Intelligent control system for battery vehicles going upstairs based on deep learning

Abstract

The purpose of the present invention is to provide a battery vehicle going upstairs control system with high recognition accuracy and accurate recognition, which can prevent battery vehicles from going upstairs without affecting the normal use of residents. The battery vehicle going upstairs intelligent control system of the present invention includes: an image acquisition module, which is arranged in the elevator, and is used to acquire real-time images of the interior of the elevator to obtain images inside the elevator; a battery vehicle detection module, which performs target detection on the images inside the elevator acquired by the image acquisition module in sequence, so as to detect whether there is a battery vehicle in the image inside the elevator; and a judgment control module, which determines whether there is a battery vehicle in the elevator according to the detection result of the battery vehicle detection module, and controls the elevator to keep the door open when it is determined that there is a battery vehicle, so as to prevent the battery vehicle from going upstairs through the elevator, wherein the battery vehicle detection module includes a target detection neural network composed of an improved Tiny‑YOLOV3 network EyevatorNet.

Description

Technical Field

[0001] The present invention relates to an intelligent management and control system for battery vehicles going upstairs based on deep learning. Background Art

[0002] Electric vehicles are easy to use and inexpensive, so they are widely used as a means of transportation.

[0003] As the use of electric bikes increases, their need to be charged all the time has gradually brought about problems in the management of residential communities. For example, for their own convenience, some residents privately carry electric bikes upstairs through elevators and charge them indoors or in public areas such as corridors. After using electric bikes for a period of time, the connection lines in the car are prone to aging and short circuits. When a short circuit occurs and the external temperature is high, it is easy to burn and cause a fire, leading to tragedy.

[0004] In order to prevent accidents, property management departments in many residential areas have set up regulations prohibiting electric bicycles from going upstairs; and in order to effectively prevent residents from carrying electric bicycles upstairs and to implement control over electric bicycles going upstairs, some electric bicycle control systems have emerged in the prior art. For example, electric bicycles are detected by laying ground sensor coils, or by traditional image recognition methods (such as manually designed features), and when electric bicycles are detected, residents with electric bicycles are prevented from going upstairs. These methods have some disadvantages, including: inaccurate detection, which results in the release of electric bicycles and failure to achieve the effect of preventing them; high false recognition rate, misidentification of people or bicycles, strollers, etc. as electric bicycles, preventing them from getting on the elevator, affecting residents' normal use; and so on. Summary of the invention

[0005] In order to solve the above problems, a battery vehicle going upstairs control system is provided, which has high recognition accuracy and can prevent battery vehicles from going upstairs without affecting the normal use of residents. The present invention adopts the following technical solutions:

[0006] The present invention provides an intelligent control system for electric vehicle going upstairs based on deep learning, which is characterized by comprising: an image acquisition module, which is arranged in the elevator, and is used to acquire real-time images of the interior of the elevator to obtain images inside the elevator; an electric vehicle detection module, which performs target detection on the images inside the elevator acquired by the image acquisition module in sequence, so as to detect whether there is an electric vehicle in the images inside the elevator; and a judgment control module, which determines whether there is an electric vehicle in the elevator according to the detection result of the electric vehicle detection module, and controls the elevator to keep the door open when it is determined that there is an electric vehicle, so as to prevent the electric vehicle from going upstairs through the elevator, wherein the electric vehicle detection module includes a target detection neural network composed of an improved Tiny-YOLOV3 network EyevatorNet.

[0007] The deep learning-based intelligent control system for electric vehicle going upstairs provided by the present invention may also include: a voice module, which is arranged in the elevator. When it is determined that there is an electric vehicle in the elevator, the judgment control module controls the voice module to play a voice prompt into the elevator while controlling the elevator to keep the door open.

[0008] The deep learning-based intelligent management and control system for electric vehicle going upstairs provided by the present invention may also include: a status monitoring module, used to detect the working status of each module and form corresponding working status information; and a communication module, used to send the working status information to the outside.

[0009] In the deep learning-based intelligent control system for battery vehicles going upstairs provided by the present invention, the image acquisition module may also include a wide-angle digital camera arranged in the elevator.

[0010] In the deep learning-based intelligent control system for battery vehicle going upstairs provided by the present invention, all parts of the intelligent control system for battery vehicle going upstairs except the wide-angle digital camera can be integrated in the control box to form a corresponding control terminal.

[0011] The deep learning-based intelligent management and control system for battery vehicle going upstairs provided by the present invention may also include: a power supply module, which is used to supply power to each module in the intelligent management and control system for battery vehicle going upstairs.

[0012] The deep learning-based intelligent management and control system for electric vehicle going upstairs provided by the present invention may also include: a remote control module, which is connected to the switch unit, and is used to cooperate with the switch unit so that the management personnel can start or shut down the intelligent management and control system for electric vehicle going upstairs through the switch unit.

[0013] Function and Effect of the Invention

[0014] According to the deep learning-based intelligent control system for battery vehicles going upstairs provided by the present invention, since the target detection neural network composed of the improved Tiny-YOLOV3 network EyevatorNet is used to detect targets in the images in the elevator, it has a stronger ability to extract image information and higher recognition accuracy, and is not prone to misjudgment. The equipment has been tested on site, and the system runs stably and the control effect is excellent. The target detection of the present invention is based on AI image recognition technology. Compared with traditional pattern recognition and buried ground sensor coil detection methods, it is more reliable and stable, and can effectively solve the safety hazards caused by the problem of battery vehicles going upstairs to charge. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a block diagram of the structure of the battery vehicle upstairs intelligent management and control system based on deep learning in an embodiment of the present invention;

[0016] Figure 22 is a schematic diagram of the structure of the Tiny-YOLOV3 network EyevatorNet according to an embodiment of the present invention;

[0017] Figure 3 It is a schematic diagram of the detection results of the Tiny-YOLOV3 network EyevatorNet according to an embodiment of the present invention;

[0018] Figure 4 It is a workflow diagram of the management and control system of an embodiment of the present invention. DETAILED DESCRIPTION

[0019] Hereinafter, specific implementations of the present invention will be described in detail with reference to the accompanying drawings and embodiments.

[0020] <Example>

[0021] Figure 1 It is a block diagram of the structure of the smart control system for battery vehicles going upstairs based on deep learning in an embodiment of the present invention.

[0022] like Figure 1 As shown, the deep learning-based intelligent control system for electric vehicle going upstairs (hereinafter referred to as the control system) 100 of the embodiment can be divided into a core part and a peripheral part, specifically including an image acquisition module 10, an electric vehicle detection module 20, a judgment control module 30, a voice module 40, a status monitoring module 50, a communication module 60, a power supply module 70 and a remote control module 80.

[0023] The image acquisition module 10 is arranged in the elevator. In this embodiment, the image acquisition module 10 includes a wide-angle digital camera, which can acquire real-time images in the elevator car (i.e., images in the elevator). The acquired real-time image data is divided into two paths, one path is transmitted to the community monitoring room (for example, transmitted to the security monitoring display in the community monitoring room for display), and the other path is used as the input of the battery vehicle detection module 20. In other words, the camera in the image acquisition module 10 of this embodiment is also used as a security monitoring camera. In other embodiments, the camera can also be obtained by modifying the line of the original camera in the elevator in the community monitoring system, and its real-time image data is separated into one path and input into the battery vehicle detection module 20.

[0024] The battery vehicle detection module 20 is used to perform target detection on the elevator image acquired by the image acquisition module 10, so as to detect whether there is a battery vehicle in the elevator image. In this embodiment, the camera acquires the real-time image in the elevator in the form of a video, and its video stream contains a plurality of real-time image frames in chronological order. The battery vehicle detection module 20 sequentially acquires these real-time image frames in chronological order and sequentially performs target detection on these real-time image frames.

[0025] In this embodiment, the electric vehicle detection module 20 includes a target detection neural network, which is specifically an improved Tiny-YOLOV3 network EyevatorNet.

[0026] Figure 2 1 is a schematic diagram of the structure of the Tiny-YOLOV3 network EyevatorNet according to an embodiment of the present invention.

[0027] like Figure 2 As shown, the Tiny-YOLOV3 network EyevatorNet adopted in this embodiment is improved based on the existing Tiny-YOLOV3 neural network model, which is specifically composed of the following layer structure: 416*416*3 input layer Input, multiple convolutional layers (Conv / Bn / Lk*n, n=1 or 7), multiple maximum pooling layers (Maxpooling), 26*26*256 feature layer X1, 13*13*256 feature layer X2, upsampling layer UpSampling2D, 26*26*128 feature layer X2, fusion layer Concatenate[X1,X2] and output Y1, output Y2.

[0028] The EyevatorNet of this embodiment is constructed and trained using an image training set involving electric vehicles. After the real-time image frames are uniformly converted into grayscale images for input, the detection result of whether the real-time image contains an electric vehicle can be obtained. In addition, the hardware form of the EyevatorNet is implemented using Jetson-nano.

[0029] Figure 3 It is a schematic diagram of the detection results of the Tiny-YOLOV3 network EyevatorNet according to an embodiment of the present invention.

[0030] like Figure 3 As shown, the Tiny-YOLOV3 network EyevatorNet can detect targets in real-time images in the elevator and detect the battery car (part in the frame). In addition, in addition to using the YOLOV3 dimensional clustering method for all targets, the target detection template frame in this embodiment also performs dimensional clustering on each detected category, and averages the two groups of anchors obtained as the final template frame. The template frame obtained by the above strategy has been significantly improved in recognition accuracy after field testing. Not only that, the target detection network of this embodiment adds image enhancement strategies (illumination, angle, stretching transformation, etc.) to the training mechanism, which can make the model more robust and effectively reduce the dependence on the environment under the same recognition accuracy.

[0031] The judgment control module 30 is used to judge whether there is a battery car in the current elevator according to the detection result of the battery car detection module 20, and perform corresponding control according to the judgment result, and the control includes: when it is judged that there is a battery car, the elevator door is controlled not to close (i.e., to remain open). Since the elevator door is always open at this time, the elevator car cannot rise, that is, the purpose of preventing the battery car from going upstairs through the elevator is achieved.

[0032] In this embodiment, the process by which the control module 30 determines whether there is an electric vehicle in the current elevator is as follows: when the electric vehicle detection module 20 detects that a real-time image frame contains an electric vehicle, it further determines whether more than three real-time image frames with an electric vehicle are detected within the next second. If more than three real-time image frames are detected, it is determined that there is an electric vehicle in the current elevator.

[0033] In addition, the hardware form of the judgment control module 30 of this embodiment is implemented by a main control board, and its control over the opening of the elevator door can be achieved by connecting to the elevator control system and sending an elevator door opening signal to it.

[0034] In this embodiment, the voice module 40 includes a speaker electrically connected to the main control board. After the judgment control module 30 determines that there is a battery car in the current elevator, it controls the elevator door not to close and also controls the voice module 40 to play a voice alarm to remind residents with battery cars, for example, playing "to ensure fire safety, battery cars are prohibited from going upstairs".

[0035] The status monitoring module 50 is used to detect the working status of each module in the management and control system 100 and form corresponding working status information.

[0036] The communication module 60 is a 4G cellular network communication module, which is used to send the working status information to the outside, for example, to the background management system responsible for maintaining the control system 100, so as to notify the maintenance personnel to repair it in time when a fault occurs.

[0037] The power module 70 is used to supply power to each module in the management and control system 100 to ensure the normal operation of the management and control system 100.

[0038] The remote control module 80 is connected to a switch unit (not shown) provided in the community monitoring room, and is used to allow the community manager to start or shut down the entire control system 100 through the switch unit (i.e., connect or disconnect the power module 70, so as to start or shut down the entire control system 100). The remote control module 80 can also be connected to the monitoring system in the community monitoring room, so that the community manager can control the start or shut down of the control system 100 through the monitoring system terminal PC.

[0039] In this embodiment, the battery vehicle detection module 20, the judgment control module 30, the voice module 40, the state monitoring module 50, the communication module 60, the power module 70 and the remote control module 80 are all integrated in a control box to form a control terminal. The control box housing can retain the network port, USB3.0 interface, HDMI interface, etc., so that the real-time image of the camera can be transmitted to the community monitoring room through the local area network. At the same time, since the hardware of the control system 100 is integrated except for the camera of the image acquisition module 10, it is small in size and light in weight (the actual manufactured control terminal has a size of 145mm*85mm*65mm and weighs only 0.82Kg), so it is easy to install, repair or replace. When there is a surveillance camera in the elevator, the camera can be connected after the control terminal is installed, and the layout of the entire control system 100 does not require complicated operations and is easy to implement. In addition, even if there is no camera in the elevator, the control terminal of this embodiment only needs to install and connect an external camera to work.

[0040] The following describes the working process of the management and control system 100 of this embodiment with reference to the accompanying drawings.

[0041] Figure 4 It is a workflow diagram of the management and control system of an embodiment of the present invention.

[0042] like Figure 4 As shown, after the management and control system 100 is powered on and initialized, its workflow includes the following steps:

[0043] Step S1, the state monitoring module 50 determines whether the image acquisition module 10 has an image input, if there is no image input, an abnormal state prompt is generated and sent by the communication module 60, if there is an image input, the process proceeds to step S2;

[0044] Step S2, the battery vehicle detection module 20 continuously detects the battery vehicle on the real-time image acquired by the image acquisition module 10, and if a battery vehicle is detected, the process proceeds to step S3;

[0045] Step S3, the control module 30 determines whether more than three real-time image frames with a battery car are detected within the next 1 second. If it is detected again, it is determined that there is a battery car in the elevator, and then the process goes to step S4;

[0046] Step S4, the judgment control module 30 controls the voice module 40 to issue a voice alarm, and at the same time controls the elevator not to close the door.

[0047] The above process is repeated in real time until the entire control system 100 is shut down through the remote control module 80.

[0048] Example Function and Effect

[0049] According to the control system provided in this embodiment, since the target detection neural network composed of the improved Tiny-YOLOV3 network EyevatorNet is used to detect targets in the elevator, it has a stronger ability to extract image information and higher recognition accuracy, and is not prone to misjudgment. The equipment has been tested on site, and the system runs stably and the control effect is excellent. The target detection of this embodiment is based on AI image recognition technology, which is more reliable and stable than traditional detection methods such as pattern recognition and buried ground sensor coils, and can effectively solve the safety hazards caused by the problem of charging electric vehicles upstairs.

[0050] In addition, compared with the original Tiny-YOLOV3 network structure, the improved EyevatorNet of this embodiment has a stronger ability to extract image information and can greatly improve the recognition accuracy: its training network strategy adds an image enhancement strategy, and can show good recognition effects without additional fill light under low light conditions; the detection pre-selection box is obtained by clustering each type of detection object dimension and averaging the overall clustering, which has better recognition effects than the pre-selection box obtained only for the overall clustering. Field tests also show that the EyevatorNet of this embodiment has better recognition effects.

[0051] In the embodiment, when the battery vehicle detection module detects that a real-time image frame contains a battery vehicle, the judgment control module further judges whether more than three real-time image frames with the battery vehicle are detected within the next second. If more than three real-time image frames are detected, it is judged that there is an battery vehicle in the current elevator. This judgment method can further effectively reduce misjudgment.

[0052] In this embodiment, all components except the camera are integrated in the control box, so it has a small size, light weight, high integration, compact design, and convenient and quick installation.

[0053] In this embodiment, the real-time image data obtained by the camera is not only used as the input of the battery vehicle detection module, but also separated as the input of the community monitoring room. Therefore, the management and control system of this embodiment also has the effect of one machine with multiple uses.

[0054] The above embodiments are only used to illustrate different implementation forms of the present invention, and the protection scope of the present invention is not limited to the description scope of the above embodiments.

Claims

1. A deep learning-based intelligent control system for battery vehicles going upstairs, characterized in that: include: An image acquisition module is provided in the elevator and is used to acquire real-time images of the interior of the elevator to obtain images inside the elevator; A battery car detection module performs target detection on the elevator interior image acquired by the image acquisition module in sequence, thereby detecting whether there is a battery car in the elevator interior image; as well as a judgment control module, which judges whether there is a battery car in the elevator according to the detection result of the battery car detection module, and controls the elevator to keep the door open when it is determined that there is a battery car, so as to prevent the battery car from going upstairs through the elevator, The battery vehicle detection module includes a target detection neural network composed of an improved Tiny-YOLOV3 network EyevatorNet. In addition to using the YOLOV3 dimensional clustering method for all targets, the target detection template box also performs dimensional clustering for each detected category, and takes the average of the two groups of anchors obtained as the final template box.

2. According to the deep learning-based intelligent control system for battery vehicles going upstairs in claim 1, it is characterized in that: Also includes: A voice module is arranged in the elevator. When it is determined that there is a battery car in the elevator, the judgment control module controls the elevator to keep the door open and also controls the voice module to play a voice prompt into the elevator.

3. According to the deep learning-based intelligent control system for battery vehicles going upstairs in claim 1, it is characterized in that: Also includes: Status monitoring module, used to detect the working status of each module and form corresponding working status information; as well as The communication module is used to send the working status information externally.

4. According to the deep learning-based intelligent control system for battery vehicles going upstairs in claim 1, it is characterized by: in, The image acquisition module includes a wide-angle digital camera arranged in the elevator.

5. According to claim 4, the deep learning-based intelligent control system for battery vehicles going upstairs is characterized by: in, All parts of the battery vehicle stair-going intelligent control system except the wide-angle digital camera are integrated into the control box to form a corresponding control terminal.

6. The deep learning-based intelligent control system for battery vehicles going upstairs according to claim 1 is characterized in that: Also includes: The power supply module is used to supply power to each module in the electric vehicle upstairs intelligent management and control system.

7. The deep learning-based intelligent control system for battery vehicles going upstairs according to claim 6 is characterized in that: Also includes: The remote control module is connected to the switch unit and is used to cooperate with the switch unit so that the management personnel can start or shut down the battery vehicle upstairs intelligent management and control system through the switch unit.

Images