An electric vehicle charging port identification method and system

Abstract

This invention pertains to the field of electric vehicle charging and provides a method and system for identifying electric vehicle charging ports. Both methods include: using a camera mounted on a charging robot arm to capture a predetermined number of images of the charging port of the electric vehicle to be charged, obtaining a first set of images; selecting any image from a preset neural network library whose similarity to any image in the first set reaches a preset second ratio threshold T2, and recording it as the target image; the second ratio threshold T2 being greater than or equal to 99%; retrieving the distance between the charging port of the electric vehicle corresponding to the target image and the zero point of the robot arm stored in the neural network library; and using the retrieved distance to obtain the pose coordinates of the charging port of the electric vehicle to be charged. This invention aims to improve the accuracy and efficiency of identifying the charging port of the electric vehicle to be charged.

Description

Technical Field

[0001] This invention relates to the field of electric vehicle charging, and specifically to a method and system for identifying electric vehicle charging ports. Background Technology

[0002] Charging technology is a key technology for building an internet-based public charging service network for electric vehicles, and the charging experience largely determines the user's driving experience. With the widespread application of electric vehicle technology, charging stations have emerged. Existing charging stations are either manually operated, requiring human assistance, or automated, enabling automatic charging of electric vehicles. As electric vehicles become more common and widely used, automated charging at these stations is becoming increasingly important. For a charging station to achieve automated charging, it needs to identify the charging port of the vehicle being charged.

[0003] Currently, charging stations typically use charging robots to identify the charging ports of vehicles waiting to be charged. Specifically, the charging robot is equipped with a robotic arm with a binocular camera. When identifying the charging port of a vehicle, the binocular camera on the robotic arm captures two images of the charging port. The pose coordinates of the charging port are then calculated using the angle between these two images. Furthermore, other features on the vehicle near the charging port are extracted to aid in the calculation. This method yields low-precision pose coordinates for the charging port, which hinders the charging robot from successfully connecting the charging gun to the vehicle's charging port. This often negatively impacts the user's charging experience and, consequently, their overall vehicle usage experience.

[0004] Therefore, the present invention provides a method and system for identifying electric vehicle charging ports. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for identifying electric vehicle charging ports, which improves the accuracy of identifying the charging ports of electric vehicles to be charged, thereby enabling a smooth connection between the charging gun and the charging port of the electric vehicle to be charged, improving the user's charging experience to a certain extent, and ultimately enhancing the user's vehicle usage experience.

[0006] In a first aspect, the present invention provides a method for identifying an electric vehicle charging port. This method is based on a charging robot, which has a robotic arm equipped with a camera. The electric vehicle charging port identification method includes:

[0007] Step S1: Use the camera to capture a predetermined number of images of the charging port of the electric vehicle to be charged, and obtain a set of first images;

[0008] Step S2: Select any image from the preset neural network library whose similarity to any image in the first image reaches a preset second ratio threshold T2, and record it as the target image; the neural network library contains several convolutional neural network recognition models that can calculate the distance between the charging port of the electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of the electric vehicle captured by the camera, as well as the corresponding image of the charging port of the electric vehicle captured by the camera and the distance between the charging port of the electric vehicle and the zero point of the robotic arm calculated by the model; the value of the second ratio threshold T2 is greater than or equal to 99%;

[0009] Step S3: Retrieve the distance between the charging port of the electric vehicle corresponding to the target image stored in the neural network library and the zero point of the robotic arm.

[0010] Step S4: Using the retrieved distance, obtain the pose coordinates of the charging port of the electric vehicle to be charged.

[0011] Furthermore, the electric vehicle charging port identification method also includes:

[0012] If there is no image in the neural network library that has a similarity to any image in the first image that reaches a preset second ratio threshold T2, then the convolutional neural network recognition model corresponding to the image with the highest similarity to the first image in the neural network library is selected, and deep self-learning is performed on each of the captured first images to obtain a new convolutional neural network recognition model. The distance between the electric vehicle charging port and the zero point of the robotic arm, calculated by inputting the corresponding input image into the new convolutional neural network recognition model, is stored in the neural network library, and then the process proceeds to step S2 to continue execution.

[0013] Furthermore, the neural network library contains a pre-trained convolutional neural network recognition model and its corresponding image of the charging port of the electric vehicle captured by the camera, as well as the calculated distance between the charging port of the electric vehicle and the zero point of the robotic arm.

[0014] The trained convolutional neural network recognition model is a pre-trained convolutional neural network model that can calculate the distance between the charging port of the electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of the electric vehicle captured by the camera.

[0015] Furthermore, the electric vehicle charging port identification method also includes:

[0016] Select images from the first images whose similarity to the target image reaches a preset first ratio threshold T1 to obtain several second images; the first ratio threshold T1 is less than the second ratio threshold T2, and the value of the first ratio threshold T1 is greater than or equal to 90%.

[0017] The second images are input into the convolutional neural network recognition models corresponding to the target images stored in the neural network library, and the distance between the charging port of the electric vehicle and the zero point of the robotic arm corresponding to each second image is calculated.

[0018] The second image, its corresponding convolutional neural network recognition model, and the distance between the electric vehicle's charging port and the robotic arm's zero point, calculated above, are stored in the neural network library.

[0019] Furthermore, the camera is a monocular camera.

[0020] Secondly, the present invention provides an electric vehicle charging port identification system, the system being based on a charging robot, the charging robot having a robotic arm equipped with a camera; the system includes:

[0021] The charging port image acquisition module is configured to use the camera to capture a preset number of images of the charging port of the electric vehicle to be charged, thereby obtaining a set of first images;

[0022] The image filtering module is configured to select any image from a preset neural network library whose similarity to any image in the first image reaches a preset second ratio threshold T2, and record it as the target image; the neural network library contains several convolutional neural network recognition models that can calculate the distance between the charging port of an electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of an electric vehicle captured by the camera, as well as the corresponding images of the charging ports of electric vehicles captured by the camera and the distances between the charging ports of the electric vehicles and the zero points of the robotic arm calculated by the models; the value of the second ratio threshold T2 is greater than or equal to 99%;

[0023] The distance retrieval module is configured to retrieve the distance between the charging port of the electric vehicle corresponding to the target image stored in the neural network library and the zero point of the robotic arm.

[0024] The charging port identification module is configured to obtain the pose coordinates of the charging port of the electric vehicle to be charged by using the retrieved distance.

[0025] Furthermore, the system also includes:

[0026] The system optimization module is configured to, when there is no image in the neural network library that has a similarity to any image in the first image that reaches a preset second ratio threshold T2, select the convolutional neural network recognition model corresponding to the image with the highest similarity to the first image in the neural network library, perform deep self-learning on each captured first image to obtain a new convolutional neural network recognition model, store the new convolutional neural network recognition model, its corresponding input image, and the distance between the electric vehicle charging port and the robotic arm zero point calculated after inputting the corresponding input image into the neural network library, and then call the image filtering module to continue execution.

[0027] Furthermore, the neural network library contains a pre-trained convolutional neural network recognition model and its corresponding image of the charging port of the electric vehicle captured by the camera, as well as the calculated distance between the charging port of the electric vehicle and the zero point of the robotic arm.

[0028] The trained convolutional neural network recognition model is a pre-trained convolutional neural network model that can calculate the distance between the charging port of the electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of the electric vehicle captured by the camera.

[0029] Furthermore, the system also includes:

[0030] Select images from the first images whose similarity to the target image reaches a preset first ratio threshold T1 to obtain several second images; the first ratio threshold T1 is less than the second ratio threshold T2, and the value of the first ratio threshold T1 is greater than or equal to 90%.

[0031] The second images are input into the convolutional neural network recognition models corresponding to the target images stored in the neural network library, and the distance between the charging port of the electric vehicle and the zero point of the robotic arm corresponding to each second image is calculated.

[0032] The second image, its corresponding convolutional neural network recognition model, and the distance between the electric vehicle's charging port and the robotic arm's zero point, calculated above, are stored in the neural network library.

[0033] Furthermore, the camera is a monocular camera.

[0034] The electric vehicle charging port identification method and system provided by this invention have the following advantages compared with the prior art:

[0035] (1) This invention is based on a pre-set neural network library. The neural network library contains several convolutional neural network recognition models that can calculate the distance between the charging port of an electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of the electric vehicle captured by the camera, as well as the corresponding images of the charging port of the electric vehicle captured by the camera and the calculated distances between the charging port of each corresponding electric vehicle and the zero point of the robotic arm. In use, a set of first images can be obtained by first using the camera on the robotic arm of the charging robot to capture a pre-set number of images of the charging port of the electric vehicle to be charged. Then, a similarity image from the pre-set neural network library to any image in the first image can be selected. Any image that reaches a preset second ratio threshold T2 is used as the target image. Then, the distance between the charging port of the electric vehicle and the zero point of the robotic arm corresponding to the target image stored in the neural network library is retrieved to obtain the pose coordinates of the charging port of the electric vehicle to be charged. The entire recognition process does not require additional extraction of features outside the vehicle's charging port for charging port location recognition, which is more convenient than existing technologies. Moreover, the recognition process is based on a trained convolutional neural network recognition model, which increases the recognition accuracy to a certain extent. This helps to facilitate the smooth connection between the charging gun and the charging port of the electric vehicle to be charged, which helps to improve the user's charging experience and, in turn, the user's vehicle use experience.

[0036] (2) After confirming the target image, the present invention directly retrieves the distance between the electric vehicle charging port corresponding to the target image and the zero point of the robotic arm in the neural network library, thereby obtaining the pose coordinates of the charging port of the electric vehicle to be charged, thus achieving the purpose of identifying the charging port of the electric vehicle to be charged, avoiding the angle calculation method used in the prior art to identify the car charging port, and to a certain extent helping to improve the recognition efficiency.

[0037] (3) When there is no image in the neural network library that has a similarity to any image in the first image that reaches a preset second ratio threshold T2, the present invention can select the convolutional neural network recognition model corresponding to the image with the highest similarity to the first image in the neural network library, perform deep self-learning on each of the captured first images, obtain a new convolutional neural network recognition model, and store the new convolutional neural network recognition model, its corresponding input image, and the distance between the electric vehicle charging port and the zero point of the robotic arm calculated after inputting the corresponding input image into the neural network library, which further helps to improve the recognition accuracy.

[0038] (4) For each electric vehicle charging port identified in this invention, several second images can be obtained by selecting images from the first image whose similarity to the target image reaches a preset first ratio threshold T1. The second images are then input into the convolutional neural network recognition model corresponding to the target image stored in the neural network library. The distance between the charging port of the electric vehicle corresponding to each second image and the zero point of the robotic arm is calculated. The second image, its corresponding convolutional neural network recognition model, and the distance between the charging port of the electric vehicle and the zero point of the robotic arm calculated above are then stored in the neural network library. This further enriches the content of the neural network library, helps to reduce the probability of deep self-learning in this invention, and thus helps to further improve the recognition speed.

[0039] (5) The camera on the robotic arm of the charging robot described in this invention can be a monocular camera, without the need for the binocular camera in the prior art, which helps to reduce costs to some extent.

[0040] Furthermore, the design principle of this invention is reliable, the structure is simple, and it has a very wide range of application prospects. Attached Figure Description

[0041] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention.

[0043] Figure 2 This is a schematic block diagram of a system according to an embodiment of the present invention. Detailed Implementation

[0044] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0045] The core of this invention is to provide a method and system for identifying electric vehicle charging ports, which improves the accuracy of identifying the charging ports of electric vehicles to be charged, so as to achieve a smooth connection between the charging gun and the charging port of the electric vehicle to be charged, thereby improving the user's charging experience to a certain extent and thus improving the user's vehicle use experience.

[0046] Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention.

[0047] The method in this embodiment is based on a charging robot, which has a robotic arm with a camera mounted on it.

[0048] Figure 1 The implementing entity can be the aforementioned charging robot, or it can be the control center of the charging station (which can be a server or a PC).

[0049] like Figure 1 As shown, the method 100 includes steps 110 to 140.

[0050] Step 110: Use the camera to capture a predetermined number of images of the charging port of the electric vehicle to be charged, and obtain a set of first images.

[0051] In this embodiment, the camera is a monocular camera, which avoids the use of binocular cameras in the prior art and reduces costs to a certain extent.

[0052] Step 120: Select any image from the preset neural network library whose similarity to any image in the first image reaches a preset second ratio threshold T2, and record it as the target image.

[0053] The aforementioned neural network library contains several convolutional neural network recognition models that can calculate the distance between the charging port of an electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of the electric vehicle captured by the camera, as well as the corresponding image of the charging port of the electric vehicle captured by the camera and the distance between the charging port of each corresponding electric vehicle and the zero point of the robotic arm calculated by the model.

[0054] Specifically, the initial neural network library contains a pre-trained convolutional neural network recognition model, an image of the electric vehicle's charging port captured by the corresponding camera, and the calculated distance between the electric vehicle's charging port and the robotic arm's zero point.

[0055] The aforementioned trained convolutional neural network recognition model is a pre-trained convolutional neural network model that can calculate the distance between the charging port of the electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of the electric vehicle captured by the camera.

[0056] The training set of the above-mentioned trained convolutional neural network recognition model is constructed as follows: take a set of images of the charging port of any electric vehicle using the camera (for example, 1000 images can be taken), and measure the distance between the charging port of the electric vehicle and the zero point of the robotic arm on the charging robot for each image taken; use the set of images taken and the corresponding distances between the charging port of the electric vehicle and the zero point of the robotic arm on the charging robot to construct the above-mentioned training set.

[0057] The value of the second proportional threshold T2 mentioned above is greater than or equal to 99%, which helps to ensure the accuracy of recognition to a certain extent.

[0058] In specific implementation, if there is no image in the neural network library that has a similarity to any image in the first image that reaches a preset second ratio threshold T2, then the convolutional neural network recognition model corresponding to the image with the highest similarity to the first image in the neural network library is selected, and deep self-learning is performed on each of the captured first images to obtain a new convolutional neural network recognition model. The distance between the electric vehicle charging port and the zero point of the robotic arm, calculated by inputting the corresponding input image into the new convolutional neural network recognition model, is stored in the neural network library, and then the process proceeds to step 120 to continue execution.

[0059] Step 130: Retrieve the distance between the charging port of the electric vehicle corresponding to the target image stored in the neural network library and the zero point of the robotic arm.

[0060] This application directly retrieves the distance between the charging port of the electric vehicle corresponding to the target image stored in the neural network library and the zero point of the robotic arm, which improves the recognition efficiency to a certain extent.

[0061] Step 140: Using the retrieved distance, obtain the pose coordinates of the charging port of the electric vehicle to be charged.

[0062] As an illustrative embodiment of the present invention, the method 100 further includes:

[0063] Select images from the first image whose similarity to the target image reaches a preset first ratio threshold T1 to obtain several second images; the first ratio threshold T1 is less than the second ratio threshold T2, and the value of the first ratio threshold T1 is greater than or equal to 90%.

[0064] The second images are input into the convolutional neural network recognition models corresponding to the target images stored in the neural network library, and the distance between the charging port of the electric vehicle and the zero point of the robotic arm corresponding to each second image is calculated.

[0065] The second image, its corresponding convolutional neural network recognition model, and the distance between the electric vehicle's charging port and the robotic arm's zero point, calculated above, are stored in the neural network library.

[0066] This enriches the data stored in the neural network library, which to some extent helps reduce the probability of deep self-learning occurring in this invention, and thus helps to further improve the recognition rate.

[0067] Figure 2 This is a schematic block diagram of a system according to an embodiment of the present invention.

[0068] The system in this embodiment is based on a charging robot, which has a robotic arm with a camera mounted on it.

[0069] Figure 2 As shown, the system 200 includes:

[0070] The charging port image acquisition module 201 is configured to use the camera to capture a preset number of images of the charging port of the electric vehicle to be charged, and obtain a set of first images;

[0071] The image filtering module 202 is configured to select any image from a preset neural network library whose similarity to any image in the first image reaches a preset second ratio threshold T2, and record it as the target image; the neural network library contains several convolutional neural network recognition models that can calculate the distance between the charging port of an electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of an electric vehicle captured by the camera, as well as the corresponding images of the charging ports of electric vehicles captured by the camera and the distances between the charging ports of each corresponding electric vehicle and the zero point of the robotic arm calculated by the models; the value of the second ratio threshold T2 is greater than or equal to 99%;

[0072] The distance retrieval module 203 is configured to retrieve the distance between the charging port of the electric vehicle corresponding to the target image stored in the neural network library and the zero point of the robotic arm.

[0073] The charging port identification module 204 is configured to obtain the pose coordinates of the charging port of the electric vehicle to be charged by using the retrieved distance.

[0074] As an illustrative embodiment of the present invention, the system 200 further includes:

[0075] The system optimization module is configured to, when there is no image in the neural network library that has a similarity to any image in the first image that reaches a preset second ratio threshold T2, select the convolutional neural network recognition model corresponding to the image with the highest similarity to the first image in the neural network library, perform deep self-learning on each captured first image to obtain a new convolutional neural network recognition model, store the new convolutional neural network recognition model, its corresponding input image, and the distance between the electric vehicle charging port and the robotic arm zero point calculated after inputting the corresponding input image into the neural network library, and then call the image filtering module 202 to continue execution.

[0076] As an illustrative embodiment of the present invention, the neural network library contains a pre-trained convolutional neural network recognition model and its corresponding image of the charging port of the electric vehicle captured by the camera, as well as the distance between the charging port of the electric vehicle and the zero point of the robotic arm calculated therefrom.

[0077] The trained convolutional neural network recognition model is a pre-trained convolutional neural network model that can calculate the distance between the charging port of the electric vehicle and the zero point of the robotic arm by inputting an image of the charging port of the electric vehicle captured by the camera.

[0078] As an illustrative embodiment of the present invention, the system further includes:

[0079] Select images from the first images whose similarity to the target image reaches a preset first ratio threshold T1 to obtain several second images; the first ratio threshold T1 is less than the second ratio threshold T2, and the value of the first ratio threshold T1 is greater than or equal to 90%.

[0080] The second images are input into the convolutional neural network recognition models corresponding to the target images stored in the neural network library, and the distance between the charging port of the electric vehicle and the zero point of the robotic arm corresponding to each second image is calculated.

[0081] The second image, its corresponding convolutional neural network recognition model, and the distance between the electric vehicle's charging port and the robotic arm's zero point, calculated above, are stored in the neural network library.

[0082] As an illustrative embodiment of the present invention, the camera is a monocular camera.

[0083] The same or similar parts between the various embodiments in this specification can be referred to mutually. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the description in the method embodiments.

[0084] The distances mentioned in this manual are all of the same type, such as Euclidean distances.

[0085] The charging ports of electric vehicles mentioned in this manual all belong to the same standard, such as the national standard or the European standard.

[0086] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the invention should also be covered within the protection scope of the invention. Therefore, the protection scope of the invention should be determined by the scope of the claims.

Claims

1. An electric vehicle charging port identification method, the method is based on a charging robot, the charging robot is provided with a mechanical arm, and a camera is arranged on the mechanical arm; characterized in that, The electric vehicle charging port recognition method comprises: Step S1, a pre-set number of images of the charging port of the electric vehicle to be charged are captured by the camera to obtain a first image set; Step S2, any one image in the pre-set neural network library that has a similarity to any one of the first images reaching a pre-set second proportion threshold T2 is selected as a target image; the neural network library stores a plurality of convolutional neural network recognition models capable of calculating the distance between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm by inputting the images of the charging port of the electric vehicle captured by the camera, the images of the charging port of the electric vehicle captured by the camera corresponding to the convolutional neural network recognition models, and the distances between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm calculated by the convolutional neural network recognition models; the value of the second proportion threshold T2 is greater than or equal to 99%; Step S3, the distance between the charging port of the electric vehicle corresponding to the target image and the mechanical arm zero point of the mechanical arm stored in the neural network library is called; Step S4, the pose coordinates of the charging port of the electric vehicle to be charged are obtained by using the called distance; The electric vehicle charging port recognition method further comprises: If there is no image in the neural network library that has a similarity to any one of the first images reaching the pre-set second proportion threshold T2, the convolutional neural network recognition model corresponding to the image having the highest similarity to the first image in the neural network library is selected, the first images captured are subjected to deep self-learning, a new convolutional neural network recognition model is obtained, the new convolutional neural network recognition model, the input image corresponding to the new convolutional neural network recognition model, and the distance between the charging port of the electric vehicle and the mechanical arm zero point of the mechanical arm calculated by inputting the input image into the new convolutional neural network recognition model are stored in the neural network library, and then the step S2 is executed.

2. The method of claim 1, wherein, The neural network library pre-stores a trained convolutional neural network recognition model, the images of the charging port of the electric vehicle captured by the camera corresponding to the trained convolutional neural network recognition model, and the distances between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm calculated by the trained convolutional neural network recognition model; The trained convolutional neural network recognition model is a pre-trained convolutional neural network model capable of calculating the distance between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm by inputting the images of the charging port of the electric vehicle captured by the camera.

3. The method of claim 1, wherein, The electric vehicle charging port recognition method further comprises: The images in the first image set that have a similarity to the target image reaching a pre-set first proportion threshold T1 are selected to obtain a plurality of second images; the first proportion threshold T1 is less than the second proportion threshold T2, and the value of the first proportion threshold T1 is greater than or equal to 90%; The second images are respectively input into the convolutional neural network recognition model corresponding to the target image stored in the neural network library to calculate the distance between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm corresponding to each second image; and The second image, the corresponding convolutional neural network recognition model, and the distance between the charging port of the electric vehicle and the mechanical arm zero point of the mechanical arm obtained through the above calculation are stored in the neural network library.

4. The method of claim 1, wherein, The camera is a monocular camera.

5. An electric vehicle charging port identification system, the system is based on a charging robot, the charging robot is provided with a mechanical arm, and a camera is arranged on the mechanical arm; characterized in that, The system comprises: The charging port image acquisition module is configured to use the camera to take a preset number of images of the charging port of the electric vehicle to be charged to obtain a group of first images. The image screening module is configured to select any one image in the preset neural network library that has a similarity to any one of the first images reaching a preset second proportion threshold T2, and record the image as a target image; the neural network library stores a plurality of convolutional neural network recognition models capable of calculating the distance between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm by inputting the image of the charging port of the electric vehicle taken by the camera, the corresponding image of the charging port of the electric vehicle taken by the camera, and the distance between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm calculated therefrom; the value of the second proportion threshold T2 is greater than or equal to 99%. The distance calling module is configured to call the distance between the charging port of the electric vehicle corresponding to the target image and the mechanical arm zero point of the mechanical arm stored in the neural network library. The charging port recognition module is configured to use the called distance to obtain the pose coordinates of the charging port of the electric vehicle to be charged. The system further comprises: The system optimization module is configured to, when there is no image in the neural network library that has a similarity to any one of the first images reaching the preset second proportion threshold T2, select the convolutional neural network recognition model corresponding to the image in the neural network library that has the highest similarity to the first image, perform deep self-learning on each of the taken first images to obtain a new convolutional neural network recognition model, store the new convolutional neural network recognition model, the corresponding input image, and the distance between the charging port of the electric vehicle and the mechanical arm zero point of the mechanical arm calculated after inputting the corresponding input image in the neural network library, and then call the image screening module to continue execution.

6. The electric vehicle charge port identification system of claim 5, wherein, The neural network library pre-stores a trained convolutional neural network recognition model, the corresponding image of the charging port of the electric vehicle taken by the camera, and the distance between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm calculated therefrom. The trained convolutional neural network recognition model is a convolutional neural network model that is pre-trained to be capable of calculating the distance between the charging port of the corresponding electric vehicle and the mechanical arm zero point of the mechanical arm by inputting the image of the charging port of the electric vehicle taken by the camera.

7. The electric vehicle charge port identification system of claim 5, wherein, The system further comprises: The images in the first images that have a similarity to the target image reaching a preset first proportion threshold T1 are selected to obtain a plurality of second images; the first proportion threshold T1 is less than the second proportion threshold T2, and the value of the first proportion threshold T1 is greater than or equal to 90%. The second image is input into a convolutional neural network recognition model corresponding to the target image stored in the neural network library, and the distance between the charging port of the electric vehicle and the mechanical arm zero point of the mechanical arm corresponding to each second image is calculated; The second image, its corresponding convolutional neural network recognition model, and the distance between the charging port of the electric vehicle and the mechanical arm zero point of the mechanical arm obtained by the above calculation are stored in the neural network library.

8. The electric vehicle charge port identification system of claim 5, wherein, The camera adopts a monocular camera.

Images