An elevator car passenger distribution monitoring method for elevator energy feedback optimization

Abstract

The present application relates to the technical field of image data processing, and more particularly to an elevator car passenger distribution monitoring method for elevator energy feedback optimization, comprising: acquiring an overhead image of an elevator car, identifying passenger shape centers and quantity, constructing a deformation action matrix, generating a deformed two-dimensional Gaussian function, superimposing to generate a quality density field, calculating the coordinates of the actual centroid of the elevator car, based on the coordinate deviation of the actual centroid and the geometric center of the elevator car, calculating the additional friction power loss, dynamically compensating the energy feedback control parameter, and completing the elevator car passenger distribution monitoring for elevator energy feedback optimization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing. More particularly, the present application relates to an elevator car passenger distribution monitoring method for elevator energy feedback optimization. BACKGROUND

[0002] In modern building transportation systems, elevators as the core vertical transportation tool, its energy consumption problem is concerned, for this, energy feedback technology emerged as the times require, the core of energy feedback technology lies in, when the operating state of the elevator meets certain conditions, such as the weight relationship between the car and the counterweight makes the overall potential energy of the system surplus, the control system will drive the frequency converter of the elevator to make the traction machine work in generator mode, and the surplus mechanical energy, including potential energy and kinetic energy, in the car operation is converted into electric energy and sent back to the power grid, so as to realize energy saving benefit.

[0003] The Chinese patent document with publication number CN116823673B discloses a high-speed elevator car passenger state visual perception method based on image processing. The present application obtains a video enhancement image for each video gray scale image through guided filtering, obtains the noise evaluation index of each video enhancement image according to the isolation and abnormality analysis of noise points; obtains the fitting edge according to the pixel point offset condition, analyzes the similar situation between the fitting edge and the edge, and obtains the edge preservation evaluation index of each video enhancement image; obtains the enhancement effect index through the noise evaluation index and the edge preservation evaluation index, determines the optimal regularization parameter according to the enhancement effect index of all video enhancement images, and obtains the optimal video enhancement image to judge the passenger state according to the guided filtering corresponding to the optimal regularization parameter.

[0004] The above patent document only judges the state of the passengers, and does not obtain the distribution of the passengers in the elevator car. At the same time, the energy feedback control strategy commonly used in the industry at present mainly derives its decision from the weighing device installed at the bottom of the car. The weighing device can measure the total load in the car in real time and output a single, scalar form of total mass data to the control system. The control system compares the total mass data with the preset counterweight mass, and when the difference between the two meets the triggering condition of energy feedback, the energy feedback function is started or adjusted.

[0005] However, this control model relying only on total mass information will equate the complex spatial distribution of all passengers in the car to a concentrated force acting on the geometric center of the elevator car floor, which is a highly idealized treatment that ignores the actual spatial distribution characteristics of the actual load in the car. In actual operation, the position of the passengers in the car is random and uneven, which will inevitably cause the actual center of mass of the elevator car to deviate from the geometric center of the elevator car. According to the principle of basic mechanics, the product of the deviation of the center of mass and the gravity will produce a tilting moment acting on the car, which will cause the car to tilt towards one side of the guide rail, thereby increasing the normal pressure between the guide shoe and the guide rail on that side and increasing the friction, while the other side decreases accordingly. This asymmetric additional friction caused by uneven load distribution constitutes a resistive load, thereby consuming a portion of the energy that could have been recovered. SUMMARY

[0006] To solve the problem that the prior art only relies on total mass information and cannot perceive and evaluate the tilting moment caused by uneven passenger distribution and the additional friction loss it produces, resulting in reduced efficiency of elevator energy feedback, the present application proposes a method for monitoring passenger distribution in an elevator car for optimizing elevator energy feedback, which comprises the following steps:

[0007] Obtain an overhead image of the elevator car; use semantic segmentation to output mask values for each pixel point in the overhead image to identify the shape center of each passenger in the elevator car and determine the number of passengers; for any passenger in the elevator car, construct a deformation action matrix representing the squeezing effect of other passengers on the passenger; based on the deformation action matrix, generate a deformation-corrected two-dimensional Gaussian function for each passenger, wherein the covariance matrix of the two-dimensional Gaussian function reflects the change in passenger occupancy shape caused by the squeezing effect; weight and superimpose the deformation-corrected two-dimensional Gaussian function of each passenger to generate a mass density field representing the mass distribution in the elevator car. Based on the mass density field, use the center of mass formula to obtain the coordinates of the actual center of mass of the elevator car; calculate the additional friction power loss caused by uneven mass distribution according to the coordinate deviation of the actual center of mass from the geometric center of the elevator car; dynamically compensate the energy feedback control parameters of the elevator according to the additional friction power loss to complete the monitoring of passenger distribution in the elevator car for optimizing elevator energy feedback.

[0008] This invention acquires an elevator car top-view image, classifies each pixel in the image using a semantic segmentation model, and outputs mask values ​​that distinguish between occupied and vacant areas, accurately identifying the morphological center and number of each passenger, thus achieving precise perception of the spatial distribution of passengers within the car. By constructing a deformation action matrix to quantify the mutual squeezing effect among passengers, based on the pixel coordinates of the passenger morphological center, it reflects the influence of the relative positions of different passenger pixel regions on the degree of squeezing, and generates a deformation-corrected two-dimensional Gaussian function, allowing the function's covariance matrix to dynamically match the morphological changes of passenger pixel regions in the image caused by squeezing. By weighted superposition of the deformation-corrected Gaussian functions of all passengers, a mass density field corresponding to the image pixel space is generated, achieving scientific modeling of the continuous mass distribution within the car. The actual centroid coordinates are calculated based on the mass density field, overcoming the limitation of traditional methods that simplify complex pixel distributions to concentrated forces, thus improving the accuracy of centroid calculation. Additional frictional power loss is calculated by the coordinate deviation between the actual centroid and the geometric center, evaluating the impact of uneven load pixel distribution on energy feedback efficiency. Based on the additional frictional power loss, the energy feedback control parameters are dynamically compensated, achieving precise optimization of the control strategy.

[0009] Furthermore, obtaining the top-down view image inside the elevator car includes: performing perspective transformation on the original image captured by a camera located at the top of the elevator car to eliminate perspective distortion introduced by the camera's shooting angle, thereby generating a top-down view image inside the elevator car that reflects the true two-dimensional plane of the elevator car floor.

[0010] Furthermore, the step of identifying the shape center of each passenger in the elevator car and determining the number of passengers includes: using semantic segmentation to output the mask values ​​of each pixel in the top view image, generating a binarized occupancy mask image that distinguishes between passenger-occupied areas and vacant areas; performing a distance transformation on the binarized occupancy mask image to generate a distance map; searching for local maximum pixels in the distance map, determining the position of each local maximum pixel as the shape center of a passenger, and determining the total number of local maximum pixels as the number of passengers.

[0011] Furthermore, the deformation action matrix satisfies:

[0012] In the formula, For the first Deformation effect matrix of each passenger and The first The passenger and the first The coordinates of the center of the passenger's form. These are preset scalar parameters. This is the matrix transpose operator. It is a natural exponential function. is a length of the symbol.

[0013] The application realizes the evaluation of the deformation action matrix by constructing a composite form based on the exponential decay function of the distance between passengers and the outer product of the direction vector; the exponential decay term ensures that the force between passengers at close distance is stronger, and the force between passengers at long distance is weaker; the outer product of the direction vector can effectively represent the directional characteristics of the interaction between passengers, so that the deformation action has a clear directionality, and when the distance between passengers approaches zero, the force is maximized along the connecting line direction.

[0014] Further, the application generates a deformed two-dimensional Gaussian function for each passenger, including: setting a prior covariance matrix representing the standard occupancy form of passengers in a non-crushing state; adding the inverse matrix of the prior covariance matrix and the deformation action matrix to obtain the inverse matrix of a posterior covariance matrix; inverting the inverse matrix of the posterior covariance matrix to obtain the covariance matrix of the deformed two-dimensional Gaussian function.

[0015] The application sets the prior covariance matrix to represent the standard occupancy form of passengers in a non-crushing state, providing a benchmark reference for deformation correction; through the calculation method of adding the inverse of the prior covariance matrix and the deformation action matrix, the covariance matrix update under the Bayesian inference framework is realized; the deformation action matrix as an external information contribution term can effectively correct the prior estimate, so that the covariance matrix can reflect the actual crushing deformation.

[0016] Further, the covariance matrix of the two-dimensional Gaussian function is a diagonal matrix, and the diagonal elements are equal, corresponding to a circular standard occupancy form.

[0017] Further, the application generates a mass density field representing the mass distribution in the elevator car, including: dividing the total mass of all passengers in the elevator car by the number of passengers to obtain the average passenger mass; multiplying the average passenger mass as a weight with the deformed two-dimensional Gaussian function of each passenger, and then summing all the weighted two-dimensional Gaussian functions to obtain the mass density field.

[0018] The application constructs a continuous mass density field by superimposing and summing the weighted two-dimensional Gaussian functions, which can accurately reflect the spatial distribution characteristics of the mass in the car, ensuring that the mass density field has good continuity and smoothness, facilitating subsequent centroid calculation and analysis, fusing the spatial distribution information and mass information of all passengers, generating an accurate mass density field, and improving the accuracy of mass distribution modeling through the superposition of deformed Gaussian functions, providing a reliable physical basis for elevator energy feedback optimization.

[0019] Further, the additional friction power loss satisfies:

[0020] ; wherein, is the additional friction power loss, is the total mass of all passengers in the elevator car, is the gravitational acceleration, is the elevator operating speed, is the vertical distance between the upper and lower guide shoes of the elevator car, and are the effective friction coefficients of the guide shoes and guide rails of the elevator car in the axis and axis directions, and are the distances of the actual center of mass of the elevator car from the geometric center of the elevator car in the axis and axis directions, is the absolute value symbol.

[0021] The present application realizes accurate evaluation of the additional friction power loss by constructing a composite function containing mass, gravity, speed, geometric size and friction coefficient, accurately reflects the influence of asymmetric friction force caused by center of mass offset on energy feedback efficiency, the product of mass and gravity embodies the size of tilting moment, the center of mass offset directly reflects the degree of uneven load distribution, and the effective friction coefficient considers the friction characteristics in different directions, thereby ensuring the comprehensiveness and accuracy of power loss calculation; by calculating the additional friction power loss in real time, the influence of uneven load distribution on system efficiency can be accurately evaluated, and the accuracy and effectiveness of energy feedback optimization control are improved.

[0022] Further, the dynamic compensation of the energy feedback control parameters of the elevator comprises: adding the additional friction power loss value to an initial feedback power reference value calculated based on the total mass of all passengers in the elevator car to obtain a corrected target feedback power; and a frequency converter driving the elevator adjusts the output electromagnetic torque according to the corrected target feedback power, and the energy lost due to additional friction is fed back to the power grid.

[0023] Further, the total mass of all passengers in the elevator car is obtained by a weighing device installed at the bottom of the elevator car.

[0024] The present application has the following beneficial effects:

[0025] (1) Abandon the traditional control strategy to equivalent the total load as the idealized treatment of the geometric center of the concentrated force, through the top view image acquisition pixel level space information, using semantic segmentation model to identify the corresponding pixel area of the passenger, combined with the distance map to locate the morphological center pixel point of each passenger, and then through the mass density field to associate the pixel area with the mass distribution, describe the actual mass distribution in the elevator car, the spatial station difference of the passenger is included in the energy feedback decision, solves the problem that the traditional method cannot perceive the additional friction loss caused by ignoring the centroid offset, provides real and fine load distribution data basis for subsequent energy optimization.

[0026] (2) By introducing the deformation action matrix taking the pixel coordinates of the passenger morphological center as the basis for calculation, the mutual extrusion effect between passengers is evaluated, and the influence of the distance and direction of different passenger pixel areas on the extrusion intensity is quantified; based on the covariance matrix of the modified two-dimensional Gaussian function, the function can dynamically simulate the passenger occupancy form change caused by extrusion, such as the flattening of the pixel area and the mass superposition of the overlapping area in the crowd, compared with the traditional model which simply regards the passenger as a rigid point and ignores the pixel form change, the processing method of the present application is more in line with the real form presented by the image pixels in the actual passenger carrying scene, so that the accuracy of the actual centroid coordinates is improved, and the calculation of the additional friction loss is more accurate.

[0027] (3) Based on the coordinate deviation between the actual centroid and the geometric center, the additional friction power loss caused by uneven load distribution is calculated, the deviation calculation takes the image pixel coordinate system as the reference, which can accurately reflect the influence of the asymmetric distribution of the passenger pixel area on the centroid position; and the energy feedback control parameters are dynamically adjusted according to the loss, for example, when the passengers are concentrated on one side of the car, the passenger pixel area on that side in the image is larger, resulting in an increase in the centroid offset, at this time the compensation parameter can actively adapt to the additional loss, the energy consumed by the asymmetric friction force is brought back into the feedback range, avoiding energy waste, and effectively improving the efficiency of elevator energy feedback. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is a step flow chart of a passenger distribution monitoring method for elevator car energy feedback optimization according to an embodiment of the present application.

[0029] Figure 2 is a schematic diagram of a two-dimensional Gaussian function of a passenger distribution monitoring method for elevator car energy feedback optimization according to an embodiment of the present application. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present application will be described clearly and completely below, and the described embodiments are part of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0031] The specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0032] Please refer to Figure 1 which shows a step flow chart of a passenger distribution monitoring method for elevator car energy feedback optimization provided by an embodiment of the present application, the method comprising the following steps:

[0033] S01: Obtain an overhead image of the elevator car.

[0034] It should be noted that in order to perceive the passenger distribution inside the car, an image acquisition device needs to be installed in the elevator car, a wide-angle camera is installed at the central position of the top of the car, and the central position of the top is selected to minimize the dead angle of shooting and ensure that the field of view can completely cover the entire car floor. The camera real-time collects the video stream inside the car and decodes it into a series of independent image frames. Due to the installation angle of the camera and the optical properties of the lens itself, there will be perspective distortion, i.e. objects close to the camera appear large and objects far away appear small. The rectangular floor may appear trapezoidal in the image, making it impossible to directly measure the position and distance on the image. Therefore, the image must be corrected first to convert it into an overhead view that can truly reflect the physical space of the car floor, and the geometric center of the elevator car floor is taken as the origin, the axis points to the elevator door, the axis points to the side wall of the elevator car, and a two-dimensional coordinate system is constructed.

[0035] Specifically, the overhead image of the elevator car is obtained, comprising:

[0036] At the first installation or maintenance of the system, a one-time calibration procedure such as Zhang Zhengyou calibration method is performed to obtain the internal and external parameter matrices of the camera. These parameters are used to perform perspective transformation on each original image to generate a non-distorted, overhead view of the elevator car floor image, i.e. the overhead image of the elevator car.

[0037] S02: Use semantic segmentation to output the mask value of each pixel point in the overhead image to identify the shape center of each passenger in the elevator car and determine the number of passengers.

[0038] It should be noted that after obtaining the overhead image, it is necessary to distinguish from the overhead image which areas are occupied by passengers and which are free floor, and to determine the center position of each independent passenger, which is the basis for establishing an independent model for each passenger.

[0039] Specifically, the shape center of each passenger in the elevator car is identified and the number of passengers is determined, comprising:

[0040] Segmenting the passenger area, using a semantic segmentation model based on deep learning, such as a lightweight U-Net or SegNet model, inputting the overhead view image into the model, and outputting a binary occupancy mask image, in which the mask value of the passenger-occupied area is 1 (such as white), and the mask value of the idle floor is 0 (such as black);

[0041] In a crowded scene, the occupancy mask may be connected in one piece, and it is necessary to separate individuals from it, for which the binary occupancy mask is distance transformed, and the distance transformation will generate a distance map, in which the brightness value of each foreground pixel (pixel point with mask value 1) is equal to the distance to the nearest background pixel (pixel point with mask value 0);

[0042] Since the center area of each passenger's body is farthest from the idle area, a local brightness peak will be formed in the distance map, and by searching for these local maximum value pixels in the distance map, the morphological center of each passenger can be accurately determined, and the total number of local maximum value pixels is determined as the number of passengers.

[0043] S03: For any passenger in the elevator car, a deformation action matrix is constructed to represent the squeezing effect of other passengers on the passenger.

[0044] It should be noted that the traditional method is to identify the center of each person's head, and then simply draw a circle of the same size at each center point to represent the passenger, so that in a crowded situation, these circles will overlap a lot, and cannot reflect the real space occupation situation, and the calculated total center of gravity will also have a large deviation. In order to reflect the real space occupation situation, the space occupation situation of each passenger should be simulated. The present application uses a two-dimensional Gaussian function to approximately simulate the space occupation situation of each passenger. The morphological center of each passenger corresponds to the center point of its two-dimensional Gaussian function, and each passenger's two-dimensional Gaussian function has a 2x2 covariance matrix. The element in the first row and the first column corresponds to the width of the axis, corresponding to the width of the passenger axis, the elements in the first row and the second column and the second row and the first column are equal, which control the rotation direction of the two-dimensional Gaussian function, corresponding to the rotation direction of the passenger; the element in the second row and the second column corresponds to the width of the axis, corresponding to the width of the passenger axis. Thus, referring to Figure 2Each passenger corresponds to a one-to-one two-dimensional Gaussian function to simulate their spatial occupancy within the elevator. Since the height of the Gaussian function is fixed, this invention essentially assumes that each passenger has the same height, focusing on the mass distribution of passengers within the elevator car. Inside the elevator car, as passenger density increases, the effective space occupied by each person deforms due to the presence of neighboring passengers. The occupancy shape of an independent passenger can be approximated as a circle or ellipse (i.e., the cross-section of the two-dimensional Gaussian function). However, in crowded conditions, the circle or ellipse will be compressed under pressure; therefore, a mathematical model is needed to describe this deformation.

[0045] Specifically, the deformation action matrix satisfies:

[0046] ;

[0047] In the formula, For the first Deformation effect matrix of each passenger and The first The passenger and the first The coordinates of the center of the passenger's form. These are preset scalar parameters. This is the matrix transpose operator. It is a natural exponential function. This is the modulus symbol.

[0048] in, It is a standard Gaussian function in form, representing the first Gaussian function. The passenger on the first The influence strength of each passenger, with a value range of [value range]. When passengers With passengers distance When the distance is very close, the value of this term is close to 1, indicating a strong squeezing effect. As the distance increases, the value of this term decays exponentially and rapidly, approaching 0, indicating that the squeezing effect can be ignored. The projection matrix on the right is a 2×2 matrix. In linear algebra, the cross product of a unit vector and itself generates a projection matrix. When the projection matrix is ​​multiplied by any other vector, it yields the projection of that vector onto the direction of the unit vector. This represents the unit vector mentioned above. For passengers... The formula iterates through all other passengers. For each other passenger First, the intensity of its influence is calculated. Then, a projection matrix containing only the direction of influence is constructed. Finally, the two are multiplied to obtain the values ​​of other passengers. For passengers The generated stresses are ultimately summed to obtain the total deformation matrix. .

[0049] S04: Based on the deformation effect matrix, a deformation-corrected two-dimensional Gaussian function is generated for each passenger, where the covariance matrix of the two-dimensional Gaussian function reflects the change in passenger occupancy pattern caused by the squeezing effect.

[0050] Specifically, generating a deformation-corrected two-dimensional Gaussian function for each passenger includes:

[0051] Assuming the first [uncompressed] The covariance matrix of the two-dimensional Gaussian function for each passenger is: (Based on the preset dimensions of the elevator car space, representing the standard occupancy pattern when passengers are not crowded), its inverse matrix It is a priori;

[0052] Deformation action matrix As a likelihood term, it quantifies the degree to which squeezing modifies passenger occupancy patterns;

[0053] According to Bayes' theorem, the covariance matrix after compression The inverse matrix satisfies: That is, the superposition of the prior inverse matrix and the likelihood matrix reflects the correction of the occupancy pattern by the squeeze;

[0054] Again Find the inverse to get the first... The covariance matrix of a two-dimensional Gaussian function after a passenger is squeezed. .

[0055] Specifically, the covariance matrix of the two-dimensional Gaussian function is a diagonal matrix with equal diagonal elements, corresponding to a standard circular occupancy shape.

[0056] S05: The deformation-corrected two-dimensional Gaussian functions of each passenger are weighted and superimposed to generate a mass density field that characterizes the mass distribution inside the elevator car. Based on the mass density field, the coordinates of the actual centroid of the elevator car are obtained using the centroid formula.

[0057] It should be noted that in order to obtain the actual center of mass of the elevator car, it is necessary to know the mass of every point inside the elevator car. The effective space occupied by each passenger has been obtained above, but the mass corresponding to each space is unknown. Therefore, it is necessary to first assign a two-dimensional Gaussian function mass to each passenger.

[0058] Specifically, the generation of the mass density field characterizing the mass distribution within the elevator car includes:

[0059] The total mass of all passengers inside the elevator car is obtained by a weighing device installed at the bottom of the elevator car.

[0060] The average passenger mass is obtained by dividing the total mass of all passengers in the elevator car by the number of passengers.

[0061] The average passenger mass is used as a weight, multiplied by the deformation-corrected two-dimensional Gaussian function of each passenger, and then all weighted two-dimensional Gaussian functions are summed to obtain the mass density field.

[0062] S06: Calculate the additional frictional power loss caused by uneven mass distribution based on the coordinate deviation between the actual center of mass and the geometric center of the elevator car.

[0063] Specifically, the additional frictional power loss satisfies:

[0064] ;

[0065] In the formula, To account for additional frictional power loss, This refers to the total mass of all passengers inside the elevator car. It is the acceleration due to gravity. For elevator operating speed, This is the vertical distance between the upper and lower guide shoes of the elevator car. and The guide shoes and guide rails of the elevator car were obtained through on-site measurements. shaft and The effective coefficient of friction in the axial direction, for example, in this embodiment, and Both are 0.15. and The actual center of mass of the elevator car is at shaft and The distance from the geometric center of the elevator car along the axial direction. It is the absolute value symbol.

[0066] The additional friction power loss is constructed based on the additional friction mechanism caused by passenger mass eccentricity, when the total mass is... The passenger's center of gravity is off from the center of the elevator car. At that time, its gravity A tilting moment is generated relative to the center of the elevator car. This tilting moment is caused by the upper and lower guide shoes (vertical spacing is...). The force couple formed by the force couple is used to balance the force, thereby introducing an additional horizontal normal force on the guide shoe; specifically, Normal force in the axial direction and Proportional Normal force in the axial direction and The normal force is directly proportional to the coefficient of kinetic friction in the corresponding direction. and The multiplication of the two additional friction force components in two directions, whose equivalent contribution is ; as the elevator runs in the vertical direction with a speed , the power of the friction force working in the direction of motion is equal to the equivalent additional friction force multiplied by the speed , so the gravity is multiplied by the speed , and divided by the guide shoe spacing , and finally multiplied by the additional friction force in two directions , to reflect the influence of the mass distribution, structural geometry, friction characteristics and running state on the energy consumption, to obtain the additional friction power loss.

[0067] S07: According to the additional friction power loss, the energy feedback control parameters of the elevator are dynamically compensated, and the passenger distribution monitoring of the elevator car for elevator energy feedback optimization is completed.

[0068] Specifically, the dynamic compensation of the energy feedback control parameters of the elevator includes:

[0069] The additional friction power loss value is added to the initial feedback power reference value calculated based on the total mass of all passengers in the elevator car, to obtain the corrected target feedback power;

[0070] The correction of the motor electromagnetic torque in the vector control mode ensures that the energy feedback current matches the actual friction loss, and in the direct torque control mode, the electromagnetic torque output / absorbed by the motor is directly corrected by switching the voltage vector, to ensure that the feedback current is not disturbed by the additional friction;

[0071] After PWM modulation, the inverter feeds back the corrected current to the power grid, to realize the effective recovery of additional energy.

[0072] The above only describes the preferred embodiments of the present application, and does not limit the present application, any modification, equivalent replacement, improvement, etc. within the principles of the present application, should be included in the protection scope of the present application.

Claims

1. A method for elevator car passenger distribution monitoring for elevator energy feedback optimization, characterized by, The method comprises the following steps: obtaining an overhead image in an elevator car; using semantic segmentation to output mask values of each pixel point in the overhead image to identify the shape center of each passenger in the elevator car and determine the number of passengers; for any passenger in the elevator car, constructing a deformation action matrix for representing the deformation effect of other passengers on the passenger; based on the deformation action matrix, generating a deformation-corrected two-dimensional Gaussian function for each passenger, wherein the covariance matrix of the two-dimensional Gaussian function reflects the change in the passenger's occupation shape caused by the deformation effect; weighting and superimposing the deformation-corrected two-dimensional Gaussian function of each passenger to generate a mass density field representing the mass distribution in the elevator car, and using the mass center formula to obtain the coordinates of the actual mass center of the elevator car based on the mass density field; calculating the additional friction power loss caused by the uneven mass distribution according to the coordinate deviation between the actual mass center and the geometric center of the elevator car; dynamically compensating the energy feedback control parameters of the elevator according to the additional friction power loss to complete the passenger distribution monitoring of the elevator car for optimizing the energy feedback of the elevator.

2. The method for elevator car passenger distribution monitoring for elevator energy feedback optimization of claim 1, wherein, The method comprises the following steps: performing perspective transformation on the original image captured by the camera located at the top of the elevator car to eliminate the perspective distortion introduced by the camera shooting angle, and generating an overhead image in the elevator car reflecting the real two-dimensional plane of the floor of the elevator car.

3. The method for elevator car passenger distribution monitoring for elevator energy- feedback optimization of claim 1, wherein, The method comprises the following steps: using semantic segmentation to output mask values of each pixel point in the overhead image to generate a binary occupancy mask image that distinguishes between passenger occupied areas and free areas; performing distance transformation on the binary occupancy mask image to generate a distance map; searching for local maximum pixel points in the distance map, determining the position of each local maximum pixel point as the shape center of a passenger, and determining the total number of local maximum pixel points as the number of passengers.

4. The method for elevator car passenger distribution monitoring for elevator energy- feedback optimization of claim 1, wherein, The deformation action matrix satisfies: ； wherein is the deformation matrix of the and is the is the is the is a predetermined scalar parameter, is the transpose operator of a matrix, is the natural exponential function, is the modulus length symbol.

5. The method for elevator car passenger distribution monitoring for elevator energy- feedback optimization of claim 1, wherein, The method comprises the following steps: setting a prior covariance matrix representing the standard occupation shape of a passenger in a non-deformation state; adding the inverse matrix of the prior covariance matrix to the deformation action matrix to obtain the inverse matrix of a posterior covariance matrix; inverting the inverse matrix of the posterior covariance matrix to obtain the covariance matrix of the deformation-corrected two-dimensional Gaussian function.

6. The method according to claim 1 or 5, characterized in that, The covariance matrix of the two-dimensional Gaussian function is a diagonal matrix with equal diagonal elements, corresponding to a circular standard occupation shape.

7. The method for elevator car passenger distribution monitoring for elevator energy- feedback optimization of claim 1, wherein, The method comprises the following steps: dividing the total mass of all passengers in the elevator car by the number of passengers to obtain the average passenger mass; multiplying the average passenger mass by the deformation-corrected two-dimensional Gaussian function of each passenger, and then summing all the weighted two-dimensional Gaussian functions to obtain the mass density field.

8. The method for elevator car passenger distribution monitoring for elevator energy- feedback optimization of claim 1, wherein, The additional friction power loss satisfies: ； wherein is the additional frictional power loss, is the total mass of all passengers inside the elevator car, is the gravitational acceleration, is the elevator operating speed, is the vertical distance between the upper and lower guide shoes of the elevator car, is the vertical distance between the upper and lower guide shoes of the elevator car, are the guide shoes and guide rails of the elevator car, respectively, in the axis direction, are the effective friction coefficients in the axis direction, are the distances of the actual center of mass of the elevator car from the geometric center of the elevator car in the axis direction, axis direction, is the absolute value symbol.

9. The method for elevator car passenger distribution monitoring for elevator energy- feedback optimization of claim 1, wherein, The method comprises the following steps: adding the additional friction power loss value to the initial feedback power reference value calculated based on the total mass of all passengers in the elevator car to obtain a corrected target feedback power; The frequency converter driving the elevator adjusts the output electromagnetic torque according to the corrected target feedback power, and feeds back the energy lost due to additional friction to the power grid.

10. A method for elevator car passenger distribution monitoring for elevator energy feedback optimization according to claim 7 or 9, characterized in that, The total mass of all passengers in the elevator car is obtained by a weighing device installed at the bottom of the elevator car.

Images