System for preventing electric vehicle from entering unit door

Abstract

The invention discloses a system for preventing an electric vehicle from entering a cell gate, and relates to the technical field of electric vehicle access control management and control. Obtaining three views of the vehicle; performing three-dimensional modeling to obtain a to-be-judged model, and distinguishing the to-be-judged model based on the convolutional neural network; controlling the alarm device to give an alarm; the method is used for solving the problems that in the prior art, a traditional access control mode is generally adopted for controlling personnel or an electric vehicle to enter a unit door, or a single camera is adopted for automatic recognition during recognition, and although recognition can be conducted to a certain degree, the recognition cannot be conducted; if an object similar to the electric vehicle appears or the object is extremely similar to the electric vehicle in an image shot at an angle where a camera is located, the problem of false alarm caused by inaccurate identification can occur.

Description

Technical Field

[0001] The present invention relates to the technical field of electric vehicle access control, in particular to a system for preventing electric vehicles from entering a unit door. Background Art

[0002] Methods for preventing electric vehicles from entering unit doors include setting up physical barriers, using intelligent monitoring systems, and strengthening community management. Specifically, physical barriers are set up to prevent electric vehicles from entering unit doors, and warnings are issued before electric vehicles enter unit doors through a combination of technical means and manual supervision. In existing technologies, electric vehicles entering unit doors are controlled by setting up access control at unit doors. Access control can use electronic license plate recognition or camera recognition to identify when electric vehicles enter unit doors, and issue warnings or close unit doors to prevent electric vehicles from entering.

[0003] In the existing method for preventing electric vehicles from entering the unit door, in the technology using the intelligent monitoring system, an electronic tag or a Bluetooth signal transmission module is usually installed in the electric vehicle in the community. When the electric vehicle enters the unit door or elevator, the entry status of the electric vehicle is judged by scanning the electronic tag, thereby issuing an early warning; this method can only be used to prevent electric vehicles installed with electronic tags or Bluetooth signal transmission modules, and has limitations. It is impossible to effectively control electric vehicles outside the community. For example, in the Chinese patent with the publication number CN117351630A, a system for prohibiting electric vehicles from entering buildings and elevators is disclosed. This scheme is to set both the electronic tag and the Bluetooth signal transmission module on the electric bicycle, and to detect whether there is an identifiable electronic tag in the unit door area to determine whether there is an electric vehicle in the unit door area. Other improvements for preventing electric vehicles from entering the unit door usually use a single camera for automatic recognition. Although this method can perform a certain degree of recognition, if an object similar to the electric vehicle appears or the object is extremely similar to the electric vehicle in the image taken at the angle of the camera, there will be a problem of false alarm due to inaccurate recognition. In view of this, it is necessary to improve the existing technology using the intelligent monitoring system for preventing electric vehicles from entering the unit door. Summary of the invention

[0004] The present invention aims to solve one of the technical problems in the prior art to at least a certain extent. By proposing a system for preventing electric vehicles from entering a unit door, it is used to solve the problem that the prior art usually adopts the traditional access control method to control the entry of people or electric vehicles into the unit door, or adopts a single camera for automatic recognition during identification. Although this method can perform recognition to a certain extent, if an object similar to an electric vehicle appears or the object is extremely similar to the electric vehicle in the image taken at the angle of the camera, the problem of false alarm due to inaccurate recognition may occur.

[0005] To achieve the above purpose, a system for preventing electric vehicles from entering a unit door includes a camera arrangement module, a three-dimensional image acquisition module, a model building and identification module, and an access control warning module;

[0006] The camera arrangement module is used to install a camera on the outside of the unit door, and to shoot objects close to the unit door based on the camera;

[0007] The three-dimensional image acquisition module is used to analyze the images captured by the camera arrangement module and acquire the three views of the vehicle based on the analysis results, wherein the three views of the vehicle include a front view contour, a top view contour and a left view contour;

[0008] The model building module is used to perform three-dimensional modeling of the vehicle approaching the unit door based on the three views of the vehicle to obtain the model to be judged, and to identify the model to be judged based on the convolutional neural network;

[0009] The access control warning module is used to control the alarm device in the control unit door to sound an alarm when the identification result of the model building and identification module is an electric vehicle.

[0010] Furthermore, the camera arrangement module includes an environment analysis arrangement unit, and the environment analysis arrangement unit is configured with an environment analysis arrangement strategy, and the environment analysis arrangement strategy includes:

[0011] Place an AI camera on the wall above the unit door and adjust the angle of the AI ​​camera based on the location of the unit door so that the AI ​​camera can monitor objects in front of the unit door in real time;

[0012] Get the length and width of the unit door, recorded as door length and door width respectively; place a camera at the center of the outer door panel of the unit door, recorded as the front camera; establish a spatial coordinate system, recorded as the camera layout coordinate system, wherein the units of the X-axis, Y-axis and Z-axis of the camera layout coordinate system are all meters; draw a rectangle in the XZ plane of the camera layout coordinate system based on the door length and door width, recorded as the door rectangle, place the door rectangle in the first quadrant of the XZ plane, and make one vertex of the door rectangle coincide with the coordinate origin, and make the width of the door rectangle coincide with the X-axis.

[0013] Furthermore, the environmental analysis layout strategy also includes:

[0014] Obtain the intersection of the diagonal lines of the door rectangle and record it as (X1, 0, Z1), record the point (X1, Z1, 2×Z1) and the point (0, Z1, Z1) in the camera arrangement coordinate system as the overhead point and the left point, respectively, and record the intersection of the straight line drawn from the left point and parallel to the X axis and the straight line drawn from the overhead point and parallel to the Z axis as the positioning point; correspond the side of the door rectangle facing the first quadrant to the side of the unit door facing the outside, and based on the positional relationship between the overhead point and the left point and the door rectangle, install cameras at positions outside the unit door corresponding to the overhead point and the left point, respectively, and record them as the overhead camera and the left camera, wherein the orientations of the overhead camera and the left camera are both the positions of the positioning points;

[0015] The area enclosed by points (0, 0, 0), (2×X1, 0, 0), (0, 2×Z1, 0) and (2×X1, 2×Z1, 0) in the camera layout coordinate system is recorded as the door front range, and the area corresponding to the door front range on the ground outside the unit door is recorded as the door front area.

[0016] Furthermore, the three-dimensional image acquisition module includes an electric vehicle image acquisition unit, and the electric vehicle image acquisition unit is configured with an electric vehicle image acquisition strategy, and the electric vehicle image acquisition strategy includes:

[0017] When there is a moving object in the AI ​​camera, the positional relationship between the moving object and the area in front of the door is detected. When the moving object moves into the area in front of the door and remains stationary, the front camera, the top camera and the left camera are started to shoot, and the images captured are recorded as the front real-time image, the top real-time image and the left real-time image, respectively.

[0018] Furthermore, the electric vehicle image acquisition strategy also includes:

[0019] The front real-time image, the top real-time image and the left real-time image are analyzed using the old background analysis method. The old background analysis method is as follows: for any one of the front real-time image, the top real-time image and the left real-time image, obtain the image taken by the camera corresponding to the actual image when there is no moving object in the AI ​​camera, and record it as a static image; grayscale the static image, and record the static image obtained by the grayscale processing as a static grayscale image; establish a plane rectangular coordinate system, recorded as the old background analysis coordinate system, wherein the units of the X-axis and the Y-axis of the old background analysis coordinate system are both m; binarize the static grayscale image, and put the binarized static grayscale image into the old background analysis coordinate system; and put the binarized static grayscale image in the old background analysis coordinate system. Recorded as the old background image; grayscale processing is performed on the real-time image, and the real-time image after grayscale processing is recorded as the real-time grayscale image; the real-time grayscale image is binarized, and the real-time grayscale image after binarization processing is recorded as the binary real-time image; the binary real-time image is placed in the old background analysis coordinate system, and the border of the binary real-time image is overlapped with the border of the old background image, and the lines in the binary real-time image that overlap with the old background image are removed, and the complete contour in the binary real-time image after removal is recorded as the contour to be analyzed; the object corresponding to the contour to be analyzed in the real-time image is recorded as the object to be analyzed, AI is used to identify the object to be analyzed, and the area where the human body is located in the object to be analyzed is removed based on the recognition result, and the contour to be analyzed corresponding to the object to be analyzed at this time is recorded as the face contour;

[0020] The surface contours obtained from the front real-time image, the top real-time image, and the left real-time image are recorded as the front view contour, the top view contour, and the left view contour, respectively.

[0021] Furthermore, the model building and identification module includes a model building unit and an electric vehicle identification unit. The model building unit is configured with an electric vehicle model building strategy. The electric vehicle model building strategy includes:

[0022] A spatial coordinate system is established, recorded as the model establishment coordinate system, wherein the units of the X-axis, Y-axis and Z-axis of the model establishment coordinate system are all in meters; based on three-dimensional drawing software, the corresponding model is drawn in the model establishment coordinate system with the front view contour, the top view contour and the left view contour as the three views, and the obtained model is recorded as the model to be judged.

[0023] Furthermore, the electric vehicle identification unit is configured with an electric vehicle identification strategy, and the electric vehicle identification strategy includes:

[0024] A convolutional neural network is established, and models of various electric vehicles and objects similar to electric vehicles obtained from the Internet are put into the convolutional neural network for training. Electric vehicles and objects similar to electric vehicles are identified based on the training results.

[0025] Furthermore, the electric vehicle identification strategy also includes:

[0026] The model to be judged is put into the convolutional neural network for analysis. In the convolution layer: the three views of the model to be judged are recorded as the feature three views, and the feature three views are analyzed for similarity based on the three views of all models stored in the convolutional neural network. The similarity analysis is as follows: the model whose similarity between each image in the three views and the corresponding image in the feature three views is greater than the standard similarity is recorded as a selectable model; and all selectable models are obtained.

[0027] Furthermore, the electric vehicle identification strategy also includes:

[0028] In the pooling layer, for any selectable model, the three similarities corresponding to the three views of the selectable model in the similarity analysis are recorded as a1, a2 and a3 respectively, and the average of a1, a2 and a3 is recorded as the similarity mean. The similarity fluctuation value of the selectable model is obtained using the similarity fluctuation algorithm. The similarity fluctuation algorithm is: Among them, F is the similarity fluctuation value, a0 is the similarity mean; obtain the similarity fluctuation values ​​and similarity means of all selectable models; record the average of all similarity fluctuation values ​​as the fluctuation mean; record the average of all similarity means as the phase mean; record the selectable models whose similarity fluctuation values ​​are less than the similarity mean and whose similarity means are greater than the phase mean as the screened models.

[0029] Furthermore, the electric vehicle identification strategy also includes:

[0030] In the fully connected layer, based on the objects corresponding to all the screened models, the object with the greatest similarity to the model to be judged is obtained and recorded as the recognized object; when the recognized object is an electric vehicle, the recognition result is sent to the access control warning module as an electric vehicle; when the recognized object is not an electric vehicle, the recognition result is not sent to the access control warning module.

[0031] Beneficial effects of the present invention: The present invention firstly shoots an object near a unit door based on a camera; analyzes the shot image, and obtains the three views of the vehicle based on the analysis result. This has the advantage that, by shooting the object near the unit door and obtaining the three views of the vehicle, the object near the unit door and not a person can be separated from the image and the three views of the object can be obtained, so that in subsequent analysis, the entity model of the object can be obtained based on the three views of the object, thereby making the recognition of the electric vehicle more accurate, and preventing the problem of false alarm caused by inaccurate recognition;

[0032] The present invention also obtains a model to be judged by three-dimensional modeling of the vehicle approaching the unit door, and distinguishes the model to be judged based on a convolutional neural network; finally, when the identification result of the model building and identification module is an electric vehicle, the alarm device in the unit door is controlled to alarm, and image recognition can be combined with access control to realize intelligent management and control of electric vehicles entering the unit door. The advantage of this is that by obtaining the model to be judged based on the three views of the vehicle and using the convolutional neural network to identify the model to be judged, the accuracy of the recognition result can be further improved, while ensuring that the electric vehicle can be quickly identified and an alarm can be issued when it is in front of the unit door to prevent congestion in the unit door. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 is a functional block diagram of the system of the present invention;

[0034] Figure 2 Schematic diagram of arranging the coordinate system for the camera;

[0035] Figure 3 It is a schematic diagram of obtaining the contour to be analyzed according to the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0037] Example 1, first aspect, please refer to Figure 1 As shown, the present application provides a system for preventing electric vehicles from entering a unit door, including a camera arrangement module, a three-dimensional image acquisition module, a model building and identification module, and a door access warning module.

[0038] The camera arrangement module is used to install a camera outside the unit door, and to shoot objects close to the unit door based on the camera; the camera arrangement module includes an environment analysis arrangement unit, and the environment analysis arrangement unit is configured with an environment analysis arrangement strategy, and the environment analysis arrangement strategy includes:

[0039] Place an AI camera on the wall above the unit door and adjust the angle of the AI ​​camera based on the location of the unit door so that the AI ​​camera can monitor objects in front of the unit door in real time;

[0040] Get the length and width of the unit door, record them as door length and door width respectively; place a camera at the center of the door panel outside the unit door, record it as the front camera; establish a spatial coordinate system, record it as the camera layout coordinate system, where the units of the X-axis, Y-axis and Z-axis of the camera layout coordinate system are all meters; draw a rectangle in the XZ plane of the camera layout coordinate system based on the door length and door width, record it as the door rectangle, place the door rectangle in the first quadrant of the XZ plane, and one vertex of the door rectangle coincides with the coordinate origin, and the width of the door rectangle coincides with the X axis;

[0041] Obtain the intersection of the diagonal lines of the door rectangle and record it as (X1, 0, Z1), record the point (X1, Z1, 2×Z1) and the point (0, Z1, Z1) in the camera arrangement coordinate system as the overhead point and the left point, respectively, and record the intersection of the straight line drawn from the left point and parallel to the X axis and the straight line drawn from the overhead point and parallel to the Z axis as the positioning point; correspond the side of the door rectangle facing the first quadrant to the side of the unit door facing the outside, and based on the positional relationship between the overhead point and the left point and the door rectangle, install cameras at positions outside the unit door corresponding to the overhead point and the left point, respectively, and record them as the overhead camera and the left camera, wherein the orientations of the overhead camera and the left camera are both the positions of the positioning points;

[0042] The area enclosed by the points (0, 0, 0), (2×X1, 0, 0), (0, 2×Z1, 0) and (2×X1, 2×Z1, 0) in the camera arrangement coordinate system is recorded as the door front range, and the area on the ground outside the unit door corresponding to the door front range is recorded as the door front area;

[0043] In the specific implementation process, by installing the front camera, the top camera and the left camera, the three views of the objects in the area in front of the door, that is, the area in front of the unit door, can be effectively photographed. The three views obtained by shooting can provide effective data support for the subsequent establishment of the model to be analyzed; for example, if the length and width of a unit door are 3 meters and 2 meters respectively, the camera arrangement coordinate system established can refer to Figure 2 As shown, the rectangle corresponding to DY1 is the door rectangle, point ZM is the intersection of the diagonals of the door rectangle, and the coordinates of point ZM are (1,0,1.5), point FM and point GM are the overhead point and the left point, respectively, and the coordinates of point FM are (1,1.5,3), and the coordinates of point GM are (0,1.5,1.5). In addition, the area formed by point QY1, point QY2, point (2,0,0) and the coordinate origin is the door front range; by setting the door front range to obtain the door front area, and installing a front camera, a overhead camera and a left camera, the three-view image of the object in front of the unit door can be effectively obtained, which helps to make the subsequent model to be judged more accurate than the real object.

[0044] The three-dimensional image acquisition module is used to analyze the images captured by the camera arrangement module, and obtain the three views of the vehicle based on the analysis results, wherein the three views of the vehicle include a front view profile, a top view profile, and a left view profile; the three-dimensional image acquisition module includes an electric vehicle image acquisition unit, and the electric vehicle image acquisition unit is configured with an electric vehicle image acquisition strategy, and the electric vehicle image acquisition strategy includes:

[0045] When there is a moving object in the AI ​​camera, the positional relationship between the moving object and the area in front of the door is detected. When the moving object moves into the area in front of the door and is stationary, the front camera, the top camera and the left camera are started to shoot, and the images captured are recorded as the front real-time image, the top real-time image and the left real-time image respectively;

[0046] The front real-time image, the top real-time image and the left real-time image are analyzed using the old background analysis method. The old background analysis method is as follows: for any one of the front real-time image, the top real-time image and the left real-time image, obtain the image taken by the camera corresponding to the actual image when there is no moving object in the AI ​​camera, and record it as a static image; grayscale the static image, and record the static image obtained by the grayscale processing as a static grayscale image; establish a plane rectangular coordinate system, recorded as the old background analysis coordinate system, wherein the units of the X-axis and the Y-axis of the old background analysis coordinate system are both m; binarize the static grayscale image, and put the binarized static grayscale image into the old background analysis coordinate system; and put the binarized static grayscale image in the old background analysis coordinate system. Recorded as the old background image; grayscale processing is performed on the real-time image, and the real-time image after grayscale processing is recorded as the real-time grayscale image; the real-time grayscale image is binarized, and the real-time grayscale image after binarization processing is recorded as the binary real-time image; the binary real-time image is placed in the old background analysis coordinate system, and the border of the binary real-time image is overlapped with the border of the old background image, and the lines in the binary real-time image that overlap with the old background image are removed, and the complete contour in the binary real-time image after removal is recorded as the contour to be analyzed; the object corresponding to the contour to be analyzed in the real-time image is recorded as the object to be analyzed, AI is used to identify the object to be analyzed, and the area where the human body is located in the object to be analyzed is removed based on the recognition result, and the contour to be analyzed corresponding to the object to be analyzed at this time is recorded as the face contour;

[0047] In the specific implementation process, the binarization process is to set the grayscale value of the pixel on the image to 0 or 255, so that the entire image presents an obvious black and white effect, thereby effectively obtaining the outline of the object in the image. In the old background analysis method of this embodiment, by using the binarization process, the outlines of all objects in the real-time grayscale image and the static grayscale image can be obtained, and the outline to be analyzed in the binary real-time image is obtained by overlapping the images in the old background analysis coordinate system, that is, the outline of the object in the door front area before entering the unit door, please refer to Figure 3 As shown, the image in the coordinate system where JD1 is located is the old background image, and the image in the coordinate system where JD2 is located is the binary real-time image. By overlapping JD1 and JD2 and removing the overlapping lines, the contour in the coordinate system where JD3 is located is the contour to be analyzed; through the front view contour, the top view contour and the left view contour obtained from the front real-time image, the top view real-time image and the left real-time image, the three views of the object in the door front area in front of the unit door can be accurately obtained, so that the obtained model to be judged is more accurate than the real object;

[0048] The surface contours obtained from the front real-time image, the top real-time image, and the left real-time image are recorded as the front view contour, the top view contour, and the left view contour, respectively.

[0049] The model building module is used to perform three-dimensional modeling of the vehicle near the unit door based on the three views of the vehicle to obtain the model to be judged, and to identify the model to be judged based on the convolutional neural network; the model building and identification module includes a model building unit and an electric vehicle identification unit, and the model building unit is configured with an electric vehicle model building strategy, and the electric vehicle model building strategy includes:

[0050] A spatial coordinate system is established, recorded as the model establishment coordinate system, wherein the units of the X-axis, Y-axis and Z-axis of the model establishment coordinate system are all in meters; based on the three-dimensional drawing software, the corresponding model is drawn in the model establishment coordinate system with the front view contour, the top view contour and the left view contour as the three views, and the obtained model is recorded as the model to be judged; in the specific implementation, the model can be established with the front view contour, the top view contour and the left view contour as the three views according to the three-dimensional drawing software that can be actually used, such as Autodesk Inventor, Google SketchUp, Autodesk Inventor Fusion, GraphiSoftArchicad and SpaceClaim and other software.

[0051] The electric vehicle identification unit is configured with an electric vehicle identification strategy, which includes:

[0052] Convolutional neural networks are built, and models of various electric vehicles and objects similar to electric vehicles obtained from the Internet are put into the convolutional neural network for training. Electric vehicles and objects similar to electric vehicles are identified based on the training results:

[0053] The model to be judged is put into the convolutional neural network for analysis. In the convolutional layer: the three views of the model to be judged are recorded as the feature three views, and the feature three views are analyzed for similarity based on the three views of all models stored in the convolutional neural network. The similarity analysis is as follows: the model whose similarity between each image in the three views and the corresponding image in the feature three views is greater than the standard similarity is recorded as a selectable model; all selectable models are obtained; in specific implementation, the standard similarity can be adjusted according to the actual situation in the actual scene. In this embodiment, the standard similarity is set to 75%;

[0054] In the pooling layer, for any selectable model, the three similarities corresponding to the three views of the selectable model in the similarity analysis are recorded as a1, a2 and a3 respectively, and the average of a1, a2 and a3 is recorded as the similarity mean. The similarity fluctuation value of the selectable model is obtained using the similarity fluctuation algorithm. The similarity fluctuation algorithm is: Among them, F is the similarity fluctuation value, a0 is the similarity mean; obtain the similarity fluctuation values ​​and similarity means of all selectable models; record the average value of all similarity fluctuation values ​​as the fluctuation mean; record the average of all similarity means as the phase mean; record the selectable models whose similarity fluctuation values ​​are less than the similarity mean and whose similarity mean is greater than the phase mean as the screened models; in specific implementation, for example, if the three similarities of the three views corresponding to a selectable model are 74%, 76% and 75% respectively, then the similarity fluctuation value of the selectable model can be calculated to be approximately 0.0 00067; By obtaining the similarity fluctuation value, the fluctuation of the similarity between the three views of the selectable model and the three views of the model to be judged can be obtained. For example, when the similarity fluctuation value is large, it means that when at least two images of the three views of the selectable model are compared with the three views of the model to be judged for similarity, the difference in similarity is too large. Therefore, at least one image of the three views of the selectable model is greatly different from the image of the model to be judged at the same angle, which means that the selectable model has little correlation with the model to be judged, and subsequent judgment can be omitted to reduce the load of data processing;

[0055] In the fully connected layer, based on the objects corresponding to all the screened models, the object with the greatest similarity to the model to be judged is obtained and recorded as the recognized object; when the recognized object is an electric vehicle, the recognition result is sent to the access control warning module as an electric vehicle; when the recognized object is not an electric vehicle, the recognition result is not sent to the access control warning module.

[0056] The access control warning module is used to control the alarm device in the control unit door to sound an alarm when the identification result of the model building and identification module is an electric vehicle.

[0057] Embodiment 2, second aspect, the difference between this embodiment and embodiment 1 is that the three-dimensional image acquisition module also includes a three-dimensional missing processing unit, and the three-dimensional missing processing unit is configured with a three-dimensional missing processing strategy, and the three-dimensional missing processing strategy includes: when using the old background analysis method for analysis, when any group of binarized real-time images in the old background analysis coordinate system overlaps with the old background image and the lines are removed, the remaining lines in the binarized real-time image cannot form a complete outline, the old background analysis method is interrupted, and the identification results are not sent to the access control warning module during the analysis of the objects in the area in front of the door; in specific implementation, when the old background analysis method is used for analysis, when the binarized real-time image overlaps with the old background image and the lines are removed, the remaining lines in the binarized real-time image cannot form a complete outline, it means that the object in the area in front of the door has almost no shape when viewed from one direction, and may be an object such as cardboard or glass, so there is no need to further process this analysis.

[0058] Through the description of the above implementation methods, the embodiments of the present invention can be provided as methods, systems or computer program products. Based on such an understanding, the above technical solutions can be essentially or partly contributed to the prior art in the form of software products, which can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and include several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0059] In the embodiments provided in the present application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or units can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of systems, modules and units can be electrical, mechanical or other forms.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. Prevent electric vehicles from entering the unit door system, characterized in that: It includes camera arrangement module, 3D image acquisition module, model building and identification module and access control warning module; The camera arrangement module is used to install a camera on the outside of the unit door, and to shoot objects close to the unit door based on the camera; The three-dimensional image acquisition module is used to analyze the images captured by the camera arrangement module and acquire the three views of the vehicle based on the analysis results, wherein the three views of the vehicle include a front view contour, a top view contour and a left view contour; The model building module is used to perform three-dimensional modeling of the vehicle approaching the unit door based on the three views of the vehicle to obtain the model to be judged, and to identify the model to be judged based on the convolutional neural network; The access control warning module is used to control the alarm device in the control unit door to sound an alarm when the identification result of the model building and identification module is an electric vehicle.

2. The system for preventing electric vehicles from entering a unit door according to claim 1, characterized in that: The camera arrangement module includes an environment analysis arrangement unit, and the environment analysis arrangement unit is configured with an environment analysis arrangement strategy, and the environment analysis arrangement strategy includes: Place an AI camera on the wall above the unit door and adjust the angle of the AI ​​camera based on the location of the unit door so that the AI ​​camera can monitor objects in front of the unit door in real time; Get the length and width of the unit door, recorded as door length and door width respectively; place a camera at the center of the outer door panel of the unit door, recorded as the front camera; establish a spatial coordinate system, recorded as the camera layout coordinate system, wherein the units of the X-axis, Y-axis and Z-axis of the camera layout coordinate system are all meters; draw a rectangle in the XZ plane of the camera layout coordinate system based on the door length and door width, recorded as the door rectangle, place the door rectangle in the first quadrant of the XZ plane, and make one vertex of the door rectangle coincide with the coordinate origin, and make the width of the door rectangle coincide with the X-axis.

3. The system for preventing electric vehicles from entering a unit door according to claim 2, characterized in that: Environmental analysis layout strategies also include: Obtain the intersection of the diagonal lines of the door rectangle and record it as (X1, 0, Z1), record the point (X1, Z1, 2×Z1) and the point (0, Z1, Z1) in the camera arrangement coordinate system as the overhead point and the left point, respectively, and record the intersection of the straight line drawn from the left point and parallel to the X axis and the straight line drawn from the overhead point and parallel to the Z axis as the positioning point; correspond the side of the door rectangle facing the first quadrant to the side of the unit door facing the outside, and based on the positional relationship between the overhead point and the left point and the door rectangle, install cameras at positions outside the unit door corresponding to the overhead point and the left point, respectively, and record them as the overhead camera and the left camera, wherein the orientations of the overhead camera and the left camera are both the positions of the positioning points; The area enclosed by points (0, 0, 0), (2×X1, 0, 0), (0, 2×Z1, 0) and (2×X1, 2×Z1, 0) in the camera layout coordinate system is recorded as the door front range, and the area corresponding to the door front range on the ground outside the unit door is recorded as the door front area.

4. The system for preventing electric vehicles from entering a unit door according to claim 3, characterized in that: The three-dimensional image acquisition module includes an electric vehicle image acquisition unit, and the electric vehicle image acquisition unit is configured with an electric vehicle image acquisition strategy. The electric vehicle image acquisition strategy includes: When there is a moving object in the AI ​​camera, the positional relationship between the moving object and the area in front of the door is detected. When the moving object moves into the area in front of the door and remains stationary, the front camera, the top camera and the left camera are started to shoot, and the images captured are recorded as the front real-time image, the top real-time image and the left real-time image, respectively.

5. The system for preventing electric vehicles from entering a unit door according to claim 4, characterized in that: Electric vehicle image acquisition strategies also include: The front real-time image, the top real-time image and the left real-time image are analyzed using the old background analysis method. The old background analysis method is as follows: for any one of the front real-time image, the top real-time image and the left real-time image, obtain the image taken by the camera corresponding to the actual image when there is no moving object in the AI ​​camera, and record it as a static image; grayscale the static image, and record the static image obtained by the grayscale processing as a static grayscale image; establish a plane rectangular coordinate system, recorded as the old background analysis coordinate system, wherein the units of the X-axis and the Y-axis of the old background analysis coordinate system are both m; binarize the static grayscale image, and put the binarized static grayscale image into the old background analysis coordinate system; and put the binarized static grayscale image in the old background analysis coordinate system. Recorded as the old background image; grayscale processing is performed on the real-time image, and the real-time image after grayscale processing is recorded as the real-time grayscale image; the real-time grayscale image is binarized, and the real-time grayscale image after binarization processing is recorded as the binary real-time image; the binary real-time image is placed in the old background analysis coordinate system, and the border of the binary real-time image is overlapped with the border of the old background image, and the lines in the binary real-time image that overlap with the old background image are removed, and the complete contour in the binary real-time image after removal is recorded as the contour to be analyzed; the object corresponding to the contour to be analyzed in the real-time image is recorded as the object to be analyzed, AI is used to identify the object to be analyzed, and the area where the human body is located in the object to be analyzed is removed based on the recognition result, and the contour to be analyzed corresponding to the object to be analyzed at this time is recorded as the face contour; The surface contours obtained from the front real-time image, the top real-time image, and the left real-time image are recorded as the front view contour, the top view contour, and the left view contour, respectively.

6. The system for preventing electric vehicles from entering a unit door according to claim 5, characterized in that: The model building and identification module includes a model building unit and an electric vehicle identification unit. The model building unit is configured with an electric vehicle model building strategy. The electric vehicle model building strategy includes: A spatial coordinate system is established, recorded as the model establishment coordinate system, wherein the units of the X-axis, Y-axis and Z-axis of the model establishment coordinate system are all in meters; based on three-dimensional drawing software, the corresponding model is drawn in the model establishment coordinate system with the front view contour, the top view contour and the left view contour as the three views, and the obtained model is recorded as the model to be judged.

7. The system for preventing electric vehicles from entering a unit door according to claim 6, characterized in that: The electric vehicle identification unit is configured with an electric vehicle identification strategy, which includes: A convolutional neural network is established, and models of various electric vehicles and objects similar to electric vehicles obtained from the Internet are put into the convolutional neural network for training. Electric vehicles and objects similar to electric vehicles are identified based on the training results.

8. The system for preventing electric vehicles from entering a unit door according to claim 7, characterized in that: Electric vehicle identification strategies also include: The model to be judged is put into the convolutional neural network for analysis. In the convolution layer: the three views of the model to be judged are recorded as the feature three views, and the feature three views are analyzed for similarity based on the three views of all models stored in the convolutional neural network. The similarity analysis is as follows: the model whose similarity between each image in the three views and the corresponding image in the feature three views is greater than the standard similarity is recorded as a selectable model; and all selectable models are obtained.

9. The system for preventing electric vehicles from entering a unit door according to claim 8, characterized in that: Electric vehicle identification strategies also include: In the pooling layer, for any selectable model, the three similarities corresponding to the three views of the selectable model in the similarity analysis are recorded as a1, a2 and a3 respectively, and the average of a1, a2 and a3 is recorded as the similarity mean. The similarity fluctuation value of the selectable model is obtained using the similarity fluctuation algorithm. The similarity fluctuation algorithm is: Among them, F is the similarity fluctuation value, a0 is the similarity mean; obtain the similarity fluctuation values ​​and similarity means of all selectable models; record the average of all similarity fluctuation values ​​as the fluctuation mean; record the average of all similarity means as the phase mean; record the selectable models whose similarity fluctuation values ​​are less than the similarity mean and whose similarity means are greater than the phase mean as the screened models.

10. The system for preventing electric vehicles from entering a unit door according to claim 9, characterized in that: Electric vehicle identification strategies also include: In the fully connected layer, based on the objects corresponding to all the screened models, the object with the greatest similarity to the model to be judged is obtained and recorded as the recognized object; when the recognized object is an electric vehicle, the recognition result is sent to the access control warning module as an electric vehicle; when the recognized object is not an electric vehicle, the recognition result is not sent to the access control warning module.

Images