阅读说明：本技术 一种电梯防夹手方法 (method for preventing hands of elevator from being clamped ) 是由 邓道举 钟亚林 曹学升 李京乐 杨莉 许小康 徐越翰 赵雷杰 杨嘉炀 寿梦娜 孔 于 2019-10-17 设计创作，主要内容包括：本发明公开了一种电梯防夹手方法,包括以下步骤：S1、摄像头获取视频图像；S2、对视频图像进行预处理,得到电梯门目标区域图像；S3、将目标区域图像输入防夹手模型,经过学习,得到有手或无手的预测结果；S4、根据预测结果,进行相应动作。本申请通过拍摄视频图像,依据防夹手模型,对电梯门边的图像进行有手或无手的预测,实现了对电梯门边细小物体存在时的判断,降低事故发生的机率。(The invention discloses a elevator hand-clamping prevention method which comprises the following steps of S1, acquiring a video image by a camera, S2, preprocessing the video image to obtain an elevator target area image, S3, inputting the target area image into a hand-clamping prevention model, obtaining a prediction result of hands or no hands through learning, and S4, performing corresponding action according to the prediction result.)

1, method for preventing hands of elevator from being clamped, which is characterized by comprising the following steps:

s1, acquiring a video image by the camera;

s2, preprocessing the video image to obtain an image of a target area of the elevator ;

s3, inputting the target area image into the anti-pinch model, and obtaining a prediction result with or without hands through learning;

and S4, performing corresponding actions according to the prediction result.

2. The elevator hand-pinch prevention method according to claim 1, characterized in that: in step S1, the video image is decimated to obtain a color image.

3. The method of claim 2, wherein the color image is clipped in step S2 to obtain an image of the target area within the predetermined range of the elevator .

4. The elevator hand-pinch prevention method according to claim 1, characterized in that: in step S3, the establishment of the anti-pinch model includes the following steps:

a1, collecting image data samples of the elevator ;

a2, cleaning and preprocessing the image data sample to obtain an image sample of a target area of the elevator ;

a3, inputting the target area image sample into a neural network for training;

and A4, obtaining the learned hand-clamping prevention model after finishing training.

5. The method of claim 4, wherein in step A1, the camera is installed right above the elevator , and the camera takes pictures of the elevator in a certain area to obtain a plurality of image data samples of the elevator in a certain range, with or without hands in each case.

6. The elevator hand-pinch prevention method according to claim 2, characterized in that: in step a2, the image data samples obtained in step a1 are cleaned to obtain effective image data samples, and the effective image data samples are cut to obtain target area image samples.

7. The elevator hand-pinch prevention method according to claim 4, characterized in that: in step a3, the target area image sample is input to a neural network for training, to obtain the characteristics of the target area when the hand is present or absent, and to obtain the prediction capability of the image.

8. The elevator hand-pinch prevention method according to claim 1, characterized in that: in step a4, after training is completed, an anti-pinch model capable of predicting presence or absence of a hand is obtained.

9. The elevator hand-pinch prevention method according to claim 1, characterized in that: in step S4, if it is predicted that a hand approaches, a voice alarm is presented, and if it is predicted that no hand is present, no alarm is given.

10. The elevator hand-pinch prevention method according to claim 1, characterized in that: the anti-pinch model is arranged in an elevator control system after being trained in advance.

Technical Field

The invention relates to the technical field of elevator control, in particular to an elevator anti-pinch method.

Background

At present, the obstacle at the elevator is detected by adopting an infrared light curtain mode, an infrared transmitting tube and an infrared receiving tube are arranged at the edge of the elevator at equal intervals, in the process of a switch , if the infrared receiving tube cannot receive an infrared signal, the existence of an obstacle is indicated, if the obstacle meets a transparent object, the obstacle cannot be detected, and compared with a smaller object, such as a dog rope, a small hand of a child or a part of a big hand, the obstacle cannot be detected with high probability, the detection effect is poor, the phenomenon that the small dog is outside, the owner enters the elevator, the elevator is closed can occur, and thus a safety accident occurs, if the part of the small hand of the child or the big hand is positioned at the edge of the elevator , and the elevator cannot be detected, the hand clamping accident can occur when the elevator is closed.

Furthermore, the light curtain has a length equal to of the height of the elevator , which makes the installation process troublesome and the installation and transportation efficiency low.

Therefore, how to improve the detection accuracy of the elevator and prevent accidents is a problem to be solved.

Disclosure of Invention

The invention aims to provide elevator anti-pinch methods, which adopt a camera to detect the state of an elevator , process video images shot by the camera to obtain images of a target area of the elevator , input the images into an anti-pinch model to train, and obtain a prediction result of whether a hand is at the elevator , so that corresponding actions are performed, accidents are prevented, and the safety of obstacles is protected.

The above object of the present invention is achieved by the following technical solutions:

A method for preventing hands of an elevator, comprising the following steps:

s1, acquiring a video image by the camera;

s2, preprocessing the video image to obtain an image of a target area of the elevator ;

s3, inputting the target area image into the anti-pinch model, and obtaining a prediction result with or without hands through learning;

and S4, performing corresponding actions according to the prediction result.

The present invention further provides that in step S1, the video image is decimated to obtain a color image.

The present invention further provides for cropping the color image to obtain an image of the target area within the fixed range of elevator in step S2.

The invention is further that the establishment of the anti-pinch model in step S3 includes the following steps:

a1, collecting image data samples of the elevator ;

a2, cleaning and preprocessing the image data sample to obtain an image sample of a target area of the elevator ;

a3, inputting the target area image sample into a neural network for training;

and A4, obtaining the learned hand-clamping prevention model after finishing training.

The invention further includes that in step A1, a camera is installed right above the elevator , the camera takes pictures of the elevator in a certain area, and a large number of image data samples of the elevator in a certain range, with or without hands in each case, are obtained.

The method further includes the steps of cleaning the image data sample obtained in step A1 to obtain an effective image data sample, and cropping the effective image data sample to obtain a target area image sample in step A2.

Step , in step A3, the target area image sample is input into a neural network for training, so as to obtain the characteristics of judging whether a hand exists or not in the target area, and obtain the prediction capability of the image.

The method is further that the method comprises step A4, after training, obtaining the hand-clamping prevention model capable of predicting whether the hands are present or not.

The present invention further includes the steps of performing a voice alarm prompt if a hand is predicted to approach the alarm in step S4, and not performing an alarm if no hand is predicted.

The present invention proceeds to where the anti-pinch model is pre-trained and deployed in the elevator control system.

Compared with the prior art, the invention has the beneficial technical effects that:

1. the camera is adopted to shoot the elevator , all obstacles on the elevator can be detected, and the small objects are prevented from being missed to be detected;

2., according to the method, a hand-clamping prevention model is established by a large number of images of the elevator with or without hands and adopting a neural network, so that a basis is provided for extracting the state of the elevator ;

3., the video images shot by the camera are captured to obtain images of a target area of the elevator , the images of the target area are input into an anti-pinch model, the images of the elevator are predicted, judgment of the elevator when small objects exist is achieved, and the probability of accidents is reduced.

Drawings

FIG. 1 is a flow chart illustrating the anti-pinch control of embodiments of the present invention;

FIG. 2 is a flow chart of exemplary embodiments of the present invention for establishing an anti-pinch model.

Detailed Description

The present invention is further described in detail below with reference to the attached figures.

The invention discloses an elevator anti-pinch method, which comprises the following steps as shown in figure 1:

s1, acquiring a video image by the camera;

s2, preprocessing the video image to obtain an image of a target area of the elevator ;

s3, inputting the target area image into the anti-pinch model, and obtaining a prediction result with or without hands through learning;

and S4, performing corresponding actions according to the prediction result.

The camera is arranged at the side of the elevator , video shooting is carried out on the state at the side of the elevator , and the state at the side of the elevator is recorded in real time.

In step S1, frames are extracted from the video image, and frame rate images at regular intervals of in the video image are obtained, so as to obtain an RGB three-channel color image.

In step S2, the color image is cleaned and cut to obtain an image of the target area within a predetermined range of the elevator .

The color image is cleaned, namely a large number of repeated images, overexposed or too dark images and blurred images are deleted, and clear and non-repeated images are obtained.

The fixed range of the front of the elevator is set, and the cleaned image is cut to obtain a target area image.

In the embodiments of the present application, the color image is cut to obtain a color image of a small area, and then the color image of the small area is cleaned, so that the data processing amount can be reduced and the processing speed can be increased.

And after the target area image is obtained, preprocessing is carried out, including data increasing operations such as translation, rotation, noise adding and the like.

In step S3, the target area image that is in the past is input to the anti-pinch model for learning, and after learning is completed, a prediction result of whether a hand is present or absent in the target area is obtained.

In step S4, if it is predicted that a hand approaches, a voice alarm is given, the elevator is stopped and the elevator is opened in the reverse direction, and the occurrence of a hand-pinching accident is prevented.

If hands are not predicted, no alarm is given and the elevator continues to be closed.

The models of establishing anti-pinch, as shown in FIG. 2, include the following steps:

a1, collecting image data samples of the elevator ;

a2, cleaning and preprocessing the image data sample to obtain an image sample of a target area of the elevator ;

a3, inputting the target area image sample into a neural network for training;

and A4, obtaining the learned hand-clamping prevention model after finishing training.

In specific embodiments of the present application, a camera is installed directly above elevator , the camera takes video of elevator in a certain area, obtains a real-time image of elevator , frames the real-time image, and obtains a large number of image data samples of elevator and in a certain range, and elevator has hands or no hands in each case.

And cleaning the image data sample to obtain an effective image data sample, and cutting the effective image data sample to obtain a target area image sample.

And inputting the target area image sample into a neural network for training to obtain the characteristics of judging whether a hand exists or not in the target area, and obtaining the prediction capability of the image.

After training is finished, an anti-pinch model capable of predicting hands and hands-free is obtained.

And deploying the trained anti-pinch model in an elevator control system. For example, in mobile equipment, the prediction judgment of the presence or absence of hands in a real-time image is realized.

In specific embodiments of the present application, the resolution of the image is 640 x 480.

The embodiments of the present invention are preferred embodiments of the present invention, and the scope of the present invention is not limited by these embodiments, so: all equivalent changes made according to the structure, shape and principle of the invention are covered by the protection scope of the invention.