of FIG. 3. [Diagram 5] 4 is a flowchart for explaining a vehicle front image learning operation of the central processing unit of FIG. 3. [Figure 6] 4 is a flowchart for explaining a vehicle rear image learning operation of the central processing unit of FIG. 3. [Figure 7] 4 is a flowchart for explaining a main operation of the central processing unit of FIG. 3. [Figure 8] 13 is a table showing parameter values ​​and management orders for past period parking fees on weekdays. [Figure 9] 13 is a table showing parameter values ​​and management orders for parking fees for past holiday periods. [Figure 10A] 4 is a flowchart for explaining a weekday sales learning operation of the central processing unit of FIG. 3. [Figure 10B] 10B is a table showing the relationship between past periods and daily sales for the past periods used in the weekday sales learning operation of FIG. 10A. [Figure 11A] 4 is a flowchart for explaining a holiday sales learning operation of the central processing unit of FIG. 3. [Figure 11B] 11B is a table showing the relationship between past periods and daily sales for the past periods used in the holiday sales learning operation of FIG. 11A. [Figure 12A] 4 is a flowchart for explaining weekday future period sales estimation / parking fee setting by the central processing unit of FIG. 3; [Figure 12B]12B is a table showing the relationship between a future period and estimated daily sales for the future period used in the weekday sales estimation operation of FIG. 12A. [Figure 13A] 4 is a flowchart for explaining the future period holiday sales estimation / parking fee setting of the central processing unit of FIG. 3; [Figure 13B] 13B is a table showing the relationship between a future period and estimated daily sales for the future period used in the holiday sales estimation operation of FIG. 13A. [Figure 14] 4 is a block diagram showing another example of the parking lot management device of FIG. 3. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] First, the principle of the present invention will be described with reference to Fig. 1. That is, the setting of parameter values ​​of parking fees and their correlation with total daily sales will be described.

[0012] Parking fees consist of multiple parameters, for example, four parameters: maximum time P after entry, maximum fee Q within maximum time P, fee R per unit time during the day (8:00-22:00), and fee S per unit time at night (22:00-8:00).

[0013] FIG. 1(A) shows an example of parameter setting for parking fee on weekdays (Monday to Friday) in a parking lot that can accommodate 100 cars. In other words, the maximum time P is 12 hours, the maximum fee Q is 1,200 yen, the daytime (8:00 to 22:00) fee R is 20 minutes / 100 yen (= 5 yen / 1 minute), and the nighttime (22:00 to 8:00) fee S is 60 minutes / 100 yen (= 5 / 3 yen / 1 minute). In this case, the sales distribution for the usage time is as shown in FIG. 1(B), and the sales distribution for the entry time is as shown in FIG. 1(C), and the total sales for one day is about 100,000 yen. If the maximum time P, the maximum fee Q, the daytime fee R per unit time, and the nighttime fee S per unit time are changed, the sales distribution for the usage time shown in FIG. 1(B) and the sales distribution for the entry time shown in FIG. 1(C) are expected to change. For example, if the maximum time P is reduced, it is predicted that there will be no effect according to the sales distribution for usage time shown in Figure 1(B), but sales are predicted to increase according to the sales distribution for entry time shown in Figure 1(C). Also, if the daytime rate R per unit time is increased, it is predicted that there will be no effect according to the sales distribution for entry time shown in Figure 1(C), but it is predicted that the number of customers who use the parking lot for about an hour will decrease, resulting in a decrease in sales, according to the sales distribution for usage time shown in Figure 1(B). In this way, changing the parameter values ​​for parking fees leads to fluctuations in customers, and sales trends become complicated, especially when the parking lot has a large capacity.

[0014] The present invention aims to learn daily sales based on parameter values ​​of parking fees for past periods, estimate (predict) daily sales for parameter values ​​of parking fees for future periods, and achieve optimal daily sales.

[0015] FIG. 2 is a diagram showing a parking lot including an embodiment of a management device according to the present invention.

[0016] In FIG. 2, six vehicle compartments 10-1, 10-2, ..., 10-6 are provided on one side of the parking lot 10, and six vehicle compartments 10-7, 10-8, ..., 10-12 are provided on the other side of the parking lot 10. The parking lot 10 is also provided with an unregulated entrance / exit 10-3 and a parking fee signboard 10-14. A management device 1 is provided at one end of the parking lot 10. At least one camera (e.g., a wide-angle monocular camera) 2 is fixed to a pole (not shown) at a height of, for example, 3 m, so that the entire parking lot 10 can be imaged at once. That is, the vehicles in the parking lot 10 and the vehicle compartments 10-1 to 10-12 can be imaged at once. In particular, the camera can image the front of the vehicles parked in the vehicle compartments 10-1 to 10-6 and the rear of the vehicles parked in the vehicle compartments 10-7 to 10-12. The camera 2 is connected to the management device 1. A payment machine 1a located in the parking lot 10 or close to but distant from the parking lot 10 is connected to the management device 1. Furthermore, the management device 1 is connected to an information terminal 4 and the like via the Internet 3 as necessary. An operator can operate the management device 1 via the information terminal 4 or directly.

[0017] FIG. 3 is a detailed block circuit diagram of the management device 1 of FIG.

[0018] In FIG. 3, the management device 1 is configured by a computer, and in detail, includes a central processing unit (CPU) 11, a read-only memory (ROM) 12 for storing programs, a flash memory 13, a random access memory (RAM) 14 for storing temporary data, an input / output (I / O) interface 15 having a USB (registered trademark) terminal, interrupt terminals INT0, INT1, INT1', INT2, INT2', INT3, INT3', etc., an image interface 16 connected to the camera 2, a communication interface 17 connected to the Internet 3, a vehicle front image learning control model 18, a vehicle rear image learning control model 18', a weekday learning control model 19, a holiday learning control model 19', etc. Each learning control model 18, 18', 19, 19' is a neural network by deep learning consisting of an input layer, multiple intermediate layers, and an output layer. In the neural network, learning is performed by adjusting the weights and biases of a multilayer perceptron by the error backpropagation method.

[0019] 4, 5, 6 and 7, the normal management operation of the management device 1 in FIG. 3 will be described.

[0020] Fig. 4 is a flow chart showing a parking fee change routine executed by CPU 11 in response to an interrupt by an operator via terminal INT0 in Fig. 3. That is, when the operator changes the parking fee, he or she redraws parking fee sign 10-14 and the parameter value change routine for parking fee in Fig. 4 is executed.

[0021] First, in step 401, the contents of the parking fee sign 10-14 are input as parameter values ​​of the new parking fee into the flash memory 13. For example, the parameter values ​​of the weekday parking fee are P: Maximum time after receipt Q: Maximum charge within maximum time R: Daytime (8:00~22:00) hourly rate S: Nighttime (22:00~8:00) charge per unit time and for the parameter value of the parking fee on holidays (Saturday, Sunday, and public holidays), P': Maximum time after receipt Q': Maximum charge within maximum time R': Daytime (8:00~22:00) charge per unit time S': Nighttime (22:00~8:00) charge per unit time In this case, if the maximum fee is not adopted, enter P=P'=0 and Q=Q'=0. Also, if only the maximum fee is adopted, enter R=R'=0 and S=S'=0. Most parking fees can be expressed by the four parameter values ​​P, Q, R, S for weekdays and the four parameter values ​​P', Q', R', S' for holidays. Holiday fees are set higher, for example, for parking lots close to event venues. If holiday fees are not necessary, delete P', Q', R', S'. This establishes the new parking fees.

[0022] Next, in step 402, an end date is entered into the old parking fee in the flash memory 13 to end the use of the old parking fee.

[0023] Next, in step 403, a start date is entered into the new parking fee in the flash memory 13 to start using the new parking fee.

[0024] Then, in step 404, the routine of FIG.

[0025] 5 shows a vehicle front image learning operation routine executed by the CPU 11 in response to an interrupt from the operator via the terminal INT1. This learning operation routine is stored in the ROM 12 or the flash memory 13.

[0026] First, in step 501, a plurality of learning vehicle front image data recognized by the operator is acquired from a vehicle front camera (not shown) or a USB memory previously acquired, etc. The more the recognized learning vehicle front image data is, the greater the learning effect in step 502.

[0027] In step 502, a vehicle front image learning process is performed. That is, a vehicle front image learning control model 18 that extracts feature amounts using a large number of recognized learning vehicle front image data as input data is trained by deep learning.

[0028] Then, the routine of FIG.

[0029] 6 shows a vehicle rear image learning operation routine executed by the CPU 11 in response to an interrupt from the operator via the terminal INT1'. This learning operation routine is stored in the ROM 12 or the flash memory 13.

[0030] First, in step 601, a plurality of learning rear image data of the vehicle recognized by the operator is acquired from a vehicle rear camera (not shown) or a USB memory previously acquired, etc. The greater the number of recognized learning rear image data of the vehicle, the greater the learning effect in step 602.

[0031] In step 602, a vehicle rear image learning process is performed. That is, a vehicle rear image learning control model 18′ that extracts features using a large number of recognized learning vehicle rear image data as input data is trained by deep learning.

[0032] Then, the routine of FIG.

[0033] 5 and 6 can be executed by the operator at any time thereafter, which will increase the learning effect. Also, the learning operations can be executed on the cloud side connected to the Internet 3, and the learning results can be introduced into the learning control models 18, 18' via the Internet 3.

[0034] 7 shows a main routine executed by the CPU 11. This main routine is stored in the ROM 12 or the flash memory 13.

[0035] First, in step 701, an image of the entire parking lot 10, and in particular an image of the vehicles present in the parking lot 10, is acquired by the camera 2.

[0036] Next, in step 702, the trained vehicle front image learning control model 18, which extracts features using the recognized learning vehicle front image data as input data, is used to input image data from the camera 2 and determine whether the vehicle front image data has been extracted. In this case, the vehicle front image data is extracted together with the vehicle number whether the vehicle is inside or outside the vehicle. If the vehicle front image data has been extracted, the process proceeds to step 703, and if the vehicle front image data has not been extracted, the process proceeds to step 704.

[0037] Next, in step 703, the vehicle number is recognized from the vehicle front image data extracted in step 702. That is, the vehicle number recognition process includes a black and white binary data conversion procedure for converting the vehicle front image data extracted in step 702 into black and white binary data, a rectangular contour area extraction procedure for extracting connected contours of connected black data or white data of the black and white binary data as a plurality of rectangular contour areas, a number and dot area extraction procedure for determining and extracting a plurality of number areas and dot areas from the plurality of rectangular contour areas, and a vehicle number extraction procedure for extracting the vehicle number of the vehicle from the number area and dot area. The number and dot area extraction procedure includes a Boolean image conversion procedure for converting an image within each rectangular contour area into a Boolean image, and a Boolean image and a Boolean number template are compared by pixel-by-pixel Boolean operation to extract each rectangular contour area when the similarity between the Boolean image and the Boolean number template is equal to or greater than a first threshold or equal to or less than a second threshold that is smaller than the first threshold. The method has a number / dot discrimination and extraction procedure for discriminating that the contour area is a number / dot area and extracting it, and the vehicle number extraction processing procedure has an average value calculation procedure for calculating the average height of all the number / dot areas, a first selection procedure for selecting one of the number / dot areas as a first area, a second selection procedure for selecting a number / dot area within a range of a predetermined multiple of the average height on the right side of the first area as a second area, a third selection procedure for selecting a number / dot area between the first and second areas as a third area, a vehicle number candidate selection procedure for setting the first, second and third areas as vehicle number candidates when there is a predetermined number of third areas, and a vehicle number selection procedure for selecting the vehicle number candidate as the vehicle number of the vehicle when there is one vehicle number candidate, and selecting the vehicle number candidate that is most likely to satisfy a predetermined condition as the vehicle number of the vehicle when there is multiple vehicle number candidates (see Patent Document 1).

[0038] Next, in step 704, the trained vehicle rear image learning control model 18', which extracts features using the recognized learning vehicle rear image data as input data, is used to input image data from the camera 2 and determine whether vehicle rear image data has been extracted. At that time, if the vehicle is outside the passenger compartment, it is extracted together with the vehicle number, but if the vehicle is completely inside the passenger compartment, it is difficult to extract the vehicle number. Even in this case, when the vehicle starts entering the passenger compartment, a vehicle rear image including the vehicle number is extracted. If vehicle rear image data has been extracted, the process proceeds to step 705, and if vehicle rear image data has not been extracted, the process proceeds directly to step 706.

[0039] Next, in step 705, the vehicle number is recognized from the vehicle rear image data extracted in step 704. That is, the vehicle number recognition process includes a black and white binary data conversion procedure for converting the vehicle rear image data extracted in step 704 into black and white binary data, a rectangular contour area extraction procedure for extracting connected contours of connected black data or white data of the black and white binary data as a plurality of rectangular contour areas, a number and dot area extraction procedure for determining and extracting a plurality of number areas and dot areas from the plurality of rectangular contour areas, and a vehicle number extraction procedure for extracting the vehicle number of the vehicle from the number areas and dot areas. The number and dot area extraction procedure includes a Boolean image conversion procedure for converting an image within each rectangular contour area into a Boolean image, and a Boolean image and a Boolean number template are compared by pixel-by-pixel Boolean operation to compare each rectangular contour area with each other when the similarity between the Boolean image and the Boolean number template is equal to or greater than a first threshold or equal to or less than a second threshold that is smaller than the first threshold. The method has a number / dot discrimination and extraction procedure for discriminating that the contour area is a number / dot area and extracting it, and the vehicle number extraction processing procedure has an average value calculation procedure for calculating the average height of all the number / dot areas, a first selection procedure for selecting one of the number / dot areas as a first area, a second selection procedure for selecting a number / dot area within a range of a predetermined multiple of the average height on the right side of the first area as a second area, a third selection procedure for selecting a number / dot area between the first and second areas as a third area, a vehicle number candidate selection procedure for setting the first, second and third areas as vehicle number candidates when there is a predetermined number of third areas, and a vehicle number selection procedure for selecting the vehicle number candidate as the vehicle number of the vehicle when there is one vehicle number candidate, and selecting the vehicle number candidate that is most likely to satisfy a predetermined condition as the vehicle number of the vehicle when there is multiple vehicle number candidates (see Patent Document 1).

[0040] Next, in step 706, it is determined whether the vehicle whose vehicle number was recognized in steps 703 and 705 is in any vehicle compartment. For example, it is determined that the vehicle compartment is full due to the vehicle based on the overlap ratio between the vehicle's projection data and the vehicle compartment area data obtained by Boolean calculation. Here, if the vehicle compartment area data is 100%, when the overlap ratio of the vehicle's projection data exceeds 30%, the vehicle compartment is determined to be full, and when the overlap ratio is less than 30%, the vehicle compartment is determined to be vacant. The vehicle projection data is preferably vertical projection data, but is not limited to this.

[0041] Finally, in step 707, the management order is started or updated in the flash memory 13 or RAM 14. For example, the management order is composed of the entry time, vehicle number, vehicle compartment number, vehicle front image data, vehicle rear image data, and departure time (last time) for each management order number.

[0042] Additionally, the settlement amount of the parking fee is recorded in each management order. That is, settlement is carried out by each user operating the vehicle number of the vehicle at the settlement machine 1a, and as a result, the settlement amount is recorded in each management order. Therefore, the validity period, parking fee parameter value, and management order (management order number, entry time, settlement amount, ...) are linked by the past periods T0, T1, ..., T13; T0', T1', ..., T13' as shown in Figures 8 and 9. The validity period to which a management order belongs is determined by the entry time.

[0043] Next, learning of parking lot sales over the past period will be described.

[0044] 10A shows a weekday sales learning operation routine executed by the CPU 11 in response to an interrupt from the operator via the terminal INT2. This learning operation routine is stored in the ROM 12 or the flash memory 13.

[0045] First, in step 1001, parameter values ​​P, Q, R, and S of parking fees for one of the past periods T0, T1, . . . , T13 on weekdays shown in FIG. 8 are obtained.

[0046] Next, in step 1002, the management orders belonging to the parking fee parameter values ​​P, Q, R, and S acquired in step 1001 are used to calculate the sales per day.

[0047] Next, in step 1003, a weekday sales learning process is performed. That is, the parameter values ​​P, Q, R, and S of the parking fee acquired in step 1001 are used as input teacher data, and the daily sales acquired in step 1002 are used as output teacher data to train the weekday learning control model 19 by deep learning.

[0048] Then, the routine of FIG.

[0049] In this way, by executing the routine of FIG. 10A for each past period, daily sales U1, U2, ..., U13 for past periods T0, T1, ..., T13 are obtained as shown in FIG. 10B, and learning control is performed based on these.

[0050] 11A shows a holiday sales learning operation routine executed by the CPU 11 in response to an interrupt from the operator via the terminal INT2'. This learning operation routine is stored in the ROM 12 or the flash memory 13.

[0051] First, in step 1101, parameter values ​​P', Q', R', and S' of parking fees for one of the past holiday periods T0', T1', . . . , T13' shown in FIG. 9 are obtained.

[0052] Next, in step 1102, the management orders belonging to the parking fee parameter values ​​P', Q', R', and S' acquired in step 1101 are used to calculate the sales per day.

[0053] Next, in step 1103, holiday sales learning processing is performed. That is, the parking fee parameter values ​​P', Q', R', and S' obtained in step 1101 are used as input teacher data, and the daily sales obtained in step 1102 are used as output teacher data to train holiday learning control model 19' by deep learning.

[0054] Then, in step 1104, the routine of FIG.

[0055] In this way, by executing the routine of FIG. 11A for each past period, daily sales U1', U2', ..., U13' for past periods T0', T1', ..., T13' are obtained as shown in FIG. 11B, and learning control is performed based on this.

[0056] The learning operations in Fig. 10A and Fig. 11A can be executed by the operator at any time thereafter, and the learning effect is increased. In particular, the learning effect is increased when daily sales based on management orders with different parking fee parameter values ​​are used. Also, the operations may be executed on the cloud side connected to the Internet 3, and the learning results may be introduced into the learning control models 18, 18' via the Internet 3.

[0057] Next, we will explain the estimation (forecast) of future parking lot sales / parking fee setting.

[0058] 12A shows a weekday sales estimation / parking fee setting routine executed by the CPU 11 in response to an interrupt from the operator via the terminal INT3. This estimation / setting routine is stored in the ROM 12 or the flash memory 13.

[0059] First, in step 1201, the operator sets the parameter values ​​P, Q, R, and S of the parking fee for a future period V0 on weekdays to, for example, PV0, QV0, RV0, and SV0. However, the parameter values ​​of the parking fee may be automatically changed to the parameter values ​​of the parking fee for the current weekday.

[0060] Next, in step 1202, the parking fee parameter values ​​P, Q, R, S (=PV0, QV0, RV0, SV0) set in step 1201 are used as input data to estimate and extract daily sales W1 using weekday learning control model 19.

[0061] In this way, the parking lot's daily sales can be predicted by setting the parameter values ​​P, Q, R, and S of the parking fee for weekdays in the future period. Therefore, as shown in Fig. 12B, the parameter values ​​of the parking fee for multiple future periods V0, V1, ... can be predicted as (PV0, QV0, RV0, SV0), (PV1, QV1, RV1, SV1), ....

[0062] Next, in step 1203, the optimum daily sales, for example the maximum daily sales, is selected from the daily sales W1, W2, . . . for the future periods V0, V1, .

[0063] Next, in step 1204, the parameter value of the parking fee for the daily sales selected in step 1203 is set as the new parking fee for weekdays. In this case, the parking fee sign 10-14 in FIG. 2 is also rewritten.

[0064] Then, in step 1205, the routine of FIG. 12A ends.

[0065] 13A shows a holiday sales estimation / parking fee setting routine executed by the CPU 11 in response to an interrupt from the operator via the terminal INT3. This estimation / setting routine is stored in the ROM 12 or the flash memory 13.

[0066] First, in step 1301, the operator sets the parking fee parameter values ​​P', Q', R', and S' for a future holiday period V0' to, for example, PV0', QV0', RV0', and SV0'. However, these parking fee parameter values ​​may be automatically changed to the parameter values ​​for the current holiday parking fee.

[0067] Next, in step 1302, the parking fee parameter values ​​P', Q', R', S' (= PV0', QV0', RV0', SV0') set in step 1301 are used as input data to estimate and extract daily sales W1' using holiday learning control model 19'.

[0068] In this way, the parking lot's daily sales can be predicted by setting the parameter values ​​P', Q', R', and S' of the holiday parking fees for the future period. Therefore, as shown in Fig. 13B, the parameter values ​​of the parking fees for multiple future periods V0', V1', ... can be predicted as (PV0', QV0', RV0', SV0'), (PV1', QV1', RV1', SV1'), ...

[0069] Next, in step 1303, the optimum daily sales, for example the maximum daily sales, is selected from the daily sales W1', W2', . . . of the future periods V0', V1', .

[0070] Next, in step 1304, the parameter value of the parking fee for the daily sales selected in step 1303 is set as the new parking fee for the holiday. In this case, the parking fee sign 10-14 in FIG. 2 is also rewritten.

[0071] Then, in step 1305, the routine of FIG. 13A ends.

[0072] In this way, the present invention quickly estimates and provides future daily revenue based on daily revenue based on the parameter value of the past period parking fee, and is particularly applicable to parking lots that can accommodate tens of millions of cars.

[0073] The management device of FIG. 3 can also be configured with hardware shown in FIG.

[0074] In Figure 14, the parking lot management device has a sales estimation means 1402 that uses a trained learning control model 1401 with parameter values ​​P, Q, R, and S of the parking lot's parking fees for multiple past periods as input teacher data and the parking lot's daily sales for each of the past periods as output teacher data, and estimates the parking lot's daily sales for a future period using parameter values ​​P, Q, R, and S of the parking lot's parking fees for a future period as input data.

[0075] In the above embodiment, the parking lot has an unregulated entrance and exit gate and recognizes vehicle numbers, but the present invention is not limited to this. For example, the parking lot may have a gate machine with a bar at the entrance and exit gate and a photography system. In addition, the parameters of the parking fee are maximum time P, maximum fee Q, daytime fee R, and nighttime fee S, but are not limited to these. Furthermore, while weekday parking fees and holiday parking fees are set, parking fees that are not separated for weekdays and holidays may be adopted.

[0076] Furthermore, the present invention can be applied to any modifications within the obvious scope of the above-described embodiment. [Industrial Applicability]

[0077] The vehicle entry / exit management program and device for a parking lot according to the present invention can be used in a parking lot billing system, etc. [Explanation of symbols]

[0078] 1: Management device 1a:Payment machine 2: Camera 3: Internet 4: Information terminal 10. Parking lot 10-1, 10-2,…, 10-12: Vehicle interior 10-13: Unregulated entrance / exit gate 10-14: Parking fee sign 11: CPU 12:ROM 13: Flash memory 14: RAM 15: I / O interface 16: Image interface 17: Communication interface 18: Vehicle forward image learning control model 18': Vehicle rear image learning control model 19: Weekday learning control model 19': Holiday learning control model

Claims

1. A parking lot management program for causing a computer to execute a sales estimation procedure using a learned learning control model with parameter values ​​of a parking lot's parking fees for multiple past periods as input teacher data and the parking lot's daily sales for each of the past periods as output teacher data, and with parameter values ​​of the parking lot's parking fees for a future period as input data to estimate the parking lot's daily sales for the future period.

2. a sales estimation step of estimating daily sales for each future period of the parking lot using a learned learning control model with parameter values ​​of parking fees for a plurality of past periods of the parking lot as input teacher data and daily sales for each of the past periods of the parking lot as output teacher data, and using parameter values ​​of parking fees for a plurality of future periods of the parking lot as input data; an optimal sales selection step for selecting an optimal daily sales from the estimated daily sales for the plurality of future periods; a parking fee setting step of setting a parameter value of the parking fee for the future period when the selected daily sales are estimated as a new parking fee; A parking lot management program for causing a computer to execute the above.

3. 3. A parking lot management program executed by a computer according to claim 1 or 2, wherein the period is a weekday period.

4. 3. The parking lot management program executed by a computer according to claim 1 or 2, wherein the period is a holiday period.

5. 3. A parking lot management program executed by a computer according to claim 1, wherein the parameters of the parking fee are a maximum time, a maximum fee up to the maximum time, a daytime fee and a nighttime fee.

6. A parking lot management device comprising a sales estimation means for using a trained learning control model with parameter values ​​of parking fees for a parking lot for multiple past periods as input teacher data and daily sales for each of the past periods for the parking lot as output teacher data, and for estimating daily sales for the parking lot for a future period as input data for the parking fee parameter values ​​for the parking lot.

7. a sales estimation means for estimating the daily sales of the parking lot for each future period using a learned learning control model with parameter values ​​of parking fees for a plurality of past periods of the parking lot as input teacher data and daily sales of the parking lot for each of the past periods as output teacher data, and using parameter values ​​of parking fees for a plurality of future periods of the parking lot as input data; an optimal sales selection means for selecting an optimal daily sales from the plurality of estimated daily sales for future periods; a parking fee setting means for setting a parameter value of the parking fee for the future period at the time when the selected daily sales is estimated as a new parking fee; A parking lot management device comprising:

8. The parking lot management device according to claim 6 or 7, wherein the period is a weekday period.

9. The parking lot management device according to claim 6 or 7, wherein the period is a holiday period.

10. 8. A parking lot management device according to claim 6, wherein the parameters of the parking fee are a maximum time, a maximum fee up to the maximum time, a daytime fee and a nighttime fee.

Images