Elevator door intelligent control method based on feature recognition

Abstract

The invention relates to an elevator door intelligent control method based on feature recognition, which comprises the following steps: S1, opening an elevator door, and collecting image information in an elevator door area; s2, image feature information in the image information is extracted, and age analysis is conducted on the elevator passengers according to the image feature information; and S3, if the age exceeds the age threshold value, the door closing time of the elevator door is prolonged. According to the elevator door intelligent control method based on feature recognition, a visual recognition method for old people and disabled people is designed according to the elevator access personnel scene, and face-based age recognition, human body shape-based disabled people recognition and various recognition evaluation methods based on specific carried articles (such as crutches and wheelchairs) can be achieved at the same time; the multi-angle evaluation enables the evaluation result to be more accurate, if the old or the disabled are recognized, the door closing time of the elevator door is intelligently adjusted, the elevator door closing waiting time is intervened, and the potential safety hazards when the old or the disabled move slowly go in and out of the elevator are eliminated.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to an elevator door intelligent control method based on feature recognition. Background Art

[0002] The renovation of old residential areas is a major livelihood project to improve people's well-being. However, the waiting time for the door to close in existing elevators is mostly 3.2 seconds. For the elderly, disabled people and other people with limited mobility entering and exiting the elevator, it is easy to be caught in the door or chase the elevator. Therefore, in order to solve the safety hazard of the elderly, disabled people or people with limited mobility entering and exiting the elevator, it is proposed to use a visual recognition method to intelligently control the elevator door closing time. Summary of the invention

[0003] The embodiment of the present invention provides an intelligent control method for elevator doors based on feature recognition to solve the potential safety hazard problem caused by the excessively single waiting time for elevator door closing in the prior art.

[0004] In order to achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0005] An elevator door intelligent control method based on feature recognition comprises the following steps:

[0006] S1, the elevator door opens and the image information in the elevator door area is collected;

[0007] S2, extracting image feature information from the image information, and performing age analysis on the passengers according to the image feature information;

[0008] S3. If the age exceeds the age threshold, the closing time of the elevator door is extended.

[0009] Furthermore, before S2, the training method for the face detection model includes: preprocessing human face images of different ages, then using a labeling tool to label age data for the faces in the images, extracting corresponding age features from the labeled human face images, mapping the age data and the age features, and then making the labeled images into training data sets and test data sets in a certain proportion, inputting the training data sets into the face detection model for training, and inputting the test data sets for testing after the training is completed. If the recognition rate of the test reaches the preset requirement, the training model is exported, the training is completed, and the face recognition model is obtained.

[0010] Furthermore, the preprocessing includes at least one of enhancement processing such as random erasing, random blurring, and super-resolution on the face image.

[0011] Furthermore, in S3, the face recognition model extracts feature information from the image information and performs comparative analysis, and performs evaluation based on the number and degree of age features in the image information. If the age value output by the face recognition model is lower than the age threshold, the automatic door closing interval of the elevator is not modified; conversely, if the age value output by the face recognition model is higher than the age threshold, the automatic door closing interval of the elevator is extended.

[0012] Furthermore, if facial information cannot be collected from the image information, body recognition or specific object recognition is performed. If the output result of any one of facial recognition, body recognition or specific object recognition meets the intervention condition, the elevator door closing waiting time is extended.

[0013] The embodiments of the present invention have the following advantages:

[0014] The intelligent control method for elevator doors based on feature recognition of the present invention is designed for the scene of people entering and exiting the elevator, and a visual recognition method for the elderly and disabled is designed. It can simultaneously realize age recognition based on face, disabled person recognition based on human body shape, and multiple recognition and evaluation methods based on specific carried items (such as crutches, wheelchairs, etc.). Multi-angle evaluation makes the evaluation result more accurate. If the elderly or disabled are identified, the closing time of the elevator door is intelligently adjusted, and the waiting time for the elevator door closing is intervened, which is used to solve the safety hazards of slow-moving elderly or disabled people when entering and exiting the elevator. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the implementation methods of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the implementation methods or the description of the prior art. Obviously, the drawings in the following description are only exemplary, and for ordinary technicians in this field, other implementation drawings can be derived from the provided drawings without creative work.

[0016] The structures, proportions, sizes, etc. illustrated in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with the technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantial technical significance. Any structural modification, change in proportion or adjustment of size shall still fall within the scope of the technical contents disclosed in the present invention without affecting the effects and purposes that can be achieved by the present invention.

[0017] Figure 1 A method flow chart of an elevator door intelligent control method based on feature recognition provided in Example 1 of the present invention;

[0018] Figure 2 A method detail diagram of an elevator door intelligent control method based on feature recognition provided in Example 1 of the present invention;

[0019] Figure 3 This is a system structure diagram of an elevator door intelligent control method based on feature recognition provided in Example 2 of the present invention. DETAILED DESCRIPTION

[0020] The following is a description of the implementation of the present invention by specific embodiments. People familiar with the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] An intelligent control method for elevator doors based on feature recognition. This technology mainly solves the problem of people being pinched by elevator doors due to the slowness of elderly people, disabled people and other people with limited mobility getting on the elevator. It can also solve the problem of people with limited mobility not getting on the elevator due to malicious closing of the elevator door in advance. The technical details and technical effects of this technology are specifically introduced through multiple embodiments below.

[0022] Example 1

[0023] like Figure 1 As shown, the following steps are included:

[0024] S1, the elevator door opens and the image information in the elevator door area is collected;

[0025] S2, extracting image feature information from the image information, and performing age analysis on the passengers according to the image feature information;

[0026] S3. If the age exceeds the age threshold, the closing time of the elevator door is extended.

[0027] In step S1, this technology can use the existing camera in the elevator to collect images, which are used to collect image information in the elevator door area after the elevator door is opened, and is used to identify the age or physical condition of the person entering the elevator. The situation of entering the elevator is very complicated, and there may be multiple people or a single person. When multiple people enter the elevator, there will generally be someone who will spontaneously help the person with limited mobility, such as helping to press the door opening button and waiting for the person with limited mobility to slowly enter. Therefore, this technology discusses more about the situation of a single person entering and exiting the elevator without changing his or her movements. For example, if multiple people enter the elevator, the image information of the first person entering the elevator can be identified. If the first person entering the elevator does not meet the intervention conditions, a new identification procedure will not be performed until the elevator door is closed and then opened; or if more than one person's face information is identified between the opening and closing of the elevator door, a new identification procedure will not be performed until the elevator door is closed and then opened.

[0028] In step S2, the image feature information in the image information is extracted, and the age of the passenger is analyzed based on the image feature information. The age analysis of this technology is mainly based on face recognition. Face recognition uses a machine learning algorithm. By training a large amount of face image data, it learns the complex relationship between age and various face features, and then predicts age based on the input face image feature information. Before face recognition, the face detection model is trained first. The specific training method includes the following steps:

[0029] Preprocessing is performed on facial images of different ages, and the preprocessing includes at least one of random erasing, random blurring, super-resolution and other enhancement processing on the facial images, and it is also necessary to eliminate images with unclear targets and unclear images. Random erasing refers to randomly erasing a certain position in the training image, so that the age of the included faces can be accurately recognized in the subsequent case of occlusion. Random blurring refers to blurring the training image, so that the age of the included faces can be accurately recognized in the subsequent case of blurred images. Super-resolution processing refers to super-resolution processing of the training image to obtain a relatively clear image, which is then used as a training image for training, that is, as an aid to improve the generalization of the age recognition model.

[0030] Then use the annotation tool to annotate the age data of the face in the image, extract the corresponding age features from the annotated face image, map the age data to the age features, and then make the annotated image into a training data set and a test data set in a certain proportion, input the training data set into the face detection model for training, and input the test data set for testing after the training. If the recognition rate of the test meets the preset requirements, the training model is exported, the training is completed, and the face recognition model is obtained; if the recognition rate does not meet the preset requirements, the next step is executed, wherein the preset requirements of this technology are preferably 99% recognition rate. Specifically, it is preferred to use the MTCNN face recognition model, the training set of which includes IMDB, WIKI, MOPRTH2, UTKFace, MegaAge-Asian, fer2013, Expw and other data sets, of which the age estimation is about 16w.

[0031] When the elevator door opens, the image information in the video information collected by the camera in the elevator is extracted at a fixed frequency, and the image information is pre-processed and input into the face recognition model. The face recognition model extracts the feature information in the image information and compares and analyzes it. It evaluates according to the number and degree of age features in the image information. For example, if the image information has crow's feet features and the proportion of white hair pixels exceeds 50%, it means that the age of the elevator passenger is over 60 years old. It is further evaluated in combination with other image features, and finally outputs the age of the person entering the elevator. If the age value output by the face recognition model is lower than the age threshold, the elevator passenger is identified as "young", and the automatic door closing interval of the elevator is not modified. The automatic door closing interval of a conventional elevator is 3.2 seconds; on the contrary, if the age value output by the face recognition model is higher than the age threshold, the elevator passenger is identified as "old", and the automatic door closing interval of the elevator is extended, such as to 6.4 seconds.

[0032] like Figure 2 As shown, if the face information cannot be collected from the image information, it means that the face of the person currently entering or exiting the elevator is blocked, such as wearing a scarf or a mask, and other image information is needed for identification. This technology uses body recognition or specific object recognition to assist face recognition, or uses face recognition, body recognition, and specific object recognition to perform multi-angle recognition to make the recognition result more accurate.

[0033] a. Body posture recognition

[0034] This technology adopts the invention patent technology of patent number 202310150844.3, patent name: Human body image age estimation method, system, medium and device, to identify the body shape of people entering and exiting the elevator. If the age value output by the body shape recognition model is lower than the age threshold, the passenger is identified as "young", and the automatic door closing interval of the elevator is not modified. The automatic door closing interval of a conventional elevator is 3.2 seconds; conversely, if the age value output by the body shape recognition model is higher than the age threshold, the passenger's age is identified as "old", and the automatic door closing interval of the elevator is extended, such as to 6.4 seconds.

[0035] In posture recognition, this technology also has a portrait magnification rate recognition. When the elevator door opens, the system determines that someone has entered the elevator, and then collects image information in the video stream at a fixed frequency, and calculates the number of pixels in the same part of the image information. For example, the image information of the head is extracted, and the rate of increase of the number of pixels in this part in the image stream is calculated. The magnification rate is calculated by dividing the pixel change in the same part between adjacent images or images separated by a fixed time by time. If the rate of change of the number of pixels is higher than the set rate, it means that the passenger is moving slowly and has a problem of slow movement, and the waiting time for the elevator door to close needs to be extended; conversely, if the rate of change of the number of pixels is higher than the set rate, it means that the passenger's movement is normal and there is no need to intervene in the control of the elevator.

[0036] b. Specific object identification

[0037] For the recognition of specific objects such as crutches and wheelchairs, this technology uses patent number 202110196030.4, and the patent name is the invention patent technology of elevator recognition method, device and elevator, to realize the recognition of target objects in the elevator. The target objects in this technology are crutches and wheelchairs. If the specific object recognition model recognizes the presence of specific objects in the image information, it means that the elevator passenger is moving slowly and has a problem of slow movement, and the waiting time for the elevator door to close needs to be extended; conversely, if the specific object recognition model does not recognize the presence of specific objects in the image information, it means that the elevator passenger's movement is normal and there is no need to intervene in the control of the elevator.

[0038] Since age recognition alone cannot accurately assess the movement of passengers, body recognition and specific object recognition are combined to comprehensively analyze the passengers' actions. If the output result of any of face recognition, body recognition or specific object recognition meets the intervention conditions, such as the output is "the elevator door closing waiting time needs to be extended", the elevator door closing waiting time is extended. For example, the system outputs a 3-second high-level pulse to the elevator door controller, informing the elevator door switch controller to extend the elevator door closing waiting time.

[0039] Example 2

[0040] like Figure 3 As shown, in order to control both entry and exit of the elevator, cameras are set on the front and rear sides of the elevator, namely the front camera and the rear camera, wherein the front camera is set on the elevator door side, facing the elevator, and is used to collect image information of people leaving the elevator. In this technology, it is preferably set on the upper part of the elevator operating panel. The rear camera can be installed in a conventional camera position in the elevator, facing the elevator door, and is used to collect image information of people entering the elevator. When the elevator door is opened, the front camera will collect image information of people in the elevator, and the system will perform the aforementioned feature extraction and analysis. If the person leaving the elevator is a person with mobility difficulties, the closing time of the elevator door will be extended.

[0041] The function of setting the front camera includes preventing people in the elevator from maliciously closing the elevator in advance, which may cause the elevator to trap people. The specific implementation process is that when the elevator door is opened, the rear camera collects the video stream of the passengers, the image recognition module extracts the image information in the video stream, and the face recognition submodule, the considerate recognition submodule, and the specific object recognition submodule in the image recognition module analyze the image information respectively. If the recognition result of any one of the face recognition submodule, the considerate recognition submodule, and the specific object recognition submodule meets the intervention condition, the waiting time for the elevator door to close is extended. Since the elderly are not all people with limited mobility, a front camera is set. If a closing signal is received from the elevator keyboard, the image recognition module collects the image information in the video stream of the front camera. If the person in the image information of the front camera and the rear camera is the same person, it means that the current person has entered the elevator, and the elevator controller receives the closing signal to close the elevator; if the portrait information collected by the front camera and the portrait information collected by the rear camera are not the portrait information of the same person, the elevator control system ignores the closing signal and follows the delayed door closing waiting time, which can prevent someone from maliciously closing the door in advance.

[0042] Although the present invention has been described in detail above by general description and specific embodiments, it is obvious to those skilled in the art that some modifications or improvements can be made to the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention all belong to the scope of protection claimed by the present invention.

Claims

1. An elevator door intelligent control method based on feature recognition, characterized in that: The following steps are involved: S1, the elevator door opens and the image information in the elevator door area is collected; S2, extracting image feature information from the image information, and performing age analysis on the passengers according to the image feature information; S3. If the age exceeds the age threshold, the closing time of the elevator door is extended.

2. The elevator door intelligent control method based on feature recognition according to claim 1 is characterized in that: Before S2, the training method for the face detection model includes: Preprocess the facial images of people of different ages, then use the annotation tool to annotate the age data of the faces in the images, extract the corresponding age features from the annotated facial images, map the age data to the age features, and then make the annotated images into training data sets and test data sets in a certain proportion, input the training data set into the face detection model for training, and input the test data set for testing after the training. If the recognition rate of the test reaches the preset requirement, export the training model, complete the training, and obtain the face recognition model.

3. The elevator door intelligent control method based on feature recognition according to claim 2 is characterized in that: The preprocessing includes at least one of random erasing, random blurring, super-resolution and other enhancement processing on the human face image.

4. The elevator door intelligent control method based on feature recognition according to claim 1 is characterized in that: In S3, the face recognition model extracts feature information from the image information and performs comparative analysis, and evaluates the number and degree of age features in the image information. If the age value output by the face recognition model is lower than the age threshold, the automatic door closing interval of the elevator is not modified; conversely, if the age value output by the face recognition model is higher than the age threshold, the automatic door closing interval of the elevator is extended.

5. The elevator door intelligent control method based on feature recognition according to claim 1 is characterized in that: If facial information cannot be collected from the image information, body recognition or specific object recognition is performed. If the output result of any one of facial recognition, body recognition or specific object recognition meets the intervention condition, the elevator door closing waiting time is extended.

Images