Elevator space proportion recognition method based on single camera

Abstract

This invention discloses a method for recognizing elevator space occupancy based on a single camera, comprising the following steps: Step S1, calculating the total available space volume of the elevator car, setting the space volume occupied by a baseline unit, and establishing a database of conversion ratios between passengers and preset items and the baseline unit; Step S2, installing a single camera on the top of the elevator car so that the camera's shooting angle can cover the entire interior space of the car; Step S3, acquiring real-time video images of the car and segmenting and extracting passenger and item areas; Step S4, using a lightweight deep learning model to identify and count the number of passengers and various items; Step S5, counting the total number of equivalent baseline units based on the conversion ratio database, and calculating the real-time space occupancy based on the total available space volume of the car; Step S6, outputting the space occupancy result to a display terminal or elevator control system, and issuing a congestion warning when a preset threshold is reached. This invention can accurately identify elevator space occupancy using a single camera.

Description

Technical Field

[0001] This invention relates to the field of elevator technology, and more specifically to a method for recognizing the spatial occupancy of an elevator based on a single camera. Background Technology

[0002] In existing technologies, the feeling of crowding in an elevator depends on the proportion of space in the car rather than the load. However, current spatial recognition relies on multi-angle cameras and 3D modeling, which has problems such as high hardware costs, large investment in multiple cameras and supporting equipment, and difficulty in promotion. At the same time, the installation of multi-camera spatial acquisition devices is complicated, requiring centered installation that may damage the elevator's decoration, making construction difficult. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method for accurately identifying the space ratio of an elevator using a single camera.

[0004] To address the aforementioned technical problems, this invention provides a method for recognizing elevator space occupancy based on a single camera, comprising the following steps: Step S1: Calculate the total available space volume of the elevator car, set the space volume occupied by the reference unit, and establish a database of conversion ratios between passengers and preset items and the reference unit. Step S2: Install a single camera on the top of the elevator car so that the camera's field of view can cover the entire interior space of the car; Step S3: Real-time acquisition of video images from the elevator car, segmentation and extraction of passenger and item areas; Step S4: Use a lightweight deep learning model to identify and count the number of passengers and various items; Step S5: Calculate the total number of equivalent benchmark units based on the conversion ratio database, and calculate the real-time space ratio in combination with the total available space volume of the car. Step S6: Output the space occupancy result to the display terminal or elevator control system, and issue a congestion warning when the preset threshold is reached.

[0005] Preferably, the space volume occupied by the reference unit is the average space volume occupied by a single standard adult.

[0006] Preferably, the single camera is a fisheye or wide-angle camera, installed at the top left corner of the elevator car at a height of 2.3m and with a tilt angle adjusted to 40°. The PTZ digital correction algorithm is used to correct distortion in the captured image.

[0007] Preferably, in step S3, Gaussian filtering algorithm is used to remove image noise for each frame of image, grayscale processing is used to simplify image data, histogram equalization algorithm is used to enhance image contrast, and threshold segmentation algorithm is used to extract the area containing passengers and items.

[0008] Preferably, in step S4, the lightweight deep learning model is a target detection model optimized based on YOLOv8.

[0009] This invention employs a single camera to identify elevator space occupancy, optimizing installation and reducing hardware costs. It proposes a conversion algorithm between a baseline unit and items to achieve accurate and efficient space occupancy calculation. A database of conversion ratios between various elevator-riding items and the baseline unit is established, using the standard adult's space occupancy as the baseline unit. The conversion relationship is determined based on the actual floor area occupied by the items. A lightweight deep learning model identifies passengers and elevator-riding items, calculates the total equivalent baseline unit count, and combines this with the real-time conversion of the total car space occupancy. This eliminates the need for 3D modeling, improving real-time recognition and accurately reflecting the actual space conditions of the car. It solves the problems of inaccurate traditional load assessment and low efficiency in existing space recognition methods.

[0010] This technology abandons the multi-angle camera deployment mode of existing technologies, using only a single wide-angle or fisheye camera installed in the top corner of the elevator car, eliminating the need for centered installation and avoiding damage to the original elevator decor. Through angle calibration and distortion correction technology, it ensures that the camera's field of view covers the entire car without blind spots. This eliminates the need for high-performance 3D modeling equipment, significantly reducing hardware procurement and installation costs. It is adaptable to various elevator scenarios, addressing the pain points of existing technologies such as high hardware costs, difficulty in promotion, and installation damage to the decor. Attached Figure Description

[0011] none Detailed Implementation

[0012] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can fully understand other advantages and technical effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through different specific embodiments, and the details in this specification can also be applied based on different viewpoints, with various modifications or changes made without departing from the overall design concept of the invention. It should be noted that, unless otherwise specified, the following embodiments and features can be combined with each other. The following exemplary embodiments of the present invention can be implemented in many different forms and should not be construed as being limited to the specific embodiments set forth herein. It should be understood that these embodiments are provided to make the disclosure of the present invention thorough and complete, and to fully convey the technical solutions of these exemplary embodiments to those skilled in the art. Example 1

[0013] This embodiment provides a method for recognizing the spatial proportion of an elevator based on a single camera. The specific steps are as follows: Step S1: Preset parameter calibration: Select a residential elevator as the test object. The elevator car has an internal length of 1.8m, a width of 1.5m, and a height of 2.5m. Calculate the total usable space volume of the car, V = 1.8 × 1.5 × 2.5 = 6.75m³. Select 10 standard-sized adults with a height of 1.6-1.8m and a weight of 50-70kg. Have them stand naturally in the car. Capture their standing images through a camera. Combined with the car space parameters, calculate the average space volume occupied by a single standard adult as 0.25m³, i.e., the baseline unit V = 0.25m³. Establish a conversion ratio database and preset the conversion ratios for common items used in elevators: trolley (supermarket shopping cart, delivery cart) conversion ratio k1 = 2, wheelchair conversion ratio k2 = 2.5, suitcase (28 inches and below) conversion ratio k3 = 1, suitcase (above 28 inches) conversion ratio k4 = 1.5, stroller conversion ratio k5 = 1.8.

[0014] Step S2, Camera Installation and Debugging: Select a wide-angle camera and install it at the top left corner (corner position) of the elevator car, at a height of 2.3m and an angle of 40° to ensure that the camera's field of view covers the entire interior space of the car without any blind spots; debug the camera, adjust the image clarity, and use the PTZ digital correction algorithm to correct the distortion of the captured image, eliminating the slight distortion caused by the wide-angle lens, ensuring that passengers and various items on the elevator can be clearly identified, while avoiding capturing passengers' faces to protect their privacy.

[0015] Step S3, Image Acquisition and Preprocessing: Video images inside the elevator car are acquired in real time using the aforementioned camera, with a frame rate of 30 frames per second. Each frame is preprocessed: Gaussian filtering is used to remove image noise, grayscale processing is used to simplify image data, histogram equalization is used to enhance image contrast, and threshold segmentation is used to extract regions of interest (ROIs) containing passengers and items on the elevator. Irrelevant background areas such as the elevator walls, control panel, and ceiling decorations are removed to reduce the amount of subsequent recognition calculations.

[0016] Step S4, Target Recognition and Classification: A target detection model optimized based on the YOLOv8 lightweight model is adopted. This model is trained with more than 10,000 sample images of passengers, trolleys, wheelchairs, and suitcases in the elevator car. The samples cover different lighting conditions (strong light, weak light, backlight) and different occlusion scenarios (passengers occluding each other, objects occluding each other). The preprocessed effective image area is input into the model to identify passengers and various items in the elevator in real time. The number of passengers in the current elevator car is N=3, the number of trolleys is N=1, and the number of suitcases (less than 28 inches) is N=1.

[0017] Step S5, Space Ratio Calculation: Based on the conversion ratio database, obtain the conversion ratio k1=2 for trolleys and k3=1 for suitcases (under 28 inches). Calculate the total number of equivalent reference units N=3+(1×2)+(1×1)=6; then calculate the real-time space ratio R=(6×0.25) / 6.75×100%≈22.2%.

[0018] Step S6, Result Output and Feedback: The space occupancy of 22.2% is output to the display terminal inside the elevator car in real time, displaying "The current elevator space is spacious, and you can ride normally"; at the same time, the space occupancy data is sent to the elevator control system to provide a basis for elevator scheduling; if subsequent passengers enter, when the space occupancy reaches 80% (i.e., N≥21.6, rounded to 22), the display terminal outputs a prompt signal "The elevator is crowded, please wait a while before riding".

[0019] The present invention has been described in detail above through specific embodiments and examples, but these are not intended to limit the invention. Many modifications and improvements can be made by those skilled in the art without departing from the principles of the invention, and these should also be considered within the scope of protection of the present invention.

Claims

1. A method for recognizing elevator space occupancy based on a single camera, characterized in that, Includes the following steps: Step S1: Calculate the total available space volume of the elevator car, set the space volume occupied by the reference unit, and establish a database of conversion ratios between passengers and preset items and the reference unit. Step S2: Install a single camera on the top of the elevator car so that the camera's field of view can cover the entire interior space of the car; Step S3: Real-time acquisition of video images from the elevator car, segmentation and extraction of passenger and item areas; Step S4: Use a lightweight deep learning model to identify and count the number of passengers and various items; Step S5: Calculate the total number of equivalent benchmark units based on the conversion ratio database, and calculate the real-time space ratio in combination with the total available space volume of the car. Step S6: Output the space occupancy result to the display terminal or elevator control system, and issue a congestion warning when the preset threshold is reached.

2. The elevator space occupancy identification method according to claim 1, characterized in that, The space volume occupied by the reference unit is the average space volume occupied by a single standard adult.

3. The elevator space occupancy identification method according to claim 1, characterized in that, The single camera is a fisheye or wide-angle camera, installed at the top left corner of the elevator car, at a height of 2.3m, with the tilt angle adjusted to 40°.

4. The elevator space occupancy identification method according to claim 3, characterized in that, The PTZ digital correction algorithm is used to correct distortion in the captured images.

5. The elevator space occupancy identification method according to claim 1, characterized in that, In step S3, Gaussian filtering algorithm is used to remove image noise for each frame of image, grayscale processing is used to simplify image data, histogram equalization algorithm is used to enhance image contrast, and threshold segmentation algorithm is used to extract the region containing passengers and items.

6. The elevator space occupancy identification method according to claim 1, characterized in that, In step S4, the lightweight deep learning model is a target detection model optimized based on YOLOv8.