Bicycle parking lot management program and bicycle parking lot management device

Abstract

To provide a bicycle parking lot management program and a bicycle parking lot management device that can reduce facility costs and maintenance costs.SOLUTION: For images of a bicycle (123456) assigned an admission code 123456, the images are recorded intermittently as a video at date and time ts, t1, ..., t2, t3, in which case a human image is also recorded. At date and time t3 when the bicycle (123456) is parked, the image of the bicycle (123456) is a still image. In this case, there is usually no human image, and the location of the bicycle is recorded. In other words, the image of the bicycle (123456) from entry to parking is recorded. The bicycle (123456) is then recorded intermittently as a video at date and time t1'..., t2' te, and in this case, the human image is also recorded. In other words, the images of bicycle (123456) are recorded from de-parking to exiting.SELECTED DRAWING: Figure 11

Description

[Technical Field]

[0001] The present invention relates to a bicycle parking lot management program and a bicycle parking lot management device for managing a bicycle parking lot having a plurality of parking positions for parking two-wheeled vehicles such as bicycles. [Background technology]

[0002] In conventional bicycle parking lots, bicycle parking machines are installed at each of a number of bicycle parking locations for parking multiple bicycles. Each bicycle parking machine is composed of a lock / unlock mechanism that locks and unlocks the bicycle wheels to prevent theft, and a control unit that detects the entry of a bicycle and locks the lock / unlock mechanism, and detects the settlement of the bicycle's fee and unlocks the lock / unlock mechanism (see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2004-237782 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the conventional bicycle parking lots described above, it is necessary to provide bicycle parking machines with locking / unlocking mechanisms at each bicycle parking location, which results in high installation costs and maintenance costs. [Means for solving the problem]

[0005] In order to solve the above-mentioned problems, a bicycle parking lot management program according to the present invention is a bicycle parking lot management program for managing a bicycle parking lot having a plurality of bicycle parking positions, and includes an overall image acquisition procedure for acquiring overall image data of the bicycle parking lot, a motorcycle image / human image estimation and extraction procedure for estimating and extracting motorcycle images and human images using the overall image data as input data using a first learning control model that has been machine-learned with motorcycle images and human images as input teacher data and motorcycle information and human information as output teacher data, and Two-wheeled vehicles, including those in motion Video images of Sometimes, the motorcycle image and the person image that comes into contact with the motorcycle image are recorded together with the date and time. motorcycle From video images Two-wheeled vehicles, including stationary two-wheeled vehicles When the image changes to a still image, the motorcycle image is displayed together with the parking position where the motorcycle is located. motorcycle The first recording procedure is to cause a computer to execute a still image together with the date and time.

[0006] The bicycle parking lot management device according to the present invention is a bicycle parking lot management device for managing a bicycle parking lot having a plurality of bicycle parking positions, and includes: an overall image acquisition means for acquiring overall image data of the bicycle parking lot; a motorcycle image / human image estimation / extraction means for estimating and extracting motorcycle images and human images using the overall image data as input data and a first learning control model that has been machine-learned using motorcycle images and human images as input teacher data and motorcycle information and human information as output teacher data; Two-wheeled vehicles, including those in motion Video images of Sometimes, the motorcycle image and the person image that comes into contact with the motorcycle image are recorded together with the date and time. motorcycle From video images Two-wheeled vehicles, including stationary two-wheeled vehicles When the image changes to a still image, the motorcycle image is displayed together with the parking position where the motorcycle is located. motorcycle and a first recording means for recording the still image together with the date and time. [Effects of the Invention]

[0007] According to the present invention, the entry / exit of bicycles and the like at each parking location can be managed without installing bicycle parking machines consisting of locking / unlocking mechanisms at each parking location, thereby reducing equipment costs and maintenance costs. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a diagram showing a bicycle parking lot including an embodiment of a bicycle parking lot management device according to the present invention; [Figure 2] FIG. 2 is a detailed block circuit diagram of the bicycle parking lot management device of FIG. 1. [Figure 3] 3 is a flowchart for explaining an entry / exit processing routine executed by the central processing unit of FIG. 2. [Figure 4] FIG. 4 is a diagram for providing a supplementary explanation of the flowchart in FIG. 3. [Figure 5] 4 is a detailed flowchart of the entrance processing steps of FIG. 3. [Figure 6] FIG. 6 is a diagram for providing a supplementary explanation of the flowchart in FIG. 5. [Figure 7] 4 is a detailed flowchart of the exit processing step of FIG. 3. [Figure 8] FIG. 8 is a diagram for providing a supplementary explanation of the flowchart in FIG. 7. [Figure 9] 3 is a flowchart illustrating a main routine executed by the central processing unit of FIG. 2. [Figure 10] FIG. 10 is a diagram for providing a supplementary explanation of the flowchart in FIG. 9. [Figure 11] 10 is a diagram showing a bicycle trajectory image obtained by the flowcharts of FIGS. 3, 5, 7, and 9 and recorded in the flash memory of FIG. 2. FIG. [Figure 12] 2. FIG. 9 shows a trajectory image of a person obtained by step 914 of FIG. 9 and recorded in the flash memory of FIG. [Figure 13] 3 is a block diagram showing another example of the bicycle parking lot management device of FIG. 2. FIG. DETAILED DESCRIPTION OF THE INVENTION

[0009] FIG. 1 is a diagram showing a bicycle parking lot including an embodiment of a bicycle parking lot management device according to the present invention.

[0010] 1, bicycle parking lot 1 is provided with, for example, nine bicycle parking positions 11-A, 11-B, ..., 11-I marked with frame marks, and bicycle parking positions 11-A, 11-B, ..., 11-I are managed by bicycle parking lot management device 12. For example, bicycle parking positions 11-A, 11-B, 11-D, 11-F, 11-H, and 11-I are in a parked state, while bicycle parking positions 11-C, 11-E, and 11-G are in an empty state. For this reason, parking lot 1 is provided with an entrance / exit monitoring camera (e.g., a monocular camera) 13 near entrance / exit G and an overall monitoring camera (e.g., a monocular camera) 14. Entrance / exit monitoring camera 13 and overall monitoring camera 14 are connected to bicycle parking lot management device 2.

[0011] Near the entrance / exit G, there are provided a two-dimensional code (e.g., QR Code (registered trademark)) sign 15, a touch panel 16, and an available / unavailable sign 17 that indicates whether the bicycle parking lot is full or empty. The touch panel 16 and the available / unavailable sign 17 are connected to the bicycle parking lot management device 12.

[0012] The bicycle parking lot management device 12 is also connected to an information terminal 3, a payment server 4, etc. via the Internet 2. An operator can operate the bicycle parking lot management device 12 via the information terminal 3 or directly.

[0013] FIG. 2 is a detailed block circuit diagram of the bicycle parking lot management device 12 of FIG.

[0014] 2, the bicycle parking lot management device 12 is configured by a computer. In detail, it includes a central processing unit (CPU) 121, a read-on memory (ROM) 122 for storing programs and the like, a random access memory (RAM) 123 for storing temporary data and the like, a flash memory 124 for storing images and the like, and an input / output (I / O) interface connected to the vacant signboard 16, the touch panel 17, etc. 125, an image interface 126 connected to the entrance / exit monitoring camera 13 and the overall monitoring camera 14, a communication interface 127 connected to the Internet 2, learning control models 128 and 129, and the like.

[0015] The learning control models 128 and 129 are deep learning neural networks that are composed of an input layer, multiple intermediate layers, and an output layer.

[0016] The learning control model 128 is trained by deep learning (machine learning) using a large number of bicycle images, such as images of adult bicycles, children's bicycles, bicycles with baskets, bicycles with two rear wheels, bicycles with two front wheels, and electrically assisted bicycles, as input training data and bicycle information as output training data. The learning control model 128 also trains by deep learning (machine learning) using a large number of human images, such as adult males, adult females, child males, child females, people carrying luggage, people holding umbrellas, and front, side, and rear views of these people, as input training data and human information as output training data. This training can be performed in advance outside the bicycle parking lot 1, and the operator can input the learning results to the learning control model 128 using the information terminal 3. However, learning may also be performed within the bicycle parking lot management device 12.

[0017] The learning control model 129 uses bicycle / human contact images, such as front, rear, and oblique images of a person pushing a bicycle, and front, rear, and oblique images of a person riding on a bicycle, as input training data, and uses bicycle / human contact information as output training data to learn using deep learning (machine learning). This learning is also performed in advance outside the bicycle parking lot 1, and the operator uses the information terminal 3 to input the learning results to the learning control model 129. Model 129 to Store However, learning may also be performed within the bicycle parking lot management device 12.

[0018] Figure 3 is a flowchart for explaining the entry / exit processing routine executed by the CPU 121 of Figure 2. This flowchart is stored in the ROM 122 or flash memory 124. The routine of Figure 3 is an interrupt routine that starts when an entry or exiting person scans the two-dimensional code sign 15. In other words, when an entry or exiting person scans the two-dimensional code sign 15 shown in Figure 4(A) with an information terminal such as a smartphone, the CPU 121 of the bicycle parking lot management device 12 detects this via the Internet 2 and starts the routine shown in Figure 3.

[0019] In step 301, the CPU 121 of the bicycle parking lot management device 12 displays the image shown in Fig. 4(B) on the touch panel 16. As a result, if a person entering touches "enter," the process proceeds to entry processing step 302, and on the other hand, if a person leaving touches "exit," the process proceeds to exit processing step 303.

[0020] Although the entry / exit routine in FIG. 3 is started by scanning the two-dimensional code signboard 15, the entry processing step 302 and the exit processing step 303 may be independent entry processing programs and exit processing programs, and the entry processing program may be started by touching "entry" on the touch panel 16, and the exit processing program may be started by touching "exit" on the touch panel 16.

[0021] FIG. 5 is a detailed flowchart of the admission process step 302 in FIG.

[0022] First, in step 501, image data from the entrance / exit monitoring camera 13 shown in FIG. 6(A) is acquired.

[0023] Next, in step 502, a bicycle image and a human image are estimated and extracted from a plurality of image data, for example, nine image data, taken by the entrance / exit monitoring camera 13 using the learned control model 128.

[0024] Next, in step 503, nine bicycle images / person images are displayed on the touch panel 16 as shown in FIG. 6(B), and the visitor is asked to select one of them.

[0025] Next, in step 504, an entrance code as a bicycle management number is assigned to the bicycle and displayed on the touch panel 16, as shown in Figure 6(C). The entrance code may also be set by the visitor.

[0026] Next, in step 505, the selected bicycle / person image is taken as an entry image along with the date and time, e.g., ts, and written to a predetermined area in flash memory 124 specified by the entry code, as shown in Figure 11. This starts recording the bicycle's trajectory. Note that recording the entry image is useful for preventing problems at the time of entry.

[0027] Then, the routine of FIG. 5 ends at step 506.

[0028] FIG. 7 is a detailed flowchart of the exit processing step 303 in FIG.

[0029] First, in step 701, the person leaving the venue inputs the entrance code into the touch panel 16. As a result, the display shown in FIG.

[0030] Next, in step 702, as shown in FIG. 8B, a still image of the bicycle at the parking position corresponding to the entry code i is displayed on the touch panel 16, along with the usage fee.

[0031] Next, in step 703, it is determined whether or not the billing process has been completed. The billing process can be either cash payment or electronic payment using the payment server 4. When the billing process has been completed, the process proceeds to step 704.

[0032] Next, in step 704, the image from the general surveillance camera 14 is used to determine whether or not a bicycle image is currently present in the parking position corresponding to entry code i. If a bicycle image is present, it is determined that the charging process (toll settlement) will proceed first and the bicycle will move after, and in step 705 the charging process end flag F(i) for that entry code i is set. This charging process end flag F(i) is used in the main routine of Figure 9, which will be described later. In other words, at this point, the exit path of bicycle (i) is not recorded. Conversely, if a bicycle image is not present, it is determined that the bicycle will move first and the charging process (toll settlement) will move after, and the process proceeds to steps 706, 707, and 708.

[0033] Steps 706, 707, and 708 will now be described. That is, in step 706, image data from the entrance / exit monitoring camera 13 shown in FIG. 8C is acquired. Next, in step 707, the learned learning control model 128 is used to estimate and extract bicycles and people from the image data from the entrance / exit monitoring camera 13. Next, in step 708, the bicycle image / people image estimated and extracted in step 707, along with the date and time, for example, te, are written as an exit image to a predetermined area specified by the entry code i in the flash memory 124, as shown in FIG. 11. This completes the exit recording of the trajectory of bicycle (i) with entry code i. Note that recording the exit image is useful for preventing problems at the time of exit.

[0034] Then, the routine of FIG.

[0035] Fig. 9 is a flowchart for explaining a main routine executed by the CPU 121 of Fig. 2. This flowchart is stored in the ROM 122 or the flash memory 124.

[0036] First, in step 901, image data from the general surveillance camera 14 is acquired.

[0037] Next, in step 902, the learned learning control model 128 is used to estimate and extract bicycle images and human images from the image data of the overall monitoring camera 14.

[0038] Next, in step 903, it is determined whether bicycle (i) with entry code i is moving or stationary, that is, whether the position of bicycle (i) is different from or the same as the previous value. If the result shows that bicycle (i) is moving, that is, if the movement trajectory of bicycle (i) is extended, proceed to steps 904 to 910. In this case, a human image is always in contact with the bicycle image. On the other hand, if bicycle (i) is stationary, that is, if the movement trajectory of bicycle (i) is extended and stopped, proceed to steps 911, 912, and 913.

[0039] Steps 904 to 910 will now be described.

[0040] In step 904, the bicycle image / human image in contact state estimated and extracted in step 902, together with the time, for example, t1, t2: t1', t2', are converted into a moving image and written to a predetermined area specified by the entry code i in the flash memory 124, as shown in Figure 11. This records the trajectory of bicycle (i) with entry code i.

[0041] Next, in step 905, it is determined whether charging processing (payment of fees) for entry code i has already begun by checking whether the charging processing end flag F(i) is "1." If charging processing has already ended (F(i)="1"), then in step 906 it is determined whether bicycle (i) is outside the surveillance of the general surveillance camera 14. Normally, bicycle (i) is outside the surveillance of the general surveillance camera 14 but within the surveillance of the entrance / exit surveillance camera 13. In this case, the trajectory recording of bicycle (i) is completed in steps 907, 908, and 909, which are the same as steps 706, 707, and 708 in FIG. 7. Then, in step 910, the charging processing end flag F(i) is reset. On the other hand, if charging processing has not ended (F(i)="0"), or if charging processing has ended but the bicycle is within the surveillance of the general surveillance camera 14, the process proceeds directly to step 914.

[0042] Steps 911, 912, and 913 will now be described.

[0043] In step 911, the state of contact between the bicycle and the person is estimated and extracted using the learned control model 129 for the bicycle image and the person image estimated and extracted in step 902.

[0044] Next, in step 912, it is determined whether the bicycle and the person have changed from the contact state shown in Fig. 10(A) to the non-contact state shown in Fig. 10(B). As a result, the process proceeds to step 913 only if the bicycle and the person have changed from a contact state to a non-contact state.

[0045] In step 913, the bicycle image estimated in step 902 is taken as a still image along with the date and time, e.g., t3, and the bicycle parking location, and written to the flash memory 124 at a predetermined location specified by entry code i, as shown in Figure 11. However, if the bicycle comes to rest in an unintended location that is not a parking location, the parking location is not written. This stops the trajectory of bicycle (i) with entry code i.

[0046] In this way, for example, the entry trajectory of a bicycle / person image with entry code 123456 changes as a moving image at times t1, ..., t2, t3, and becomes a still image at time t3, with bicycle (i) parking at a certain bicycle parking location, for example, 11-A. Next, bicycle (i) leaves the parking location at time t1', changes again as a moving image at times t1' ..., t2', and at time te, bicycle (i) exits bicycle parking lot 1. In this case, the charged time T can be measured, for example, from time t3 to t1'. In other words, bicycle parking lot 1 can be managed using only the images from entrance / exit monitoring camera 13 and the images from overall monitoring camera 14.

[0047] Furthermore, in step 914, the trajectories of only people are managed. That is, for the security of the bicycle parking lot 1, the trajectories of all people entering the bicycle parking lot 1, regardless of whether they are bicycles, are managed. In this case, the trajectories of people, including moving and still images, are recorded in the image data of the general surveillance camera 14. That is, each person is given a person code H0001, H0002, ..., and the person image estimated in step 902, along with times t1, t2, ..., is written to a predetermined area defined by the person codes H0001, H0002, ... in the flash memory 124, as shown in FIG. 12. In this way, the person trajectories of the person codes H0001, H0002, ... are recorded, ensuring security. In this case, too, if the person trajectories are recorded only when people are moving, the storage capacity of the flash memory 124 can be reduced.

[0048] Furthermore, in step 915, it is determined whether the bicycle parking positions 11-A, 11-B, . . . , 11-I are fully occupied by bicycles or have at least one empty bicycle, and the full bicycle sign 17 is controlled accordingly.

[0049] Then, the process returns to step 901.

[0050] FIG. 11 is a diagram showing a bicycle trajectory image obtained by the flowcharts of FIGS. 3, 5, 7, and 9 and recorded in the flash memory 124 of FIG.

[0051] As shown in FIG. 11, for example, images of bicycle (123456) given entry code 123456 are intermittently recorded as video at dates and times ts, t1, ..., t2, and t3, and in this case, images of people are also recorded. At date and time t3 when bicycle (123456) is parked, the image of bicycle (123456) is a still image. In this case, images of people are usually not present, and the bicycle's parking location is recorded. However, if the bicycle is parked in an unintended location other than the parking location, the parking location is not recorded. In this way, images of bicycle (123456) from entry to parking are recorded. Next, bicycle (123456) is intermittently recorded as video at dates and times t1' ..., t2', and te, and images of people are also recorded in this case. In this way, images of bicycle (123456) from unparking to leaving are recorded. Note that the parking time (chargeable time) is from t3 to t1'. In this way, bicycles can be managed and billed. Also, since images of bicycles are recorded from entry to exit, bicycle security (anti-theft) can be improved without the need to lock and unlock the bicycle.

[0052] FIG. 12 is a diagram showing a human trajectory image obtained by step 914 of FIG. 9 and recorded in flash memory 124 of FIG.

[0053] As shown in FIG. 12, a person entering the bicycle parking lot 1 is recorded as a video. In this case, the person includes all users of the bicycle parking lot 1, non-users, people simply passing through the bicycle parking lot 1, and malicious intruders. Images of bicycles may also be mixed in with the person image. For example, an image of a person (H0001) assigned the person code H0001 is intermittently recorded as a video at times ts, t1, ..., t2, and t3, and in this case, the person image is also recorded. At time t3, when the person (H0001) pauses, the image of the person (H0001) becomes a still image. When the person (H0001) moves again, the person (H0001) is intermittently recorded as a video at times t1' ..., t2', and te. In this way, images of the person (H0001) from entry to exit are recorded. This improves the security of the bicycle parking lot 1 as a whole.

[0054] The bicycle parking lot management device of Fig. 2 can also be configured with hardware shown in Fig. 13. In Fig. 13, bicycles are treated as two-wheeled vehicles, including motorbikes.

[0055] Bicycle parking with multiple bicycle parking positions as shown in FIG. The place The bicycle parking lot management device for managing the bicycle parking lot comprises an overall image acquisition means 1301 that acquires overall image data of the bicycle parking lot, a motorcycle image / human image estimation and extraction means 1303 that estimates and extracts motorcycle images and human images using the overall image data as input data using a first learning control model 1302 that has been machine-learned and that uses motorcycle images and human images as input teacher data and motorcycle information and human information as output teacher data, and a first recording means 1304 that, when the motorcycle image is a moving image, records the motorcycle image and human images that come into contact with the motorcycle image together with the date and time, and, when the motorcycle image changes from a moving image to a still image, records the still image together with the date and time and the bicycle position where the motorcycle image is located.

[0056] Furthermore, the bicycle parking lot management device is equipped with a motorcycle / human contact image estimation and extraction means 1306 that uses a second learning control model 1305 that has been machine-learned to estimate and extract motorcycle / human contact images using overall image data as input data, with the motorcycle / human contact image as input teacher data and the motorcycle / human contact information as output teacher data, and a first recording means 1304 records a still image of the motorcycle image only when it is determined that the motorcycle image and the human image that was in contact with the motorcycle image are not in contact, based on the presence or absence of the motorcycle / human contact image.

[0057] Furthermore, the bicycle parking lot management device is equipped with an entrance / exit image acquisition means 1307 that acquires entrance / exit image data of the parking lot, and an entrance / exit motorcycle image / entrance / exit human image estimation extraction means 1308 that estimates and extracts entrance / exit motorcycle images and entrance / exit human images using a first learning control model 1302 with the entrance / exit image data as input data, and a first recording means 1304 records the entrance / exit motorcycle images and entrance / exit human images as entrance or exit images of the motorcycle images and human images.

[0058] Furthermore, the bicycle parking lot management device includes a second recording means 1309 that records the estimated extracted human image together with the date and time. When the human image is a still image, the second recording means 1309 records it only when the human image changes from a moving image to a still image.

[0059] In the above embodiment, the number of stationary bicycle image data recorded is small, but the number of moving bicycle image data recorded depends on the execution cycle time of the main routine. In order to extend the execution cycle time of the main routine, dummy steps or delay steps can be introduced to reduce the number of moving bicycle image data, thereby reducing the storage capacity of the flash memory 124.

[0060] Furthermore, although an entrance / exit monitoring camera 13 and an overall monitoring camera 14 are provided, if the overall monitoring camera 14 can also capture images of the vicinity of the entrance / exit G, the overall monitoring camera 14 will be unnecessary. In other words, at least one monitoring camera is required, and the number can be increased as appropriate.

[0061] Furthermore, although the bicycle parking lot 1 has an entrance / exit G, the entrance / exit G does not have to be provided.

[0062] Furthermore, the present invention can be applied to any modifications within the obvious scope of the above-described embodiments. [Industrial Applicability]

[0063] The bicycle parking lot management device according to the present invention can be used for two-wheeled vehicles such as motorcycles in addition to bicycles. [Explanation of symbols]

[0064] 1: Bicycle parking lot 11-A, 11-B, ..., 11-I: Bicycle parking locations 12: Bicycle parking lot management device 13: Entrance / exit surveillance camera 14: Overall surveillance camera 15: 2D code signboard 16: Touch panel 17: Full Sky Sign 2: Internet 3: Information terminal 4: Payment server 121:CPU 122:ROM 123:RAM 124: Flash memory 125: Input / output interface 126: Image Interface 127: Communication interface 128, 129: Learning control model F: Charging process end flag

Claims

1. A bicycle parking lot management program for managing a bicycle parking lot having a plurality of bicycle parking positions, an overall image acquisition step of acquiring overall image data of the bicycle parking lot; a motorcycle image / human image estimation and extraction procedure for estimating and extracting motorcycle images and human images using the entire image data as input data and a first learning control model that has undergone machine learning training using motorcycle images and human images as input teacher data and motorcycle information and human information as output teacher data; a first recording step of recording, when the motorcycle image is a moving motorcycle image including the motorcycle in a moving state, the motorcycle image and the image of the person contacting the motorcycle image together with date and time, and, when the motorcycle image changes from the moving motorcycle image to a still motorcycle image including the motorcycle in a stationary state, recording the still motorcycle image together with the date and time and the parking position where the motorcycle image is located; A bicycle parking lot management program for causing a computer to execute the above.

2. Furthermore, the computer is caused to execute a motorcycle / human contact image estimation and extraction procedure for estimating and extracting a motorcycle / human contact image using the entire image data as input data and a second learning control model that has been machine-learned using the motorcycle / human contact image as input teacher data and the motorcycle / human contact information as output teacher data, A bicycle parking lot management program to be executed by a computer as described in claim 1, wherein the first recording procedure records the still image of the motorcycle image only when it is determined that the motorcycle image and the person image in contact with the motorcycle image are not in contact based on the presence or absence of the motorcycle / person contact image.

3. moreover, an entrance / exit image acquisition step for acquiring image data of the entrance / exit of the parking lot; and a step of estimating and extracting an entrance / exit motorcycle image and an entrance / exit human image using the entrance / exit image data as input data and the first learning control model, 2. The bicycle parking lot management program for causing a computer to execute the bicycle parking lot management program according to claim 1, wherein the first recording step records the entrance / exit motorcycle image and the entrance / exit person image as an entrance image or an exit image of the motorcycle image and person image.

4. 2. The bicycle parking lot management program according to claim 1, further comprising a second recording step of recording the estimated extracted human image together with a date and time.

5. A bicycle parking lot management program to be executed by a computer as described in claim 4, wherein the second recording procedure records only when the human image is a human still image including a human in a stationary state, and the human image changes from a human moving image including the human in a moving state to the still image.

6. A bicycle parking lot management device for managing a bicycle parking lot having a plurality of bicycle parking positions, an overall image acquisition means for acquiring overall image data of the bicycle parking lot; a motorcycle image / human image estimation / extraction means for estimating and extracting motorcycle images and human images using the entire image data as input data, by using a first learning control model that has undergone machine learning training using motorcycle images and human images as input teacher data and motorcycle information and human information as output teacher data; a first recording means for recording the motorcycle image and the person image in contact with the motorcycle image together with date and time when the motorcycle image is a moving motorcycle image including the motorcycle in a moving state, and for recording the motorcycle image and the person image in contact with the motorcycle image together with date and time when the motorcycle image changes from the moving motorcycle image to a still motorcycle image including the motorcycle in a stationary state, together with the parking position where the motorcycle image is located and the still motorcycle image together with date and time; A bicycle parking lot management device comprising:

7. The present invention further includes a motorcycle / human contact image estimation / extraction means for estimating and extracting a motorcycle / human contact image using the entire image data as input data and a second learning control model that has been machine-learned using the motorcycle / human contact image as input teacher data and the motorcycle / human contact information as output teacher data, The bicycle parking lot management device described in claim 6, wherein the first recording means records the still image of the motorcycle image only when it is determined that the motorcycle image and the person image that was in contact with the motorcycle image are not in contact based on the presence or absence of the motorcycle / person contact image.

8. moreover, an entrance / exit image acquisition means for acquiring image data of the entrance / exit of the parking lot; an entrance / exit motorcycle image / entrance / exit human image estimating / extracting means for estimating / extracting an entrance / exit motorcycle image and an entrance / exit human image using the first learning control model with the entrance / exit image data as input data; Equipped with The bicycle parking lot management device according to claim 6, wherein the first recording means records the entrance / exit motorcycle image and the entrance / exit person image as an entrance image or an exit image of the motorcycle image and person image.

9. 7. The bicycle parking lot management device according to claim 6, further comprising a second recording means for recording the estimated extracted human image together with a date and time.

10. The bicycle parking lot management device described in claim 9, wherein the second recording means records the human image only when the human image changes from a human moving image including the human in a moving state to the still image when the human image is a human still image including the human in a stationary state.

Images