Charging robot risk perception method and system based on visual identification

Abstract

The invention discloses a charging robot risk perception method and system based on visual identification, which are applied to a charging robot and a charging robot risk perception system, and the system comprises a charging robot provided with an image acquisition device and a charging gun head; the risk perception method comprises the following steps: acquiring real-time image data around the charging robot by using an image acquisition device; identifying the real-time image data to obtain real-time state information of the charging robot; and risk perception information is determined according to the real-time image data and the real-time state information of the charging robot. According to the method and the system disclosed by the invention, whether the current state of the charging robot has the risk or not is sensed based on visual identification, so that the possible risk can be quickly dealt with, and the safety of the charging robot during aerial track type walking and charging is ensured.

Description

Technical Field

[0001] This invention relates to the field of mobile charging equipment control technology, and in particular to a risk perception method and system for charging robots based on visual recognition. Background Technology

[0002] To achieve flexible, efficient, and space-efficient charging, an aerial track-mounted miniature charging robot has been proposed in existing technologies. This aerial track-mounted miniature charging robot uses a track-mounted design to transfer charging facilities from the ground to the air, thus solving the charging problem. This charging robot is particularly suitable for charging new energy vehicles.

[0003] However, existing aerial track-mounted small charging robots typically only use ultrasonic radar to prevent surrounding risks when walking or charging in the track. This method has problems such as limited detection range, insufficient accuracy and poor environmental adaptability, and is prone to misjudgment and response delay, which cannot meet the requirements of system safety and reliability.

[0004] Therefore, the existing technology needs further improvement. Summary of the Invention

[0005] The purpose of this invention is to provide a risk perception method and system for charging robots based on visual recognition, overcoming the shortcomings of low overall safety and reliability of existing charging robots when they are walking on tracks or charging.

[0006] The technical solution adopted by this invention to solve the technical problem is as follows: In a first aspect, this application discloses a risk perception method for charging robots based on visual recognition, wherein it is applied to a risk perception system for charging robots based on visual recognition, the risk perception system for charging robots includes: a charging robot; the charging robot is equipped with an image acquisition device and a charging gun head; The risk perception method includes: Real-time image data of the area surrounding the charging robot is acquired using an image acquisition device; Real-time image data is identified to obtain real-time status information of the charging robot; wherein, the real-time status information includes one or more of the following: the plugging / unplugging status of the charging gun head, the status information of the target risk object, and the relative position information of the target risk object and the charging robot; Risk perception information is determined based on the real-time image data and the real-time status information of the charging robot.

[0007] Optionally, the charging robot risk perception system further includes: a backend server connected to the charging robot; wherein the backend server is equipped with an AI data analysis model for analyzing and judging the received data information; The steps for determining risk perception information based on the real-time image data and the current real-time status information of the charging robot include: The real-time image data and the real-time status information of the charging robot are sent to the backend server; The AI ​​data analysis model is used to analyze the received real-time image data and real-time status information to determine the corresponding risk perception information.

[0008] Optionally, the image acquisition device includes at least one AI camera; the AI ​​camera is located at the front end of the charging robot; the charging robot is mounted on a slide rail and moves along the slide rail, or the charging robot is in a fixed position on the slide rail and is in a charging state; the real-time image data includes: frontal image data and surrounding environment image data captured by the AI ​​camera in multiple consecutive time periods; the frontal image data is the image data in front of the charging robot when it moves along the slide rail; the surrounding environment image data is the surrounding environment data when the charging robot is moving or the surrounding environment data when the charging robot is stationary.

[0009] Optionally, the training method for the AI ​​data analysis model includes: Multiple images are selected from real-time images of the charging robot captured by the AI ​​camera in the past, and these selected images are used as training samples to construct a training sample set. The selected images are taken at regular intervals or at regular distances during the charging process of the charging robot during normal movement or charging. Each training sample in the training sample set is input into a preset neural network model. After the preset neural network model learns from each received training sample, a trained AI data analysis model is obtained.

[0010] Optionally, the step of identifying real-time image data to obtain the real-time status information of the current charging robot includes: Identify whether the charging gun head and target risk objects are present in the frontal image data and surrounding environment image data captured by the AI ​​camera; If the device contains a charging gun head or a target risk object, then determine the current plugging / unplugging status of the charging gun head, the status information of the target risk object, and the relative position information of the target risk object and the charging robot.

[0011] Optionally, the step of analyzing the received real-time image data and real-time status information using the AI ​​data analysis model to determine the corresponding risk perception information includes: The backend server uses an AI data analysis model to analyze the received relative position information between the target risk object and the charging robot, as well as the historical relative position information between the target risk object and the charging robot, to determine whether the distance between the target risk object and the charging robot is increasing or decreasing. If the distance between the target risk object and the charging robot is decreasing, the estimated time of collision is predicted based on the current movement speed of the charging robot and the preset distance warning threshold. Calculate the time difference between the estimated time and the current time, and determine the probability of collision risk based on the magnitude of the time difference.

[0012] Optionally, the backend server utilizes an AI data analysis model to analyze the received relative position information between the target risk object and the charging robot, as well as the historical relative position information between the target risk object and the charging robot, to determine whether the distance between the target risk object and the charging robot is increasing or decreasing. This includes the following steps: Images of the charging robot in front of it, captured at preset time intervals or distances, are selected from real-time image data. By comparing the current forward image with the historical forward image captured at the previous moment, the positional changes of the target risk object in two consecutive forward images or multiple consecutive forward images can be determined. Based on the positional changes of the target risk object in continuous forward images, determine whether the relative distance between the target risk object and the charging robot is increasing or decreasing.

[0013] Optionally, the target risk object includes: a target area with smoke characteristics; the step of using the AI ​​data analysis model to analyze the received real-time image data and the real-time status information of the charging robot to determine the corresponding risk perception information further includes: The AI ​​data analysis model is used to analyze the real-time image data, identify target areas in the real-time image data that contain smoke features, and determine the probability of fire risk based on the area size and location of the target areas containing smoke features.

[0014] Optionally, the step of analyzing the received real-time image data and the real-time status information using the AI ​​data analysis model to determine the corresponding risk perception information further includes: If the current plug-in / plug-out state of the charging gun head is the charging state, then based on the plug-in / plug-out state of the charging gun head identified in historical moments, the charging duration of the charging gun head in the current charging state is calculated, and the charging risk probability is determined based on the charging duration. Based on the charging risk probability, collision risk probability, and fire risk probability, a risk perception level corresponding to the real-time status information of the current charging robot is determined.

[0015] Secondly, this application also discloses a visual recognition-based risk perception system for charging robots, comprising: a charging robot; the charging robot is equipped with an image acquisition device, a charging gun head, and a processor; The image acquisition device is used to acquire real-time image data around the charging robot; The processor is used to identify real-time image data to obtain real-time status information of the charging robot; wherein, the real-time status information includes one or more of the following: the plugging / unplugging status of the charging gun head, the status information of the target risk object, and the relative position information of the target risk object and the charging robot, and risk perception information is determined based on the real-time image data and the current real-time status information of the charging robot.

[0016] Beneficial effects: This embodiment provides a visual recognition-based risk perception method and system for charging robots, applied to a charging robot application risk perception system. The system includes a charging robot equipped with an image acquisition device and a charging gun. The risk perception method includes: acquiring real-time image data of the charging robot's surroundings using the image acquisition device; identifying the real-time image data to obtain real-time status information of the charging robot; wherein the real-time status information includes one or more of the following: the insertion / removal status of the charging gun, the status information of a target risk object, and the relative position information of the target risk object and the charging robot; and determining risk perception information based on the real-time image data and the real-time status information of the charging robot. The method and system disclosed in this invention perceive the presence of risk in the current state of the charging robot based on visual recognition, thereby quickly responding to potential risks and improving the safety of aerial track-walking charging robots. Attached Figure Description

[0017] Figure 1 A flowchart illustrating the steps of the risk perception method for charging robots provided by the present invention; Figure 2 A side perspective view of the charging robot provided by the present invention; Figure 3 A frontal perspective view of the charging robot provided by the present invention; Figure 4 This is a schematic diagram showing the positions of various components in the charging robot provided by the present invention. Figure 5 This is a schematic diagram of the structure of the charging gun head in the charging robot provided by the present invention; Figure 6 A flowchart of the collision perception steps in the risk perception method provided by this invention; Figure 7 A flowchart illustrating the steps of smoke feature perception in the risk perception method provided by this invention; Figure 8 The flowchart shows the steps for determining the plugging / unplugging status of the charging gun head in the risk perception method provided by the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0019] Other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein.

[0020] An aerial track-mounted charging robot is a charging device that moves along a suspended track, identifies the location of the device to be charged, and connects the charging gun. This charging robot can achieve precise positioning and efficient charging via a pre-set track, and is suitable for charging needs in various scenarios such as new energy vehicles, industrial equipment, and logistics robots.

[0021] In existing technologies, aerial track-mounted charging robots often rely solely on ultrasonic radar to detect risks such as collisions when operating on the track. Therefore, this method of risk detection is limited and cannot guarantee the safe operation of the charging robot on the track.

[0022] To overcome the aforementioned problems, this invention provides a visual recognition-based risk perception method and system for charging robots. It utilizes an image acquisition device to acquire real-time image data of the charging robot's surroundings. By recognizing the acquired real-time image data, one or more of the following can be obtained: the current insertion / removal status of the charging gun, the status information of the target risk object, and the relative position information of the target risk object and the charging robot. Based on this information, the current risk perception information of the charging robot is determined, enabling rapid prediction of potential risk events that the charging robot may encounter. This facilitates the elimination of potential risks, ensuring the charging robot can successfully complete its charging operation and guaranteeing its operational safety.

[0023] The following is a detailed description of the risk perception method and system for charging robots based on visual recognition provided in this application, with reference to the accompanying drawings.

[0024] Firstly, such as Figure 1 As shown, this embodiment discloses a risk perception method and system for charging robots based on visual recognition. The risk perception system for charging robots is applied as follows: Figures 2 to 4 As shown, the charging robot risk perception system includes: a charging robot 10; the front end of the charging robot is equipped with an image acquisition device and a charging gun head 30. The image acquisition device is used to capture image information around the charging robot, and it includes at least one AI camera. In one embodiment, as... Figure 4 As shown, in order to acquire panoramic information about the area surrounding the charging robot, the image acquisition device includes a first AI camera 101 and a second AI camera 102. It is conceivable that this image acquisition device could be a single panoramic camera or multiple AI cameras positioned at different angles.

[0025] Furthermore, such as Figure 2 As shown, the charging robot has an overall hexahedral structure, which can be designed as a cuboid or cube. A pulley system is installed on one face of the charging robot, allowing it to slide back and forth on track 20. Taking the face with the pulley system as the bottom face, the face opposite the bottom face is called the top face. An image acquisition device and a charging gun are respectively located at the two edges adjacent to the top face. The image acquisition device is used to acquire images of the scene in front of and around the charging robot, so as to analyze the current state of the charging robot based on the acquired images.

[0026] The charging gun head provides a physical interface for charging the device, through which electrical energy from the charging station is transferred to the battery system of the device. For example... Figure 3 As shown, the charging gun head can be positioned below the charging robot body, facilitating the establishment of a physical interface connection. When the device to be charged (such as a new energy vehicle) needs charging, the charging robot slides along the track to the designated position, and the charging gun head can automatically or manually establish a connection with the device to be charged to achieve the transfer of electrical energy.

[0027] like Figure 1 As shown, the risk perception method includes: Step S1: Use an image acquisition device to acquire real-time image data of the area around the charging robot.

[0028] An image acquisition device located on the charging robot captures images of the area surrounding the robot to obtain real-time image data. In this embodiment, the image acquisition device includes at least one AI camera; the AI ​​camera is located at the front end of the charging robot; the charging robot is mounted on a slide rail and moves along the slide rail, or the charging robot is in a fixed position on the slide rail and is in a charging state; the real-time image data includes: frontal image data captured by the AI ​​camera within multiple consecutive time periods and image data of the surrounding environment; the frontal image data is the image data in front of the charging robot when it moves along the slide rail; the surrounding environment image data is the surrounding environment data when the charging robot is moving or the surrounding environment data when the charging robot is stationary.

[0029] In one implementation, a first AI camera and a second AI camera are respectively installed on both sides of the front end of the charging robot. The first AI camera and the second AI camera are both 180° fisheye cameras. Since the 180° fisheye camera has an ultra-wide field of view, it can extend the horizontal field of view to 180° and the vertical field of view to more than 150°. Therefore, setting 180° fisheye cameras on both sides can achieve all-round coverage of the surrounding environment of the charging robot and eliminate monitoring blind spots.

[0030] Furthermore, in combination Figure 4 As shown, a lighting device 103 is also provided between the first AI camera 101 and the second AI camera 102. Since the lighting device 103 can provide a stable light source, it can supplement the first AI camera 101 and the second AI camera 102 to compensate for the increase in camera image noise and loss of details that may be caused by insufficient ambient light, so that the camera can capture a clear image.

[0031] To obtain more accurate information about the surrounding environment, in practice, the first and second AI cameras capture images at preset time intervals, or when the charging robot moves along the track, it captures an image every preset distance. For example, it captures an image every second, or every meter it travels. The images captured by the first and second AI cameras can be stored in a historical image database. This allows for easy comparison between currently captured and historical images, identifying differences in the position of potential hazards in different images, determining the presence of a risk based on these differences, and then implementing appropriate countermeasures accordingly.

[0032] Furthermore, the images captured by the first and second AI cameras may contain other charging robots, charging gun heads, or target risk objects that may pose a risk to the operation of the charging robots. In actual operation, target risk objects can be divided into fixed-position risk objects and dynamically changing-position risk objects. Fixed-position risk objects are typically track attachments, such as track beams or supports, track end plates at the edge of buildings, walls or pillars, lighting fixtures installed on the track, and surveillance cameras. Dynamically changing-position risk objects may be operators or other automatically moving charging robots. Fixed-position risk objects can be avoided by planning the charging robot's walking route to prevent collisions. Dynamically changing risk objects, due to their randomness of appearance, require real-time monitoring and the implementation of corresponding avoidance systems. The risk perception mentioned in this embodiment includes collision events caused by dynamically changing target risk objects.

[0033] In addition, during the charging process, the target risk objects that may pose a risk to the normal operation of the charging robot may also include harmful gases such as smoke and fire generated in the charging gun head or the surrounding environment. Therefore, in addition to containing risk objects with fixed positions and risk objects with dynamically changing positions, the images captured by the first AI camera and the second AI camera may also include the charging gun heads of other charging robots and smoke and fire generated in the environment.

[0034] Step S2: Identify the real-time image data to obtain the real-time status information of the current charging robot; wherein, the real-time status information includes one or more of the following: the current plugging / unplugging status of the charging gun head, the current location information of the target risk object, and the historical location information of the target risk object.

[0035] After real-time image data is captured using the first AI camera and the second AI camera in step S1 above, the real-time image data is identified to determine whether the original image contains the aforementioned target risk object. If it does, the captured target risk object needs to be analyzed for information such as its location and historical location to determine the potential risk event that the charging robot may be in.

[0036] Furthermore, the real-time image data is identified to obtain information including: the current plugging / unplugging status of the charging gun head, the status information of the target risk object, and the relative position information of the target risk object and the charging robot.

[0037] In this step, the processor within the charging robot can be used to analyze the information contained in the real-time image data to identify whether the aforementioned charging gun head and various target risk objects are present. Alternatively, the real-time image data can be sent to a connected backend server, which can then identify the charging gun head and various target risk objects within the real-time image data.

[0038] To more accurately identify the charging gun head and various target hazards, this step involves establishing AI data analysis models for three different types of target hazards. These established and trained AI data analysis models are used to identify the objects. The three types of AI data analysis models are: a first AI data analysis model for identifying target hazards with fixed locations; a second AI data analysis model for identifying target hazards with dynamically changing locations; and a third AI data analysis model for identifying features such as smoke and fire in the image. Since each of the three categories of target hazards has certain characteristics, using three different AI data analysis models for identification improves accuracy.

[0039] In practical implementation, the training methods for the aforementioned AI data analysis model include: First, multiple images are selected from the real-time images of the charging robot captured by the AI ​​camera in the past. These selected images are used as training samples to construct a training sample set. The selected images are taken at regular intervals or at regular distances during the charging process of the charging robot during normal movement or charging.

[0040] Once the AI ​​cameras are installed, they can capture images of the surrounding environment as the charging robots move normally on their tracks and charge normally, and save these images to an image database. When it is necessary to train an AI data analysis model, multiple images are selected from the image database to construct a training sample set. In one implementation, the multiple images in the training sample set are standard images captured by the AI ​​cameras at regular intervals or after the charging robots have moved a certain distance.

[0041] Secondly, each training sample in the training sample set is input into a preset neural network model. After the preset neural network model learns from each received training sample, a trained AI data analysis model is obtained.

[0042] The pre-defined neural network model is trained using the constructed training sample set to obtain a trained AI data analysis model.

[0043] In detail, when training an AI data analysis model for fixed-location target hazards, the preset neural network model can be a network model combining a structured light camera and deep learning. This neural network model can improve the accuracy of identification. For example, an AI camera can capture images of fixed-location target hazards, and a convolutional neural network can be used to classify the target hazards in the images, thus training a first AI data analysis model for identifying fixed-location target hazards. When selecting the preset network model for training an AI data analysis model for target hazards with changing locations, a second AI data analysis model can be trained using target detection and multi-target tracking algorithms. For the third AI data analysis model for smoke and fire identification, when selecting the preset neural network model, it can be achieved by embedding SE modules or CBAM in a CNN to focus on salient features of smoke / fire (such as color and texture), or by capturing long-distance dependencies through global interaction in a Transformer architecture (a deep learning model architecture based on self-attention), improving the detection capability of scattered smoke. Furthermore, adversarial examples can be added to the training data to improve the accuracy of the trained model in identifying smoke and fire. It is conceivable that, in order to improve efficiency, a trained AI data analysis model can be used to identify the various target risks mentioned above during implementation.

[0044] Furthermore, the step of identifying real-time image data to obtain the real-time status information of the current charging robot includes: The system identifies whether the charging gun head and the target risk object are present in the frontal image data and the surrounding environment image data captured by the first AI camera and the second AI camera, respectively. If the charging gun head or the target risk object is present, the system determines the current plugging / unplugging status of the charging gun head and the current and historical location information of the target risk object.

[0045] Once real-time image data is acquired, an AI data analysis model is used to identify objects in each image to determine whether the image contains any of the aforementioned target risk objects. If any of the aforementioned target risk objects are present, the plugging / unplugging status of the charging gun head and the current and historical location information of the target risk object are obtained.

[0046] Specifically, the plugging / unplugging state of the charging gun head refers to the physical connection state between the charging gun head and the charging interface (such as an electric vehicle charging port or a charging pile socket). This is typically categorized into insertion (connection), removal (disconnection), and intermediate transition states (such as partial insertion or poor contact). In this embodiment, only the insertion and removal states are detected. Combined with... Figure 5As shown, a charging tag 301 is provided on the charging interface of the charging gun head. When the charging gun head is in the charging state, the charging tag 301 is covered by the interface and cannot be identified. When the charging gun head is not in the charging state, the charging tag 301 is exposed and can be identified. Therefore, when identifying the plugging and unplugging status of the charging gun head, the current plugging and unplugging status of the charging gun head can be determined by checking whether the charging tag is found.

[0047] The current and historical location information of the target risk object is calculated based on its position in the image. In practice, target detection and localization algorithms are first used to locate the target risk object in the image, and then the current position of the target risk object is determined based on the located coordinates. Historical location information is obtained by detecting historical images.

[0048] Step S3: Determine the risk perception information based on the real-time image data and the real-time status information of the current charging robot.

[0049] Once the aforementioned real-time image data and real-time status information are acquired, the risk perception information of the current charging robot can be determined based on these data.

[0050] In one implementation, the charging robot risk perception system further includes a backend server connected to the charging robot; wherein the backend server is equipped with an AI data analysis model for analyzing the received data information.

[0051] The steps for determining risk perception information based on the real-time image data and the current real-time status information of the charging robot include: The real-time image data and the current real-time status information of the charging robot are sent to the backend server; the AI ​​data analysis model is used to analyze the received relative position information between the target risk object and the charging robot and the historical relative position information between the target risk object and the charging robot to determine the corresponding risk perception information.

[0052] In detail, the steps of analyzing the received real-time image data and the real-time status information of the current charging robot using the AI ​​data analysis model to determine the corresponding risk perception information include: The backend server uses an AI data analysis model to analyze the relative position information of the target risk object and the charging robot, as well as the historical relative position information of the target risk object and the charging robot, to determine whether the relative distance between the target risk object and the charging robot is increasing or decreasing. If the relative distance between the target risk object and the charging robot is decreasing, the probability of a collision risk between the target risk object and the charging robot is calculated.

[0053] In detail, combined Figure 6 As shown, the specific steps for determining the probability of a collision between the target hazardous object and the charging robot are as follows: H1. The charging robot starts moving on the track.

[0054] H2, the first AI camera, and the second AI camera continuously capture images of the front and the surrounding environment, and send the captured images to the backend server.

[0055] H3. The backend server compares the received captured images with historical captured images in the image storage database to identify whether the relative distance between the target risk object and the current charging robot is gradually decreasing.

[0056] H4. If the distance between the target risk object and the current charging robot is not gradually decreasing, control the charging robot to walk normally on the track.

[0057] H5. If the distance between the target risk object and the current charging robot is gradually decreasing, determine whether there is a collision risk between the target risk object and the charging robot, and calculate the risk probability.

[0058] H6. If there is a risk of collision, notify the charging robot to stop moving on the track.

[0059] H7. Report the potential collision risk event to the operations and maintenance platform so that the platform can investigate the incident and provide timely response strategies.

[0060] Furthermore, the backend server utilizes an AI data analysis model to analyze the current location information and historical location information of the target risk object, and determines whether the distance between the target risk object and the charging robot is increasing or decreasing. This process includes: H31. Filter out images of the charging robot moving in front of it from real-time image data at preset time intervals or preset distances.

[0061] H32. Compare the current forward image captured with the historical forward image captured in the previous moment to determine the positional change of the target risk object in two consecutive forward images or multiple consecutive forward images. H33. Based on the positional changes of the target risk object in the continuous forward images, determine whether the distance between the target risk object and the charging robot is increasing or decreasing.

[0062] Furthermore, the step of calculating the probability of a collision between the target risk object and the charging robot if the distance between them is decreasing includes: H51. If the distance between the target risk object and the charging robot is decreasing, the estimated time of collision is predicted based on the current movement speed of the charging robot and the preset distance warning threshold. H52. Calculate the time difference between the estimated time and the current time, and determine the collision risk probability based on the magnitude of the time difference.

[0063] To improve the accuracy of risk perception, the target risk object in this embodiment also includes: a target area with smoke characteristics; by identifying whether there are smoke characteristics in the image, the risk of fire can be perceived.

[0064] The step of using the AI ​​data analysis model to analyze the received real-time image data and the real-time status information of the current charging robot to determine the corresponding risk perception information further includes: The AI ​​data analysis model is used to analyze the real-time image data, identify target areas containing smoke features, and determine the probability of a fire based on the size and location of the target areas containing smoke features.

[0065] Combination Figure 7 As shown, the detailed steps for identifying target regions containing smoke features in real-time image data include: K1 transmits real-time image data to the backend server.

[0066] K2. The backend server uses an AI data analysis model to identify whether there are areas in each image in the real-time image data that match the characteristics related to "fire, smoke, and fog".

[0067] K3. If an area matching the characteristics of smoke or fire is detected in the image, pause the operation of the charging robot, record the area containing the smoke feature, and proceed to step K4; otherwise, return to K1.

[0068] K4. Send the recording to the maintenance platform so that the maintenance platform can determine whether to restart the charging machine based on the recording.

[0069] Furthermore, to improve the step of using the AI ​​data analysis model to analyze the received real-time image data and the real-time status information of the current charging robot to determine the corresponding risk perception information, the method further includes: If the current plug-in / plug-out status of the charging gun head is the charging state, then based on the plug-in / plug-out status of the charging gun head identified in historical moments, the charging duration of the charging gun head in the current charging state is calculated, and the charging risk probability is determined based on the charging duration; based on the charging risk probability, collision risk probability, and fire risk probability, the risk perception level corresponding to the real-time status information of the current charging robot is determined.

[0070] To further improve the accuracy of risk perception, this step also identifies the current plugging / unplugging status of the charging gun head, and assesses the probability of charging risk that each charging gun head may fail due to long charging time based on the plugging / unplugging status and charging time. This charging risk probability is then combined with the collision risk probability and fire risk probability to obtain the overall risk perception level.

[0071] Combination Figure 8 As shown, in a specific application embodiment, the step of identifying the current plugging / unplugging state of the charging gun head includes: J1. First, the two AI cameras take images of the location of the charging gun head.

[0072] J2. The real-time image data containing the charging gun head is sent to the backend server. The backend server selects the images containing the charging gun head from the real-time data images captured by the two AI cameras.

[0073] J3. Identify the area where the charging gun head is located in the image, and identify whether the image contains a charging tag for the charging gun head.

[0074] When identifying charging tags on charging gun heads, an image acquisition device can first be used to collect images of the charging tags on the charging gun head, and these images with the charging tags can be saved to a backend server. When it is necessary to identify the area where the charging gun head is located in the image, the real-time data image acquired is compared and analyzed with the image with the charging tag saved to the backend server to determine whether the charging tag is present.

[0075] J4. If the charging tag is not detected, the charging gun head of the charging robot is not detached from the vehicle, confirming that the charging gun head is in the inserted state.

[0076] J5. If the charging tag is detected in the area where the charging gun head is located, it is determined that the charging gun head has been detached from the vehicle and is in an uninserted state. Therefore, it is necessary to notify the charging robot that the charging gun head has been detached from the vehicle.

[0077] J6. Report the results to the backend server.

[0078] In practice, the detection results of whether the charging gun head is in the inserted state can be cross-referenced with the data connection status between the charging gun head and the vehicle to accurately determine the insertion and connection status of the charging gun head.

[0079] In this step, the risk perception level of the current charging robot is obtained by comprehensively considering three aspects: the probability of charging risk, the probability of collision risk, and the probability of fire risk. The obtained risk perception level is then reported to the operation and maintenance platform so that the operation and maintenance platform can identify potential risks based on the risk perception level and handle them in a timely manner.

[0080] Furthermore, the risk perception level of the charging robot can be obtained by comprehensively considering the three aspects: charging risk probability, collision risk probability, and fire risk probability. This can be achieved by assigning a weight to each risk probability and then using that weight to determine the risk perception level. These weights can be derived from a comprehensive summary of a large amount of data.

[0081] The risk perception method disclosed in this embodiment takes pictures of the direction of travel and the surrounding environment through an AI camera and a visual recognition system. It uses image recognition to identify objects in the direction of travel and their distances, as well as the plugging and unplugging status of the charging gun cable. It also uses targeted recording to store surrounding accident monitoring videos. By obtaining the current risk perception level of the charging robot from multiple different angles, it can respond to risk events in a timely manner and improve the safety of using the charging robot.

[0082] Secondly, the present invention also provides a risk perception system for charging robots based on visual recognition. The risk perception system includes: a charging robot body, an image acquisition device disposed at the front end of the charging robot body, and a charging gun head and processor disposed on the charging robot body.

[0083] The image acquisition device is used to acquire real-time image data around the charging robot.

[0084] The processor is used to identify real-time image data to obtain real-time status information of the current charging robot; wherein, the real-time status information includes one or more of the following: the current plugging / unplugging status of the charging gun head, the current location information of the target risk object, and the historical location information of the target risk object, as well as risk perception information determined based on the real-time image data and the real-time status information of the current charging robot.

[0085] This embodiment provides a visual recognition-based risk perception method and system for charging robots, applied to a charging robot application risk perception system. The system includes: a robot body, an image acquisition device disposed at the front end of the robot body, and a charging gun disposed on the robot body. The risk perception method includes: acquiring real-time image data of the charging robot's surroundings using the image acquisition device; identifying the real-time image data to obtain the current real-time status information of the charging robot; and determining risk perception information based on the real-time image data and the current real-time status information of the charging robot. The method and system disclosed in this invention perceive the presence of risk in the current state of the charging robot based on visual recognition, thereby quickly responding to potential risks and improving the safety of aerial track-mounted charging robots.

[0086] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0087] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A risk perception method for charging robots based on visual recognition, characterized in that, A risk perception system for a charging robot based on vision recognition is provided. The risk perception system for the charging robot includes: a charging robot; the charging robot is equipped with an image acquisition device and a charging gun. The risk perception method includes: Real-time image data of the area surrounding the charging robot is acquired using an image acquisition device; Real-time image data is identified to obtain real-time status information of the charging robot; wherein, the real-time status information includes one or more of the following: the plugging / unplugging status of the charging gun head, the status information of the target risk object, and the relative position information of the target risk object and the charging robot; Risk perception information is determined based on the real-time image data and the real-time status information of the charging robot.

2. The risk perception method for charging robots according to claim 1, characterized in that, The charging robot risk perception system further includes: a backend server connected to the charging robot; wherein the backend server is equipped with an AI data analysis model for analyzing and judging the received data information; The steps for determining risk perception information based on the real-time image data and the real-time status information of the charging robot include: The real-time image data and the real-time status information of the charging robot are sent to the backend server; The AI ​​data analysis model is used to analyze the received real-time image data and real-time status information to determine the corresponding risk perception information.

3. The risk perception method for charging robots according to claim 2, characterized in that, The image acquisition device includes at least one AI camera; the AI ​​camera is located at the front end of the charging robot; the charging robot is mounted on a slide rail and moves along the slide rail, or the charging robot is in a fixed position on the slide rail and is in a charging state. The real-time image data includes: frontal image data and surrounding environment image data captured by the AI ​​camera in multiple consecutive time periods; The foreground image data is the image data in front of the charging robot when it moves along the slide rail; the surrounding environment image data is the surrounding environment data when the charging robot is moving or the surrounding environment data when the charging robot is stationary.

4. The risk perception method for charging robots according to claim 3, characterized in that, The training method for the AI ​​data analysis model includes: Multiple images are selected from real-time images of the charging robot captured by the AI ​​camera in the past, and these selected images are used as training samples to construct a training sample set. The selected images are taken at regular intervals or at regular distances during the charging process of the charging robot during normal movement or charging. Each training sample in the training sample set is input into a preset neural network model. After the preset neural network model learns from each received training sample, a trained AI data analysis model is obtained.

5. The risk perception method for charging robots according to claim 4, characterized in that, The step of identifying real-time image data to obtain the real-time status information of the current charging robot includes: Identify whether the charging gun head and target risk objects are present in the frontal image data and surrounding environment image data captured by the AI ​​camera; If the device contains a charging gun head or a target risk object, then determine the current plugging / unplugging status of the charging gun head, the status information of the target risk object, and the relative position information of the target risk object and the charging robot.

6. The risk perception method for charging robots according to claim 3, characterized in that, The step of analyzing the received real-time image data and real-time status information using the AI ​​data analysis model to determine the corresponding risk perception information includes: The backend server uses an AI data analysis model to analyze the received relative position information between the target risk object and the charging robot, as well as the historical relative position information between the target risk object and the charging robot, to determine whether the distance between the target risk object and the charging robot is increasing or decreasing. If the distance between the target risk object and the charging robot is decreasing, the estimated time of collision is predicted based on the current movement speed of the charging robot and the preset distance warning threshold. Calculate the time difference between the estimated time and the current time, and determine the probability of collision risk based on the magnitude of the time difference.

7. The risk perception method for charging robots according to claim 6, characterized in that, The backend server uses an AI data analysis model to analyze the received relative position information between the target risk object and the charging robot, as well as the historical relative position information between the target risk object and the charging robot. The steps to determine whether the distance between the target risk object and the charging robot is increasing or decreasing include: Images of the charging robot in front of it, captured at preset time intervals or distances, are selected from real-time image data. By comparing the current forward image with the historical forward image captured at the previous moment, the positional changes of the target risk object in two consecutive forward images or multiple consecutive forward images can be determined. Based on the positional changes of the target risk object in continuous forward images, determine whether the relative distance between the target risk object and the charging robot is increasing or decreasing.

8. The risk perception method for charging robots according to claim 5, characterized in that, The target risk object includes: a target area with smoke characteristics; the step of using the AI ​​data analysis model to analyze the received real-time image data and the real-time status information of the charging robot to determine the corresponding risk perception information further includes: The AI ​​data analysis model is used to analyze the real-time image data, identify target areas in the real-time image data that contain smoke features, and determine the probability of fire risk based on the area size and location of the target areas containing smoke features.

9. The risk perception method for charging robots according to claim 8, characterized in that, The step of analyzing the received real-time image data and real-time status information using the AI ​​data analysis model to determine the corresponding risk perception information further includes: If the current plug-in / plug-out state of the charging gun head is the charging state, then based on the plug-in / plug-out state of the charging gun head identified in the past, the charging duration of the charging gun head in the current charging state is calculated, and the charging risk probability is determined based on the charging duration. Based on the charging risk probability, collision risk probability, and fire risk probability, a risk perception level corresponding to the real-time status information of the current charging robot is determined.

10. A risk perception system for charging robots based on visual recognition, characterized in that, include: A charging robot; the charging robot is equipped with an image acquisition device, a charging gun head, and a processor; The image acquisition device is used to acquire real-time image data around the charging robot; The processor is used to identify real-time image data to obtain real-time status information of the charging robot; wherein, the real-time status information includes one or more of the following: the plugging / unplugging status of the charging gun head, the status information of the target risk object, and the relative position information of the target risk object and the charging robot, and risk perception information is determined based on the real-time image data and the real-time status information of the charging robot.

Images