CN116788960B - Elevator door pre-sensing anti-pinch system and method based on three-dimensional point cloud deep learning - Google Patents
Elevator door pre-sensing anti-pinch system and method based on three-dimensional point cloud deep learningInfo
- Publication number
- CN116788960B CN116788960B CN202310753474.2A CN202310753474A CN116788960B CN 116788960 B CN116788960 B CN 116788960B CN 202310753474 A CN202310753474 A CN 202310753474A CN 116788960 B CN116788960 B CN 116788960B
- Authority
- CN
- China
- Prior art keywords
- point cloud
- elevator door
- foreign object
- door
- elevator
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Elevator Door Apparatuses (AREA)
Abstract
S1, acquiring the space in the closing path of a car door and a hall door by using a three-dimensional scanning device, obtaining a point cloud image through three-dimensional scanning, and obtaining a simplified point cloud of foreign matters through the point cloud image; S2, based on the simplified point cloud, through a pre-trained convolutional neural network model, outlier point cloud data are obtained, and the outlier state is judged. By adopting the three-dimensional point cloud deep learning method, various foreign matters in the closing process of the elevator door can be accurately identified, the closing time of the elevator door is shortened, and the operation efficiency of the elevator is improved.
Description
Technical Field
The application belongs to the technical field of elevator equipment, and particularly relates to an elevator door pre-induction type anti-pinch system and method based on three-dimensional point cloud deep learning.
Background
Elevators are indispensable equipment in everyday life, elevator doors are important responsibilities for protecting passengers as key parts in an elevator system, and safety accidents caused by elevator doors are mostly caused by the fact that foreign matters on the way of closing are not detected when the elevator doors are closed.
The existing anti-pinch method of the elevator door is mainly divided into two types, wherein one type is a touch plate type, and the other type is a light curtain type. The touch plate type is mechanically controlled by an electric switch through a protruding movable touch plate, so that the elevator door is controlled to stop closing and reversely open, and contact type anti-clamping is realized. The light curtain type electric switch is controlled by a circuit of an electric switch according to an infrared induction principle, so that the elevator door is controlled to stop closing and reversely open, and non-contact anti-clamping is realized. However, when the touch panel type anti-pinch method encounters some objects with smaller resistance, the touch panel electric switch cannot be triggered, so that the anti-pinch device fails. Similarly, the infrared devices of the light curtain type anti-clamping method are not distributed on the whole elevator door from top to bottom, so that a plurality of blind areas exist, and when the infrared devices meet objects with small shading rate, foreign objects cannot be identified, and the anti-clamping device is invalid.
The existing elevator anti-pinch method based on image processing is used for identifying the foreign matters through an image comparison method, and the defects existing in the existing method are overcome to a certain extent, but when the foreign matters are not existing between elevator doors completely as in the existing two methods, the elevator door closing action is performed, the pre-induction identification of the entry and exit of the foreign matters cannot be achieved, and the operation efficiency of an elevator is greatly reduced.
Therefore, how to accurately and effectively identify the foreign matters in the closing process of the elevator door is a problem to be solved in the current stage.
Disclosure of Invention
The invention aims to provide an elevator door pre-induction type anti-pinch system and method based on three-dimensional point cloud deep learning, so as to solve the problems in the prior art.
In order to achieve the aim, the invention provides an elevator door pre-induction type anti-pinch system based on three-dimensional point cloud deep learning,
The system comprises a car door, a three-dimensional scanning device and a hall door;
the three-dimensional scanning device is arranged on the car door;
the scanning range of the three-dimensional scanning device comprises all areas of the car door and the hall door on the way of closing.
Optionally, the three-dimensional scanning device is totally 4, and each left and right car door is provided with 2.
The application also discloses an anti-pinch method of the elevator door pre-induction type anti-pinch system based on the three-dimensional point cloud deep learning, which comprises the following steps:
S1, acquiring a space in a closing path of a sedan door and a hall door by using the three-dimensional scanning device, obtaining a point cloud image through three-dimensional scanning, and obtaining a simplified point cloud of a foreign object through the point cloud image;
s2, based on the simplified point cloud, acquiring outlier point cloud data through a pre-trained convolutional neural network model, and judging the outlier state.
Optionally, the foreign matter state includes:
The elevator door closing area is free of foreign matters, the foreign matters are in an elevator door entering state, the foreign matters are in an elevator door room state, the foreign matters are in an elevator door exiting state, and the foreign matters are in an elevator door range state completely.
Optionally, the S1 includes:
s11, in the closing process of the car door and the hall door, three-dimensional scanning is carried out on the space in the closing path of the car door and the hall door through a three-dimensional scanning device arranged on the car door, so that a point cloud picture is obtained:
s12, according to the pitch angles and the distances of the 4 three-dimensional scanning devices, calculating a rotation translation matrix among 4 point clouds obtained by the 4 three-dimensional scanning devices, and transforming the 4 point clouds into 1 complete point cloud image under the same coordinate system;
S13, comparing the point cloud image after the multi-view cloud data registration with a prestored point cloud image without foreign matters, and extracting point cloud belonging to the foreign matters;
S14, removing noise of the alien point Yun Zhongfu from normal distribution by adopting a Gaussian filtering method, and eliminating burrs of the alien point cloud reconstruction curved surface by using a laser point cloud smoothing technology;
s15, downsampling the filtered and denoised foreign matter point cloud by adopting a furthest point sampling method so as to obtain a simplified point cloud.
Optionally, the S2 includes:
inputting the simplified point cloud data of the foreign matters into a pre-trained convolutional neural network model, and carrying out feature aggregation on the extracted point cloud data of the foreign matters to form a 128-dimensional global feature vector;
mapping the feature vector to interval [0,1] representing the distance of the foreign object relative to the elevator door at the time;
And comparing the distance and the position of the different object point cloud data of each frame relative to the elevator door, and pre-sensing the range of the closed area of the elevator door occupied by the foreign object through the distance change of the different object point cloud data relative to the elevator door, wherein the state of the foreign object is the state of entering the elevator door or exiting the elevator door through the position change of the different object point cloud data.
Alternatively, when there is no foreign matter in the elevator door closing area, the elevator door is closed at a normal speed;
when the foreign matters in the elevator door closing area are not found, judging that the foreign matters are in a state of entering the elevator door, and opening the elevator door at a normal speed;
when the foreign matter in the elevator door closing area still exists and the tail part does not enter the elevator, judging that the foreign matter is still in a state of entering the elevator door, and opening the elevator door at a normal speed;
when the tail part of the foreign matter in the elevator door closing area completely enters the elevator door and the foreign matter is judged to be in an elevator door outlet state, the elevator door presensories the range of the elevator door closing area occupied by the foreign matter according to the distance between the foreign matter and the elevator door, and the elevator door closing time is shortened;
when no foreign matter exists in the elevator door closing area and the foreign matter has completely exited the elevator door range, the elevator door is closed at a normal speed.
Compared with the prior art, the application has the beneficial effects that:
the method for deep learning by using the three-dimensional point cloud can accurately identify various foreign matters in the closing process of the elevator door, improves the safety performance of the elevator door, can realize the pre-sensing of the entry and exit of the foreign matters through the distance and the position change of the foreign matters relative to the elevator door, closes the elevator door along with the relative distance and the position change of the foreign matters, shortens the closing time of the elevator door and improves the operation efficiency of the elevator.
Drawings
In order to more clearly illustrate the technical solutions of the present application, the drawings that are needed in the embodiments are briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method of an elevator door pre-induction anti-pinch system and method based on three-dimensional point cloud deep learning in accordance with an embodiment of the present application;
FIG. 2 is a front view (with hall doors hidden) of an elevator door pre-feel pinch prevention system and method based on three-dimensional point cloud deep learning in accordance with an embodiment of the present application;
FIG. 3 is a top view of a method operation of an elevator door pre-induction anti-pinch system and method based on three-dimensional point cloud deep learning in accordance with an embodiment of the present application;
The reference numerals indicate 1-car door, 2-three-dimensional scanning device, 3-scanning range, 4-hall door, 5-foreign matter.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In order that the above-recited objects, features and advantages of the present application will become more readily apparent, a more particular description of the application will be rendered by reference to the appended drawings and appended detailed description.
Example 1
In this embodiment, as shown in fig. 1, an elevator door pre-induction type anti-pinch system based on three-dimensional point cloud deep learning comprises a car door 1, a three-dimensional scanning device 2 and a hall door 4.
In a further optimized scheme, the three-dimensional scanning devices 2 are specifically 4, and are respectively arranged on the 2 car doors 1, and the scanning range 3 of the three-dimensional scanning devices comprises all areas on the way of closing the car doors 1 and the hall doors 4.
Example two
An elevator door pre-induction type anti-pinch method based on three-dimensional point cloud deep learning comprises the following steps:
s1, acquiring a space in the closing path of a car door 1 and a hall door 4 by using a three-dimensional scanning device 2, obtaining a point cloud image through three-dimensional scanning, and obtaining a simplified point cloud of a foreign object through the point cloud image;
s2, based on the simplified point cloud, through a pre-trained convolutional neural network model, outlier point cloud data are obtained, and the outlier state is judged.
In a further optimization scheme, as shown in fig. 1, the processing of the S1 point cloud data includes the following steps:
S11, multi-view cloud data acquisition;
S12, multi-view cloud data registration;
s13, foreign object point cloud semantic segmentation;
s14, filtering and denoising the foreign matter point cloud;
s15, sampling the alien point cloud data.
In a further optimization scheme, as shown in fig. 2, the specific method for acquiring the S11 multi-view cloud data comprises the following steps of carrying out three-dimensional scanning on a space in the closing path of the car door 1 and the hall door 4 through a three-dimensional scanning device 2 arranged on the car door 1 in the closing process of the car door 1 and the hall door 4 to obtain a point cloud image. The specific method for registering the S12 multi-view cloud data comprises the steps of calculating a rotation translation matrix among 4 point clouds obtained by the 4 three-dimensional scanning devices 2 according to the pitching angles and the pitches of the 4 three-dimensional scanning devices 2, and transforming the 4 point clouds into 1 complete point cloud in the same coordinate system. The specific method for S13 alien point cloud semantic segmentation is that the point cloud image after multi-view point cloud data registration is compared with the prestored point cloud image without the alien substance 5, and the point cloud belonging to the alien substance 5 is extracted. The specific method for removing noise by filtering the foreign object point cloud in S14 comprises the steps of removing noise of the foreign object point Yun Zhongfu from normal distribution by adopting a Gaussian filtering method, and removing burrs of a reconstruction curved surface of the foreign object point cloud by adopting a laser point cloud smoothing technology. The specific method for sampling the S15 foreign object point cloud data comprises the step of downsampling the filtered and denoised foreign object point cloud by adopting a furthest point sampling method so as to obtain a simplified point cloud.
In a further optimization scheme, as shown in fig. 1, the S2 foreign matter 5 state analysis includes the following steps:
s21, inputting different object point cloud data;
s22, point cloud data multi-layer convolution;
s23, pooling point cloud data;
S24, enabling output by Softmax;
s25, comparing the front and rear foreign matter 5 data.
In a further optimization scheme, as shown in fig. 2, the specific method for inputting the S21 outlier point cloud data is to input the outlier simplified point cloud data into a pre-trained convolutional neural network model. The specific method of the S22 point cloud data multi-layer convolution comprises the steps of convolving input foreign object point cloud data for a plurality of times by adopting a convolution kernel shared by a plurality of weights, and extracting characteristics of each foreign object point cloud data. The specific method for pooling the S23 point cloud data comprises the step of collecting the characteristics of the extracted foreign object point cloud data into a 128-dimensional global characteristic vector. The specific method of S24Softmax activation output is to map the feature vector to interval 0,1, representing the distance of the foreign object 5 relative to the elevator door at this time. The specific method for comparing the data of the foreign matters 5 before and after S25 is that the distance and the position of the foreign matter point cloud data of each frame relative to the elevator door are compared, the foreign matters 5 occupy the closed area range of the elevator door through the distance change of the foreign matter point cloud data relative to the elevator door, and the state of the foreign matters 5 is the state of entering the elevator door or exiting the elevator door through the position change of the foreign matter point cloud data.
Example III
As shown in fig. 3, when the elevator door and the foreign matter 5 are in the state a, the foreign matter 5 is not present in the elevator door closing area at this time by the elevator door pre-sensing type anti-pinch system based on three-dimensional point cloud deep learning, and the elevator door is closed at a normal speed.
When the elevator door and the foreign matter 5 are in the state b, the foreign matter 5 in the elevator door closing area is obtained through the elevator door pre-sensing type anti-clamping system based on three-dimensional point cloud deep learning, and the foreign matter 5 is judged to be in an elevator door state when the foreign matter 5 is absent, and the elevator door is opened at a normal speed.
When the elevator door and the foreign matter 5 are in the state c, the foreign matter 5 in the elevator door closing area still exists through the elevator door pre-sensing anti-clamping system based on three-dimensional point cloud deep learning, and the tail part does not enter an elevator, so that the foreign matter 5 is judged to be still in the state of entering the elevator door, and the elevator door is opened at a normal speed.
When the elevator door and the foreign matter 5 are in the state d, the foreign matter 5 tail part in the elevator door closing area at the moment is completely entering the elevator door through the elevator door pre-sensing type anti-clamping system based on three-dimensional point cloud deep learning, the foreign matter 5 is judged to be in the elevator door exiting state, the elevator door is slowly closed according to the range of the elevator door closing area occupied by the foreign matter 5, which is pre-sensed according to the distance between the foreign matter 5 and the elevator door, so that the condition that the foreign matter 5 contacts the elevator door is avoided, and the elevator door closing time is shortened.
When the elevator door and the foreign matter 5 are in the state e, the foreign matter 5 is completely out of the elevator door range at the moment, and the elevator door is closed at a normal speed by the elevator door pre-sensing type anti-clamping system based on three-dimensional point cloud deep learning.
In the description of the present invention, it should be understood that the terms "longitudinal," "transverse," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," and the like indicate or are based on the orientation or positional relationship shown in the drawings, merely to facilitate description of the present invention, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus should not be construed as limiting the present invention.
The above embodiments are merely illustrative of the preferred embodiments of the present application, and the scope of the present application is not limited thereto, but various modifications and improvements made by those skilled in the art to which the present application pertains are made without departing from the spirit of the present application, and all modifications and improvements fall within the scope of the present application as defined in the appended claims.
Claims (5)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310753474.2A CN116788960B (en) | 2023-06-25 | 2023-06-25 | Elevator door pre-sensing anti-pinch system and method based on three-dimensional point cloud deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310753474.2A CN116788960B (en) | 2023-06-25 | 2023-06-25 | Elevator door pre-sensing anti-pinch system and method based on three-dimensional point cloud deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116788960A CN116788960A (en) | 2023-09-22 |
| CN116788960B true CN116788960B (en) | 2025-09-26 |
Family
ID=88034261
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310753474.2A Active CN116788960B (en) | 2023-06-25 | 2023-06-25 | Elevator door pre-sensing anti-pinch system and method based on three-dimensional point cloud deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116788960B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109709564A (en) * | 2018-12-05 | 2019-05-03 | 交控科技股份有限公司 | A kind of shield door anti-clipping system and method based on the detection of laser radar single line |
| CN110713087A (en) * | 2019-10-21 | 2020-01-21 | 北京猎户星空科技有限公司 | Elevator door state detection method and device |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4664394B2 (en) * | 2008-05-23 | 2011-04-06 | 株式会社日立製作所 | Elevator door safety device and safety control method |
| JP2012020823A (en) * | 2010-07-13 | 2012-02-02 | Toshiba Elevator Co Ltd | Safety device of elevator door |
| DE102014113572B4 (en) * | 2014-09-19 | 2023-01-12 | Bode - Die Tür Gmbh | Door system with sensor unit for boarding aid monitoring |
| CN110526058B (en) * | 2018-05-23 | 2022-06-03 | 奥的斯电梯公司 | Elevator door monitoring system, elevator system and elevator door monitoring method |
| US11332345B2 (en) * | 2018-08-09 | 2022-05-17 | Otis Elevator Company | Elevator system with optimized door response |
| JP7189806B2 (en) * | 2019-02-28 | 2022-12-14 | 株式会社日立製作所 | Sensor unit and elevator |
| CN110821330B (en) * | 2019-11-26 | 2021-03-12 | 交控科技股份有限公司 | Method and device for preventing people from being clamped between vehicle door and shielding door gap |
| CN114863155B (en) * | 2022-04-08 | 2025-07-01 | 河海大学 | A foreign object detection method in elevator door space based on deep learning and time of flight |
| CN116040432B (en) * | 2023-03-07 | 2023-05-30 | 成都睿瞳科技有限责任公司 | Elevator image processing method, system and storage medium |
-
2023
- 2023-06-25 CN CN202310753474.2A patent/CN116788960B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109709564A (en) * | 2018-12-05 | 2019-05-03 | 交控科技股份有限公司 | A kind of shield door anti-clipping system and method based on the detection of laser radar single line |
| CN110713087A (en) * | 2019-10-21 | 2020-01-21 | 北京猎户星空科技有限公司 | Elevator door state detection method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116788960A (en) | 2023-09-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105769120B (en) | Fatigue driving detection method and device | |
| CN109657581B (en) | Urban rail transit gate traffic control method based on binocular camera behavior detection | |
| CN108520225B (en) | A Fingerprint Detection and Classification Method Based on Spatial Transform Convolutional Neural Network | |
| CN106778668B (en) | A kind of method for detecting lane lines of robust that combining RANSAC and CNN | |
| CN106006266B (en) | A kind of machine vision method for building up applied to elevator safety monitoring | |
| CN112418103A (en) | A safety anti-collision system and method for bridge crane hoisting based on dynamic binocular vision | |
| CN108629230B (en) | People counting method and device and elevator dispatching method and system | |
| CN108275524A (en) | A kind of elevator maintenance operation monitoring and guiding device based on the assessment of the first multi-view video series of operations | |
| CN113705655A (en) | Full-automatic classification method for three-dimensional point cloud and deep neural network model | |
| CN115131377B (en) | Railway turnout spacing detection method, device and system | |
| CN111461001A (en) | A computer vision automatic door opening method and system | |
| WO2022062432A1 (en) | Passenger distribution prediction model training method and system, and passenger guidance method and system | |
| CN113470073A (en) | Animal center tracking method based on deep learning | |
| CN118230265A (en) | A method for detecting foreign matter on rail transit track bed | |
| CN113763379A (en) | Method and device for detecting broken string of dropper, computer equipment and storage medium | |
| CN116788960B (en) | Elevator door pre-sensing anti-pinch system and method based on three-dimensional point cloud deep learning | |
| CN120259213A (en) | Asphalt pavement crack detection method, system, electronic device and storage medium based on image analysis | |
| CN109410250B (en) | Visual image-based vehicle arrival identification method and terminal | |
| CN118297996B (en) | A method, device and electronic device for detecting occlusion of a surveillance camera in an elevator | |
| CN118196752B (en) | Unmanned sanitation truck target detection method based on decoupling variation self-encoder network | |
| CN119474967A (en) | Railway passenger car axle box cover bolt fault detection method based on attention enhanced dual-path network | |
| CN117145346A (en) | Internet of vehicles safety door control method based on artificial intelligence | |
| Lin et al. | Boosted vehicle detection using local and global features | |
| CN115180476A (en) | A method for detecting the shape of elevator traction sheave grooves based on three-dimensional reconstruction | |
| CN112919278B (en) | Passenger elevator entering key failure fault identification method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |