Elevator system

Abstract

Conventional elevator systems simply speed up the call process and do not perform optimal allocation adjustments for elevator cars when multiple floors are registered as landing locations. Furthermore, a special device called a passenger detection device is required. [Solution] The elevator system of this disclosure comprises surveillance cameras installed on multiple floors of a building, landing call registration devices installed on multiple floors, and a management device that assigns one of several elevator units to landing calls registered by the landing call registration devices. The management device has a feature extraction means that extracts feature quantities from images of the surveillance cameras that change according to the number of people in the shooting range of the surveillance cameras, and assigns an elevator unit based on the feature quantities corresponding to the landing calls.

Description

Technical Field

[0001] The present disclosure relates to an elevator system.

Background Art

[0002] In Patent Document 1, a passenger detection device installed in an elevator hall or a location away from the elevator hall detects passengers. When a passenger is detected, the elevator control device calculates the time when the passenger arrives at the elevator landing as an estimated passenger arrival time and makes a pre-call. Then, the historical data of the passenger arrival time is stored and learned to optimize the estimated passenger arrival time.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the case of Patent Document 1, it only makes the call earlier, and when landing registration is made on multiple floors, the optimal allocation adjustment of the car is not performed. Also, a special device called a passenger detection device is required.

[0005] The present disclosure has been made to solve the above problems. An object thereof is to provide an elevator system capable of optimally allocating cars when landing registration is made on multiple floors.

Means for Solving the Problems

[0006] The elevator system disclosed herein comprises surveillance cameras installed on multiple floors of a building, landing call registration devices installed on multiple floors, and a management device that assigns one of several elevator units to landing calls registered by the landing call registration devices, the management device comprising feature extraction means for extracting feature quantities that change according to the number of people in the field of view of the surveillance cameras from the images of the surveillance cameras, and The system includes a landing call registration time difference estimation means that responds to changes in feature quantities and estimates the time difference from the shooting range of the surveillance camera to the landing call registration device on the same floor as the surveillance camera, and extracts from images taken by a surveillance camera installed on the same floor as the landing call registration device before the time difference estimated by the landing call registration time difference estimation means. The allocation of units is based on their characteristic features. [Effects of the Invention]

[0007] This disclosure enables the optimal allocation of elevator cars when landings are registered on multiple floors. [Brief explanation of the drawing]

[0008] [Figure 1] This is a configuration diagram showing the configuration of the elevator system in Embodiment 1. [Figure 2] This is a graph showing the features extracted by the feature extraction method in a time series. [Figure 3] This is an example of learning the estimation of the time difference between boarding calls using a boarding call registration time difference estimation method. [Figure 4] This diagram illustrates the concept of the assignable floor range. [Figure 5] This is a conceptual diagram of the simulation using the assignable floor optimization method. [Figure 6] This figure shows an example of a simulation using the allocation-optimization method. [Figure 7] This diagram shows the features of the registered call level. [Figure 8] This is a graph showing the function in trials using the assignable-order optimization method. [Figure 9] This is a flowchart showing the process of assigning machine numbers in Embodiment 1. [Figure 10] This figure shows an example of hardware resources for a group management device. [Figure 11] This figure shows another example of hardware resources for a group management device. [Figure 12]It is a configuration diagram showing the configuration of the elevator system in Embodiment 2. [Figure 13] It is a diagram showing the relationship between the feature amount and the occupancy rate in the car. [Figure 14] It is a flowchart showing the process of car assignment in Embodiment 2. [Figure 15] It is a diagram showing a specific example of car assignment.

Mode for Carrying Out the Invention

[0009] The mode for carrying out the present disclosure will be described with reference to the accompanying drawings. In each figure, the same or corresponding parts are denoted by the same reference numerals, and duplicate descriptions will be appropriately simplified or omitted.

[0010] Embodiment 1. FIG. 1 is a configuration diagram showing the configuration of the elevator system in Embodiment 1. The elevator system includes a group management device 1, a landing call registration device 100, a surveillance camera 200, a car 300, and an elevator control device 400. The group management device 1 manages a plurality of elevators as a group.

[0011] The landing call registration device 100 is provided for each landing on each floor and is operated by a person. For example, the landing call registration device 100 has buttons indicating up, buttons indicating down, etc. A person who wants to go to an upper floor registers an upward landing call by pressing the button indicating up. The landing call registration device 100 may have floor buttons, and a person may directly press the button of the floor they want to go to.

[0012] The surveillance camera 200 is installed mainly in the common parts on each floor for the purpose of monitoring for suspicious persons. For example, the surveillance camera 200 is installed on a passage leading to the landing, on the wall surface near the landing, or on the ceiling surface.

[0013] The car 300 is for people to ride in and is provided for each elevator. In group management, the car of an elevator is referred to as a unit. The elevator control device 400 is provided for each elevator and controls the movement and stop of the car 300, the devices in the car 300, etc.

[0014] The group management device 1 includes a landing call assignment means 2, a floor limit means 3 assignable for each unit, an optimizable floor range means 4, a feature amount extraction means 5, a landing call registration time difference estimation means 6, and an operation history acquisition means 7.

[0015] When a landing call is registered by the landing call registration device 100, the landing call assignment means 2 assigns one of the managed units to the landing call.

[0016] The floor limit means 3 assignable for each unit then sets a floor limit on which landing calls the assigned unit can be assigned to. The optimizable floor range means 4 generates a function of an optimizable floor range that minimizes the average waiting time by machine learning using past operation performance data. As a learning algorithm, supervised learning or reinforcement learning is used to learn the relationship between feature amounts and waiting times to search for an optimal solution. Note that the learning is not only executed once, but can also be continuously executed to update the function according to changes in the operation environment.

[0017] The feature amount extraction means 5 extracts feature amounts from the images captured by the surveillance camera 200. The "feature amount" in the present disclosure refers to numerical information indicating the presence or quantity of people within the imaging range obtained by analyzing the images captured by the surveillance camera 200. Specifically, it is the number of people detected by image analysis, the total area of the person areas, the amount of movement of people, or the number of pixels of a specific color (such as skin color) in the image.

[0018] As an implementation method for image analysis using feature extraction means 5, for example, a deep learning model can be used. When using a deep learning model, a feature extraction algorithm based on a convolutional neural network (CNN) can be applied. These methods are pre-trained on a large dataset of human images and extract feature information such as the presence, position, size, and movement vector of people from images, and have robust feature extraction performance against various lighting conditions, people's posture, clothing, and overlap (occlusion) of multiple people. The extracted feature information is quantified as the number of people, dwell time, degree of overlap, movement pattern, etc.

[0019] The landing call registration time difference estimation means 6 uses machine learning to estimate the time difference between the change in feature quantities at the surveillance camera 200 and the registration at the landing call registration device 100. As learning methods, a hidden Markov model (HMM) that estimates boarding intention from the time series pattern of features, or regression analysis, can be applied. Learning is performed individually according to the installation location of the surveillance camera 200 on each floor and the walking path, and time difference predictions are made considering differences in walking speed and path. Detailed operation will be described later.

[0020] The operation history acquisition means 7 acquires and stores operation history, such as elevator operation and human operation at the landing call registration device 100, as time-series data. This time-series data of operation history includes the control details of the elevator control device 400 and the weight inside the car 300 measured by the weighing device 301.

[0021] Next, the estimation operation of the landing call registration time difference estimation means 6 will be explained in detail. First, I will explain the concept of the relationship between feature variables and boarding area registration. Figure 2 is a graph showing the features extracted by the feature extraction means 5 in time series. The feature extraction means 5 acquires images from the surveillance camera 200. If no people are in the image, no features are observed. However, if people are within the shooting range of the surveillance camera 200, features corresponding to the number of people within that shooting range can be extracted. In Figure 2, changes in features appear as people move within the shooting range of the surveillance camera 200. For example, consider a passenger flow where everyone leaves the office at the same time during lunch break to eat in the cafeteria, uses the elevator, and heads to the cafeteria. In this case, at time T1, the number of people within the shooting range of the surveillance camera 200 increases sharply. After that, they pass through the shooting range and reach the landing, and at time T2, a landing call registration is made for the direction of the floor where the cafeteria is located, i.e., a specific direction. Then, at time T3, the elevator car 300 arrives at the landing and people board. In this case, it is considered that the number of people boarding the elevator car 300 corresponds to the number of features at time T1. Therefore, the time from when the characteristic appears until the boarding call registration is made is estimated to be the time from when the person is photographed by the surveillance camera 200 until they arrive at the boarding area, i.e., the boarding call registration time difference.

[0022] The landing call registration time difference estimation means 6 learns by acquiring time-series data of image features from the surveillance cameras 200 on each floor and time-series data of landing call registration occurrences on each floor, and calculates the landing call registration time difference. Also, since the position of the surveillance cameras 200 differs on each floor, it is necessary to estimate the landing call registration time difference for each floor.

[0023] Furthermore, in situations where people frequently pass by, it is difficult to establish a correspondence between changes in feature quantities and boarding call registrations. Therefore, the time-series data will be based on off-peak hours. That is, when one person enters the field of view of the surveillance camera 200, the feature quantity changes. Subsequently, that person goes to the boarding call registration device 100 and registers their boarding call. In other words, the time difference between the person corresponding to this change in feature quantity and their arrival at the boarding call registration device is estimated.

[0024] Figure 3 shows an example of the estimation of the boarding call registration time difference by the boarding call registration time difference estimation means 6. Multiple boarding call registrations are made within the time range of the time-series data of the operation history and the time-series data of the feature quantities. Therefore, the boarding call registration time difference estimation means 6 accumulates the time difference of events where the feature quantity changes during a time period when the change in the feature quantity is almost zero, and then a boarding call registration is made afterward. The horizontal axis represents this time difference, and the vertical axis represents the frequency of occurrence of this time difference.

[0025] In this case, the walking speed, that is, the travel time from the shooting range of the surveillance camera 200 to the landing call registration device 100, is considered to be approximately constant and peaks at the median, as shown in Figure 3. The landing call registration time difference estimation means 6 outputs the time difference that is the median as the landing call registration time difference.

[0026] Next, the operation of the assignable floor optimization means 4 will be explained in detail. The assignable floor optimization means 4 receives time-series data of feature quantities and time-series data of operation history as input, and outputs a function that indicates the assignable floor range corresponding to the feature quantities.

[0027] Figure 4 illustrates the concept of the assignable floor range. The assignable floor range is the range of other floors to which a locomotive can be further assigned after a locomotive has been assigned to a locomotive based on the locomotive call registration.

[0028] Figure 4 shows the case where a unit is assigned on the 7th floor based on the downward landing call registration. In this case, if the assignable floor range is 1, the assignable floor range is one floor above and below, i.e., floors 6 to 8. If the assignable floor range is 2, the assignable floor range is two floors above and below, i.e., floors 5 to 9. If the assignable floor range is 3, the assignable floor range is three floors above and below, i.e., floors 4 to 10.

[0029] The purpose of setting the range of floors that can be assigned is to shorten the time it takes to complete a circuit. For example, if the destination floor is the 1st floor, and the stopping range of a certain train is restricted to lower floors, that train will not travel to higher floors and can return to the 1st floor immediately. Also, if a train that normally stops on higher floors does not have to stop on lower floors, it can travel a longer distance at its rated speed, thus shortening the time it takes to arrive at the 1st floor.

[0030] The assignable-floor optimization means 4 determines the optimal assignable-floor range function based on its relationship with the features. Here, as shown in the conceptual diagram in Figure 5, several trials 42 are simulated in the simulator 41 based on the time-series data 71 of the operation history and the time-series data 51 of the features. The function of the trial that minimizes the average waiting time obtained as the trial result 43 is then output. As an optimization method, for example, a genetic algorithm, particle swarm optimization, or Bayesian optimization can be applied.

[0031] Figure 6 shows an example of a simulation using the assignable floor optimization means 4. The operation history used is from an actual lunchtime period.

[0032] In Figure 6, the boarding call registration time 71a and the downward boarding call registration floor 71b are based on the actual operation history time series data 71 obtained by the operation history acquisition means 7. The call registration floor feature quantity 51a is, as shown in Figure 7, a feature quantity in the feature quantity time series data 51 obtained by the feature quantity extraction means 5, at a time before the boarding call registration time difference from the boarding call registration time 71a.

[0033] As shown in Figure 6, the actual situation was as follows: A landing call occurred on the 7th floor at 12:00:11. At that time, the feature quantity before the time difference between 12:00:11 and the registration of the landing call on the 7th floor was 50. A landing call occurred on the 9th floor at 12:00:12. At that time, the feature quantity before the time difference between 12:00:12 and the registration of the landing call on the 9th floor was 70. A landing call occurred on the 10th floor at 12:00:15. At that time, the feature quantity before the time difference between 12:00:15 and the registration of the landing call on the 10th floor was 150. A landing call occurred on the 11th floor at 12:00:17. At that time, the feature quantity before the time difference between 12:00:17 and the registration of the landing call on the 11th floor was 250.

[0034] Based on this actual data, multiple functions are set for the feature vectors and assignable floors, and trials are conducted with each setting. Note that the values ​​for the difference in landing call registration times for the 7th, 9th, 10th, and 11th floors will differ individually if the installation location of the surveillance camera 200 differs on each floor.

[0035] Figure 8 is a graph showing the function in the trial. Figure 8(a) shows the basic idea for considering the trial pattern. First, the number of features changes depending on the number of people. If there are many people, the number of features will be large. Therefore, as the number of features increases, we assume that the number of people boarding from that floor will increase, and we gradually narrow the range of floors to which we can assign.

[0036] Figure 8(b) is a graph showing the function for Trial 1. In Trial 1, the assignable rank range f(x) is 3 when the feature quantity (x) is between 0 and 100, and 1 when it exceeds 100.

[0037] Figure 8(c) is a graph showing the function for Trial 2. In Trial 2, the assignable rank range f(x) is 5 when the feature quantity (x) is between 0 and 200, and 1 when it exceeds 200.

[0038] In Figure 6, the simulation based on Trial 1 resulted in machine number 43a assigned to Trial 1.

[0039] First, for the elevator call at the 7th floor at 12:00:11, elevator A is assigned. Since the feature value is 50, the assignable floor range is from the 5th to the 9th floor.

[0040] Next, we assign a bus to the landing call on the 9th floor at 12:00:12. First, the assignable floor range for bus A is floors 5-9. Therefore, bus A is assigned. Also, since the feature value is 70, the assignable floor range is floors 7-11, but the assignable floor range for a landing call on the 7th floor has already been set to floors 5-9. Therefore, the assignable floor range is floors 7-9, where both overlap. Note that floors 9 and 7 have already been assigned. Therefore, the only floor that is actually assignable is floor 8.

[0041] Next, we assign a elevator to the 10th floor at 12:00:15. Since the 10th floor is not an assignable floor for elevator A, elevator B is assigned to it. Also, because the feature count is 150, the assignable floor range is from the 9th to the 11th floor. However, since the 9th floor is already assigned to elevator A, the only actually assignable floor is the 11th floor.

[0042] Next, an assignment is made for the call from the 11th floor at 12:00:17. The 11th floor is the available floor for assignment of bus B. Therefore, bus B is assigned. Note that there are no available floors for assignment of bus B.

[0043] Under these conditions, we calculate the average waiting time at the landings on the 7th, 9th, 10th, and 11th floors. Let's assume, for example, that it was 22.5 seconds.

[0044] Furthermore, in Figure 6, the simulation based on Trial 2 resulted in machine number 43b assigned to Trial 2.

[0045] For the elevator call at the 7th floor at 12:00:11, elevator A will be assigned. Since the feature value is 50, the assignable floor range is from the 2nd to the 12th floor.

[0046] Next, we assign a elevator to the landing call on the 9th floor at 12:00:12. First, the assignable floor range for elevator A is from the 2nd to the 12th floor. Therefore, elevator A is assigned. Also, since the feature value is 70, the assignable floor range is from the 4th to the 14th floor, but the assignable floor range for the landing call on the 7th floor has already been set to the 2nd to the 12th floor. Therefore, the assignable floor range is from the 4th to the 12th floor, where both overlap. Of these, the 9th and 7th floors have already been assigned. Therefore, the actual assignable floors are from the 4th to the 6th, 8th, and 10th to the 12th floor.

[0047] Next, we assign a elevator to the landing call at 12:00:15 on the 10th floor. Since the 10th floor is an assignable floor for elevator A, elevator A is assigned. Also, since the feature quantity is 150, the assignable floor range is 5 to 15, but the assignable floor range already overlaps with the landing calls on the 7th and 9th floors, which is 4 to 12. Therefore, the assignable floor range is 5 to 12, which overlaps with both. Of these, the 7th, 9th, and 10th floors have already been assigned. Therefore, the actual assignable floors are 5-6, 8, and 11-12.

[0048] Next, we assign a lift to the 11th floor for a call at 12:00:17. The 11th floor is among the floors to which lift A can be assigned. However, due to the feature value of 250, the range of assignable floors is 10 to 12. Lift A has already been assigned to the 7th and 9th floors, which are outside this assignable range. Therefore, lift A is excluded, and lift B is assigned.

[0049] Under these conditions, we calculate the average waiting time at each boarding area. Let's assume, for example, that it was 37.1 seconds.

[0050] This process is repeated with multiple trials, such as Trial 3, Trial 4, and so on. Finally, the function of the assignable rank range for a feature that minimizes the average waiting time is found, and that function is output.

[0051] The unit-specific floor allocation limiting means 3 performs processing based on the function output from the allocation floor optimization means 4.

[0052] Figure 9 is a flowchart showing the process of assigning elevator units in response to landing calls, as performed by the group control device 1. Note that this flowchart operates only during specific time periods when many people move towards a particular floor, such as during lunchtime or when people are leaving work. For example, during lunchtime, the specific floor would be the floor with the cafeteria. Alternatively, if there is no cafeteria and people eat outside, it would be the first floor, which is connected to the building's exit.

[0053] First, the landing call registration device 100 receives a button press from a person indicating the direction of travel and registers a landing call for a specific floor (step S001).

[0054] Based on this landing call, the landing call assignment means 2 determines the assigned landing car (step S002).

[0055] The unit-specific floor allocation restriction means 3 obtains feature quantities from the feature extraction means 5 from the time of the landing call registration to the time difference before the landing call registration. Then, based on these feature quantities and the function output by the allocation floor optimization means 4, it determines the range of floors that can be allocated for this unit (step S003).

[0056] Based on the information regarding the assignable floor range for this unit, the landing call assignment means 2 restricts assignments to landing calls from landing call registration devices 100 outside the assignable floor range that are directed towards a specific floor, and does not assign this unit (step S004).

[0057] Once this assigned elevator arrives at the specified floor, the landing call assignment restriction imposed in step S004 is lifted (step S005).

[0058] In this way, by assigning a bus unit to a registered bus stop call and then restricting the floors to which that unit can be assigned, passenger waiting times at the bus stop can be made more appropriate.

[0059] Furthermore, by using the feature data from the surveillance camera 200 images, there is no need to add any new devices, such as sensors to detect people.

[0060] Furthermore, the route to the landing may differ depending on the floor. For example, the surveillance camera 200 that captures footage when a person in one area of ​​the 7th floor goes to the landing may be different from the surveillance camera 200 that captures footage when a person in a different area of ​​the 7th floor goes to the landing. In such cases, the difference in landing call registration times at each surveillance camera 200 is estimated. The sum of the features before the difference in landing call registration times at each surveillance camera 200 can be used as the feature quantity for the floor where the call was registered.

[0061] The images captured by the surveillance camera 200 are either moving images or still images taken continuously at intervals of several seconds. Furthermore, the landing call registration time difference estimation means 6, which is a learning device, may be provided for each floor (surveillance camera).

[0062] Figure 10 shows an example of the hardware resources of the group management device 1. The group management device 1 is a computer having a processor 1a, memory 1b, and a transmit / receive circuit 1c as hardware resources. Note that there may be multiple processors 1a, memory 1b, and transmit / receive circuits 1c.

[0063] The processor 1a is also called a CPU (Central Processing Unit), central processing unit, arithmetic unit, microprocessor, or DSP. For memory 1b, semiconductor memory, magnetic disks, flexible disks, optical disks, compact disks, minidiscs, or DVDs may be used. Possible semiconductor memories include RAM, ROM, flash memory, EPROM, and EEPROM.

[0064] Figure 11 shows another example of the hardware resources of the group management device 1. In the example in Figure 11, the group management device 1 has a processing circuit that includes a processor 1a, memory 1b, a transmit / receive circuit 1c, and dedicated hardware 1d. Some of the functions of the group management device 1 are realized by the dedicated hardware 1d. Alternatively, all of the functions of the group management device 1 may be realized by the dedicated hardware 1d. The dedicated hardware 1d can be a single circuit, a composite circuit, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.

[0065] Embodiment 2. In Embodiment 2, the car occupancy rate is estimated from feature quantities, and based on this, the floors to which an elevator can be assigned are restricted. Figure 12 is a configuration diagram showing the configuration of the elevator system in Embodiment 2. In Figure 12, the same reference numerals are used for components that are the same as or equivalent to those in Figure 1, and their descriptions are omitted or simplified.

[0066] In Figure 12, the cage occupancy estimation means 8 sets a function between the feature quantity and the cage occupancy rate. Note that the cage occupancy estimation means 8 is a learning device.

[0067] Figure 13 shows the relationship between features and in-car occupancy rates. Figure 13(a) is a graph showing the relationship between time and features, and Figure 13(b) is a graph showing the relationship between time and occupancy rates.

[0068] In Figure 13(a), the person indicated by the feature quantity before the time difference between the time of boarding call registration and the time of boarding call registration boards the elevator car that arrives at that boarding station. Then, in Figure 13(b), as a result of this boarding, the occupancy rate of the elevator car, in this case the ratio of the weight measured by the weighing device 301 of the elevator car 300 to the passenger capacity weight, increases.

[0069] The elevator car occupancy rate estimation means 8 receives time-series data of feature quantities and time-series data of operation history as input, and outputs a function that shows the occupancy rate corresponding to the feature quantities.

[0070] For example, a single boarding call registration is extracted from the time-series data of the operation history, and a corresponding feature is obtained from the time-series data. Furthermore, the increase in occupancy rate corresponding to that boarding call registration is calculated from the time-series data of the operation history. This is plotted multiple times on a graph of the feature and occupancy rate, and an approximation curve is found to establish a function between the feature and the occupancy rate inside the train car. Figure 13(c) is a graph illustrating the function between the feature and the occupancy rate inside the train car.

[0071] The floor allocation restriction means 3 for each elevator unit sets the floors to which each elevator unit can be allocated based on the occupancy rate estimated by the elevator occupancy rate estimation means 9.

[0072] Furthermore, the floor allocation restriction means 3 for each elevator unit pre-determines the range of floors that can be allocated to a landing call registration, according to the occupancy rate.

[0073] For example, in the case of car 300, a full capacity value is defined as 80% of the car's capacity being occupied. If the occupancy rate is 40-59%, the assignable floor range is the 2nd floor. If the occupancy rate is 60-79%, the assignable floor range is the 1st floor. If the occupancy rate is 80% or higher, there is no assignable floor range.

[0074] Figure 14 is a flowchart showing the process of assigning elevator units in response to landing calls, as performed by the group control device 1 in Embodiment 2. In Figure 14, steps similar to those in the flowchart Figure 9 showing the elevator unit assignment process in Embodiment 1 are denoted by the same reference numerals, and their explanations are omitted or simplified. Note that this flowchart operates only during specific time periods when many people are moving in the direction of a specific floor, such as during quitting time or lunchtime.

[0075] First, in steps S001 and S002, the assigned machine number is determined.

[0076] The elevator car occupancy rate estimation means 9 obtains feature quantities from the feature quantity extraction means 5 from the time of elevator car call registration to the time difference between the time of elevator car call registration and the time of elevator car call registration. Then, the elevator car occupancy rate estimation means 9 estimates the occupancy rate based on the output function from these feature quantities (step S006).

[0077] The unit-specific floor allocation restriction means 3 sets the range of floors that can be allocated based on the occupancy rate. The landing call allocation means 2 restricts allocation to landing calls from landing call registration devices 100 that are outside the range of floors that can be allocated and does not allocate this unit (step S007).

[0078] Once this assigned elevator arrives at the specified floor, the landing call assignment restriction imposed in step S007 is lifted (step S005).

[0079] Figure 15 shows a specific example of the unit allocation process carried out using the flowchart in Figure 14. Assume there are four elevators, A, B, C, and D, and that elevators B and C are already occupied by passengers on floors above the 13th floor, with occupancy rates of 10% and 80%, respectively.

[0080] In this case, for example, a landing call registration for a downward direction is made on the 11th floor. In this case, elevator A is assigned first. Then, the feature quantity before the time difference of the landing call registration is calculated, and the occupancy rate is calculated as a function of the feature quantity and the occupancy rate. Let's say that it is 40%. In that case, the floor range to which elevator A can be assigned is the 9th to 13th floors, which is two floors above and below.

[0081] Next, let's assume a landing call registration is made on the 8th floor in the downward direction. Since the 8th floor is not within the range of floors to which machine A can be assigned, machine B will be assigned. Then, we find the feature quantity before the time difference of the landing call registration and calculate the occupancy rate as a function of the feature quantity. Let's assume that it is 20%. In that case, the range of floors to which machine B can be assigned is the 6th to 10th floors, which is two floors above and below. In reality, however, the assignable floors are the 6th, 7th, 9th, and 10th floors, excluding the 8th floor.

[0082] Next, let's assume a landing call registration is made on the 7th floor. Since the 7th floor is an available floor for elevator B, elevator B is assigned. Then, we find the feature vector before the time difference of the landing call registration and calculate the occupancy rate using a function of the occupancy rate on the feature vector. Let's assume that this rate is 30%. In that case, since the available floor range for elevator B is already floors 6-10, the available floor range becomes floors 6-9. In reality, however, the available floors are floors 6 and 9.

[0083] Now, let's assume that the occupancy rate on the 8th floor is 75%. In this case, the current occupancy rate of Unit B is 10%, and the estimated occupancy rate on the 8th floor is 75%, resulting in an occupancy rate of 85%. In this case, since it exceeds the 80% capacity value, the allocation will be reviewed, and Unit D will be allocated.

[0084] Thus, even if the occupancy rate is estimated from the feature quantities and the range of floors that can be allocated is set based on the occupancy rate, the waiting time for passengers at the boarding area can be made appropriate, just as in Embodiment 1.

[0085] Although preferred embodiments have been described in detail above, the invention is not limited to these embodiments, and various modifications and substitutions can be made to the embodiments described above without departing from the scope of disclosure.

[0086] Furthermore, when referring to the number, quantity, amount, range, etc., of each element in the embodiments, the apparatus of this disclosure is not limited to the referred number unless specifically stated or clearly defined in principle. Also, the structures, etc., described in these embodiments are not necessarily essential unless specifically stated or clearly defined in principle.

[0087] The various aspects of this disclosure are summarized below as an appendix. (Note 1) In an elevator system comprising surveillance cameras installed on multiple floors of a building, landing call registration devices installed on the multiple floors, and a management device that assigns one of several elevator units to landing calls registered by the landing call registration device, The management device includes a feature extraction means for extracting feature quantities from the images of the surveillance camera that change according to the number of people in the shooting range of the surveillance camera. An elevator system characterized by assigning the elevator unit number based on the characteristic quantity corresponding to the landing call. (Note 2) The elevator system according to Appendix 1, characterized in that the feature quantity corresponding to the landing call is the feature quantity extracted from the image taken by the surveillance camera installed on the same floor as the landing call registration device where the landing call registration was made, and the image was taken before an estimated time difference from the landing call registration. (Note 3) The elevator system according to Appendix 2, characterized in that the management device has a landing call registration time difference estimation means for estimating the time difference from the shooting range to the landing call registration device in response to the change in the feature quantity. (Note 4) The elevator system according to any one of the appendices 1 to 3, characterized in that the management device has a means for limiting the number of floors to which an elevator can be assigned can be assigned, which restricts the assignment of the elevator assigned to the landing call (Note 5) The elevator system according to Appendix 4, wherein the means for limiting the number of floors that can be allocated to each elevator unit changes the allocation limit based on the characteristic quantity prior to the time difference between the registration of the landing call and the elevator system. (Note 6) The elevator system according to Appendix 5, wherein the management device includes an operation history acquisition means for acquiring and storing the operation histories of a plurality of the aforementioned elevators, and an assignable floor optimization means for determining a function between the feature quantity that minimizes the average waiting time at each floor landing for the landing call in the operation history and the assignable floor range, and the assignable floor restriction means for each elevator changes the assignment restriction based on the function. (Note 7) The elevator system according to Appendix 5, wherein the management device comprises an operation history acquisition means for acquiring and storing operation history including the car weight of a plurality of the aforementioned elevator units, and an occupancy rate estimation means for estimating the car occupancy rate of the aforementioned elevator unit from the feature quantities based on the operation history and the feature quantities, and the allocation floor restriction means for each elevator unit changes the allocation floor range for each elevator unit to the floor where the landing call registration was made, based on the car occupancy rate estimated at the floor where the landing call registration was made. (Note 8) The elevator system according to Appendix 7, wherein the means for limiting the number of floors that can be allocated to each elevator unit is characterized in that the allocation is reviewed when the sum of the current car occupancy rate and the estimated car occupancy rate at the floor where the landing call registration was made exceeds the full occupancy value. (Note 9) The elevator system according to any one of the appendices 1 to 8, characterized in that the aforementioned feature quantity is the number of skin-colored pixels in the image captured by the surveillance camera. (Note 10) The elevator system according to any one of the appendices 1 to 8, characterized in that the surveillance camera is an AI camera, and the feature quantity is the number of people detected from the image taken by the AI ​​camera. [Explanation of symbols]

[0088] 1 Group management device, 1a Processor, 1b Memory, 1c Transceiver circuit, 1D dedicated hardware, 2. Landing call assignment means, 3. Assignable floor limiting means for each elevator, 4 Assignable floor optimization means, 41 Simulator, 42 Trial, 43 Trial results, 5 Feature extraction methods, 51 Feature time series data 6. Means for estimating the time difference between boarding location call registrations, 7. Means for acquiring operation history, 71 Operation history, 8 Car occupancy rate estimation means, 100 Platform call registration device, 200 surveillance cameras, 300 elevator cars, 400 elevator control devices

Claims

1. In an elevator system comprising surveillance cameras installed on multiple floors of a building, landing call registration devices installed on the multiple floors, and a management device that assigns one of several elevator units to landing calls registered by the landing call registration device, The management device includes a feature extraction means for extracting feature quantities from the images of the surveillance camera that change according to the number of people in the shooting range of the surveillance camera, The system includes a landing call registration time difference estimation means that estimates the time difference from the shooting range of the surveillance camera to the landing call registration device on the same floor as the surveillance camera, in response to the change in the aforementioned feature quantity. An elevator system characterized in that, from the landing call registration, the number of the elevator is assigned based on the feature quantities extracted from the image captured by the surveillance camera installed on the same floor as the landing call registration device, before the time difference estimated by the landing call registration time difference estimation means.

2. The elevator system according to Claim 1, characterized in that the landing call registration time difference estimation means acquires and learns time series data of the feature quantities of the image of the surveillance camera and time series data of landing call registration occurrences at the landing call registration device on the same floor as the surveillance camera, and calculates the time difference.

3. The elevator system according to claim 1, characterized in that the management device has a means for limiting the number of floors to which an elevator can be assigned can be assigned, which restricts the assignment of the elevator assigned to the landing call

4. The elevator system according to claim 3, characterized in that the means for limiting the number of floors that can be allocated to each elevator unit changes the allocation limit based on the characteristic quantity prior to the time difference between the registration of the landing call and the elevator system according to claim 3.

5. The elevator system according to claim 4, wherein the management device includes an operation history acquisition means for acquiring and storing the operation history of a plurality of the aforementioned elevators, and an assignable floor optimization means for determining a function between the feature quantity that minimizes the average waiting time at each floor landing for the landing call in the operation history and the assignable floor range, and the assignable floor restriction means for each elevator changes the assignment restriction based on the function.

6. The elevator system according to claim 4, wherein the management device comprises an operation history acquisition means for acquiring and storing operation history including the car weight of a plurality of the aforementioned elevator units, and an occupancy rate estimation means for estimating the car occupancy rate of the aforementioned elevator unit from the feature quantities based on the operation history and the feature quantities, and the allocation floor restriction means for each elevator unit changes the allocation floor range for each elevator unit to the floor where the landing call registration is made, based on the car occupancy rate estimated at the floor where the landing call registration is made.

7. The elevator system according to claim 6, wherein the means for limiting the number of floors that can be allocated to each elevator unit revises the allocation when the sum of the current car occupancy rate and the estimated car occupancy rate at the floor where the landing call registration was made exceeds the full occupancy value.

8. The elevator system according to any one of claims 1 to 7, characterized in that the aforementioned feature quantity is the number of skin-colored pixels in the image captured by the surveillance camera.

Images