Electric power marketing intelligent robot based on large model technology

Abstract

An electric power marketing intelligent robot based on a large model technology is provided with a control system, the control system comprises an image acquisition module and a central control analysis module, and the image acquisition module captures facial expressions of a user in real time through a camera and transmits the facial expressions to the central control analysis module; the central control analysis module calculates a user head inclination angle through a large model algorithm, and judges the height deviation between a screen and sight; detecting the actions of squinting the eyes and wrinkling the eyebrows of the user by using an expression recognition algorithm; the distance adjusting module controls the moving mechanism to adjust the distance between the screen and a user, and the height adjusting module drives the lifting device to calibrate the sight angle. When squinting is detected, the screen brightness, the contrast ratio and the font size are automatically optimized; if the rugosa is recognized, the residence time of the current page is prolonged, and the extension explanation information is superposed. According to the method, the intention of the user can be accurately understood, the interaction rhythm is adjusted in real time, the use comfort of different user groups is remarkably improved, and the intelligent and self-adaptive power marketing service experience is realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an intelligent robot for power marketing based on large model technology. Background Art

[0002] Power marketing is the process by which power companies, in a changing market environment, aim to meet consumer demand for electricity through a series of market-related operations, providing electricity products and services that meet consumer needs. Within this field, intelligent telemarketing robots can be used for customer consultation, product promotion, and troubleshooting, helping companies achieve more efficient, accurate, and personalized sales and driving innovation in the power marketing industry. Essentially, intelligent telemarketing robots are software programs based on artificial intelligence (AI) technology, designed to simulate human language and behavior, engage in intelligent conversations with customers, and achieve business goals such as sales, promotion, and customer service.

[0003] For example, the patent document with publication number CN117557222A discloses an intelligent bill collection robot system based on power marketing. The invention uses the user classification module and IVR voice platform in the intelligent bill collection robot system to realize batch intelligent call reminders and intelligent call collection for users' electricity bill payment, and can provide users with intelligent voice notifications of electricity payment policies through telephone calls, which greatly reduces the labor costs required by power companies in the electricity bill collection link, improves the efficiency and success rate of electricity bill collection, and makes the collection records traceable, further improves work efficiency, and increases user satisfaction with the power company's services; in addition, through the voice interaction module in the IVR voice platform, intelligent interaction with users can also be achieved, allowing users to understand and process electricity bill-related data through more channels and more conveniently, thereby enhancing the user experience.

[0004] With technological advancements, the technology behind power marketing robots is becoming increasingly sophisticated. In addition to virtual robots used online, similar to those in the aforementioned application, a growing number of physical robots are being deployed in power companies' offline stores. Compared to traditional human service, telemarketing robots can provide services 24 / 7 without rest, significantly increasing the reach and frequency of sales activities. Furthermore, telemarketing robots strictly adhere to pre-set scripts and logic, ensuring consistency and professionalism in every communication and avoiding the uneven communication effectiveness often experienced by human service due to factors such as emotion and experience.

[0005] However, in practice, power marketing is a continuous activity, and user sentiment (user comfort) directly impacts their patience and success rate. Therefore, it's crucial for power marketing robots to provide highly comfortable service. However, currently available intelligent robots are generally only capable of basic adaptive adjustments, such as adjusting their height based on the user's height, which is clearly insufficient for power marketing. Summary of the Invention

[0006] The purpose of the present invention is to provide an electric power marketing intelligent robot based on large model technology to solve the problems raised in the above background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solutions: an intelligent power marketing robot based on large model technology, comprising:

[0008] A control system, comprising:

[0009] An image acquisition module is configured to capture a user's facial expression image in real time through a camera and transmit the image to a central control analysis module;

[0010] Central control analysis module, including:

[0011] The first analysis unit is used to analyze the head tilt angle of the user viewing the screen based on the large model algorithm and determine the height deviation between the robot and the user's line of sight;

[0012] The second analysis unit is used to detect whether the user has a squinting expression through an expression recognition algorithm;

[0013] a third analysis unit, configured to detect whether the user has a frowning expression through an expression recognition algorithm;

[0014] a distance adjustment module, connected to the first analysis unit, controlling the movement mechanism to perform forward and backward or left and right displacement according to the height deviation, so that the robot screen maintains a preset distance range from the user;

[0015] a height adjustment module, which drives the lifting mechanism to adjust the vertical height of the robot screen according to the height deviation output by the first analysis unit, so that the robot screen maintains a suitable angle with the user's line of sight;

[0016] a screen adjustment module, responsive to the squinting detection signal from the second analysis unit, to dynamically adjust the brightness, contrast, and font size of the screen;

[0017] The response adjustment module triggers the delayed parking control of the current page of the screen after receiving the frown detection signal from the third analysis unit, and generates extended explanation information to be superimposed and displayed on the parking page.

[0018] Furthermore, the control system further includes:

[0019] Voice prompt module: used to remind users to relax and look at the screen through voice when they interact with the robot; the voice prompt module uses text-to-speech technology to convert preset text information into voice output;

[0020] Infrared ranging module: used to measure the distance between the user and the robot; the infrared ranging module emits infrared light through an infrared sensor, receives the reflected light, and calculates the distance based on the time difference or phase difference;

[0021] Distance determination module: Based on the data provided by the infrared ranging module, it determines whether the distance between the user and the robot is within an appropriate range; the distance determination module uses a PID control algorithm to adjust the robot's movement parameters by comparing the measured distance with a preset threshold.

[0022] Furthermore, the first analysis unit analyzes the head tilt angle of the user viewing the screen based on the large model algorithm to determine the height deviation between the robot and the user's line of sight, including:

[0023] Feature point detection: Using Haar features + Adaboost classifier, key feature points of the user's head are detected from the user's facial image. The feature points include facial features such as the corners of the eyes, corners of the mouth, and the tip of the nose.

[0024] Facial contour fitting: After detecting feature points, these points are used to fit the contour of the user's face. The fitting methods include ellipse fitting and polygon fitting;

[0025] Head pose estimation: After obtaining the facial contour or feature points, a geometric algorithm or machine learning algorithm is used to estimate the head pose, that is, to calculate the head tilt angle. The geometric algorithm uses geometric principles to calculate the head rotation angle based on the relative position relationship of facial feature points, or uses the fitting result of the facial contour to estimate the head tilt direction.

[0026] Inclination angle judgment and adjustment: The calculated inclination angle is compared with the preset appropriate inclination angle range. If the inclination angle is not within the appropriate range, the height adjustment module is triggered to adjust the height of the robot accordingly.

[0027] Furthermore, the machine learning algorithm trains a regression model, takes the detected feature points or facial contours as input, and trains the head tilt angle as output. During the testing phase, a newly captured facial image is input into the trained model to predict the head tilt angle.

[0028] Furthermore, the second analysis unit detects whether the user has a squinting expression by using an expression recognition algorithm, including:

[0029] Feature extraction: Using image processing technology, key facial features of the user are extracted from the pre-processed image. The feature information includes the shape, size, position of the eyes, and the skin texture around the eyes.

[0030] Squint detection: Based on the extracted feature information, the eye area is analyzed using machine learning algorithms or deep learning algorithms. Specific algorithms include the following:

[0031] 1) Geometric feature-based method: By calculating geometric features such as the width and height of the eye area and the degree of eye opening, it can be used to determine whether the user is squinting.

[0032] 2) Template matching-based method: A template image of the squinting state is predefined, and the extracted eye area is matched with the template. The degree of match is used to determine whether the user is squinting.

[0033] 3) Deep learning-based methods: A deep learning model is trained on a large number of squinting and non-squinting images, enabling the model to automatically learn the characteristic representation of squinting and accurately identify the squinting state in practical applications;

[0034] Judgment and feedback: Based on the results of squinting detection, if it is determined that the user has an expression such as squinting and cannot see the screen clearly, the corresponding feedback mechanism is triggered.

[0035] Furthermore, the feedback mechanism is specifically as follows: if the user squints, the distance adjustment module is triggered for the second time to make the robot move autonomously with the user, and during the movement, the user's expression changes are detected in real time until the robot moves to the closest distance within the safe observation range. If the problem is still not solved, the screen adjustment module is triggered to solve the problem by adjusting the screen brightness, contrast, font size, etc.

[0036] Furthermore, the third analysis unit detects whether the user has a frowning expression through an expression recognition algorithm, including:

[0037] Feature extraction: Using image processing technology, key facial features are extracted from the pre-processed image, especially the characteristic changes in the eyebrows, eyes, and mouth areas, including the degree of eyebrow curvature, the degree of eye opening and closing, and the shape of the mouth;

[0038] Expression classification and recognition: Based on the extracted feature information, machine learning algorithms or deep learning algorithms are used to classify and recognize expressions. Specifically, the algorithms are trained and learned from a large number of labeled facial expression images to automatically learn the characteristic representations of different expressions and accurately recognize specific expressions in practical applications.

[0039] Frown detection and judgment: Based on expression classification and recognition, special attention is paid to frowning expressions. According to the results output by the algorithm, if it is determined that the user has an expression such as frowning, the corresponding feedback mechanism is triggered.

[0040] Furthermore, the robot includes a body and an electric-controlled slide rail, a head is fixedly connected to the top of the body, and the head includes a display screen, a camera, a speaker and a microphone, a camera is provided on the front of the display screen, a speaker is provided on the front of the display screen, and a microphone is fixedly connected to the top of the display screen, legs are embedded in the bottom of the body, and the legs include a lifting cylinder and an electric-controlled roller, the bottom end of the lifting cylinder and the bottom of the body form an embedded structure, and the bottom end of the lifting cylinder is fixedly connected to the electric-controlled roller, an operating compartment is provided on the front of the body, and an operating keyboard is provided inside the operating compartment, and a disinfection component is provided on the inner top wall of the operating compartment, the electric-controlled slide rail is fixedly installed on the front of the body, and the top driven by the inner side of the electric-controlled slide rail is equipped with a transparent cover, the top of the electric-controlled slide rail is flush with the floor of the operating compartment, and the electric-controlled slide rail and the transparent cover plate form a sliding structure, and hands are fixedly connected to the left and right outer walls of the body.

[0041] Furthermore, the disinfection component includes a sprayer, a diverter pipe, an atomizing nozzle, a fixing frame and an alcohol storage box. The sprayer is arranged directly above the operating keyboard, and a diverter pipe is provided on the top of the sprayer, and atomizing nozzles are equidistantly provided on the bottom of the diverter pipe. The fixing frame is welded and fixed to the inner wall of the operating chamber, and an alcohol storage box is clamped and installed between the fixing frame and the inner wall of the operating chamber, and the alcohol storage box is connected to the sprayer through a hose.

[0042] Furthermore, the hand includes an air supply pump, an air intake pipe, an air exhaust pump and an air exhaust pipe. The air supply pump is fixedly installed on the left outer wall of the fuselage, and the air outlet of the air supply pump is fixedly connected to the air intake pipe, and one end of the air intake pipe passes through the left side wall of the fuselage and extends to the operating compartment. The air exhaust pump is fixedly installed on the right outer wall of the fuselage, and the air intake of the air exhaust pump is fixedly connected to the air exhaust pipe, and one end of the air exhaust pipe passes through the right side wall of the fuselage and extends to the operating compartment.

[0043] The present invention provides an intelligent power marketing robot based on large model technology, which has the following beneficial effects:

[0044] 1. The present invention is not only designed with an infrared ranging adjustment function, which enables the robot to move autonomously to adjust the distance and whether the distance between the user and the robot is within an appropriate range to avoid the user's viewing screen and service experience being affected by the distance being too close or too far, but also designed with a central control analysis module, which uses the first analysis unit, the second analysis unit and the third analysis unit to intelligently analyze the user's head inclination angle and facial expression when viewing the screen, thereby triggering the corresponding feedback mechanism, so that the robot can more flexibly adapt to the needs and preferences of different users and provide a more comfortable interactive experience. In particular, through large model technology and machine learning algorithms, the robot can more accurately understand the user's intentions and needs, provide a more intelligent response rhythm, thereby improving customer satisfaction and enhancing the brand image and market competitiveness of the power company.

[0045] 2. The present invention is provided with a transparent cover, which is used to shield the operation compartment when it is idle, so as to reduce the risk of dust and damage to the operation keyboard. In conjunction with the disinfection component, the operation keyboard after the user's operation can be disinfected and sterilized after each use, which significantly reduces the risk of disease transmission and improves health and safety. The air supply pump and the air exhaust pump operate simultaneously to achieve air renewal in the operation compartment, so as to discharge the moisture caused by the volatilization of alcohol in the operation compartment and the pungent smell of alcohol, further avoiding the influence of moisture on the sensitivity of the operation keyboard and the influence of alcohol smell on the user's senses. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a logic block diagram of a control system for an intelligent robot for power marketing based on large model technology according to the present invention;

[0047] Figure 2 This is a schematic diagram of the first analysis unit process of an electric power marketing intelligent robot based on large model technology according to the present invention;

[0048] Figure 3 This is a schematic diagram of the second analysis unit process of an electric power marketing intelligent robot based on large model technology according to the present invention;

[0049] Figure 4 This is a schematic diagram of the third analysis unit flow of the power marketing intelligent robot based on large model technology of the present invention;

[0050] Figure 5 This is a schematic diagram of the three-dimensional structure of an electric power marketing intelligent robot based on large model technology according to the present invention;

[0051] Figure 6 This is a schematic diagram of a disinfection component of an electric power marketing intelligent robot based on large model technology according to the present invention;

[0052] Figure 7 This invention is a power marketing intelligent robot based on large model technology Figure 6 A magnified schematic diagram of the structure in the middle.

[0053] In the figure: 1. Body; 2. Head; 201. Display screen; 202. Camera; 203. Speaker; 204. Microphone; 3. Leg; 301. Lifting cylinder; 302. Electric roller; 4. Operation compartment; 5. Operation keyboard; 6. Disinfection component; 601. Sprayer; 602. Diverter pipe; 603. Atomizing nozzle; 604. Fixing bracket; 605. Alcohol storage box; 7. Electric slide rail; 8. Transparent cover; 9. Hand; 901. Air supply pump; 902. Air intake pipe; 903. Air exhaust pump; 904. Air exhaust pipe. DETAILED DESCRIPTION

[0054] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0055] like Figures 1-4 As shown, an intelligent power marketing robot based on large model technology includes a robot equipped with a control system for controlling the robot's activities. The control system includes:

[0056] Voice prompt module: This module is used to provide voice prompts to the user to relax and look at the screen when interacting with the robot. In this embodiment, the voice prompt module utilizes text-to-speech (TTS) technology to convert preset text information into voice output. Specifically, the prompt is generated using pre-recorded voice messages or text-to-speech technology and played through speaker 203.

[0057] Infrared ranging module: used to measure the distance between the user and the robot. In this embodiment, the infrared ranging module uses an infrared sensor to emit infrared light and receive the reflected light, calculating the distance based on the time difference or phase difference.

[0058] Distance Determination Module: Based on the data provided by the infrared ranging module, it determines whether the distance between the user and the robot is within an appropriate range. In this embodiment, the distance determination module compares the measured distance with a preset threshold and uses a PID control algorithm to adjust the robot's movement parameters. Specifically, the measured distance is compared with the preset appropriate distance range. If it exceeds the range, the corresponding adjustment module is triggered.

[0059] Distance Adjustment Module: This module enables the robot to autonomously adjust the distance between the user and the robot when the distance is not within the appropriate range. This module controls the robot's movement mechanism (electrically controlled rollers 302) to move forward, backward, left, and right based on the results of the Distance Determination Module, adjusting the distance between the robot and the user.

[0060] Image acquisition module: uses a camera to capture real-time facial expression images of the user;

[0061] The central control analysis module specifically includes

[0062] The first analysis unit analyzes the tilt angle of the user's head when viewing the screen based on a large model algorithm to determine whether the robot's height is appropriate. In this embodiment, the first analysis unit uses image processing technology to analyze the tilt angle of the user's head and compares it with a preset appropriate angle. The specific processing flow is as follows:

[0063] Feature point detection: Haar features + Adaboost classifier (by constructing Haar features and training Adaboost classifier to detect facial feature points) are used to detect key feature points of the user's head from the user's facial image. Feature points include easily recognizable and stable facial features such as the corners of the eyes, corners of the mouth, and the tip of the nose. In this embodiment, the captured facial image of the user is preprocessed, including denoising, grayscale conversion, and binarization, to facilitate subsequent image analysis.

[0064] Facial contour fitting: After detecting feature points, these points are used to fit the contour of the user's face. Fitting methods include ellipse fitting (assuming the facial contour is approximately elliptical, fitting the ellipse parameters through methods such as least squares fitting) and polygon fitting (connecting the detected feature points into polygons to approximate the facial contour);

[0065] Head pose estimation: After obtaining the facial contour or feature points, a geometric algorithm or machine learning algorithm is used to estimate the head pose, that is, to calculate the head inclination angle. Among them, the geometric algorithm: based on the relative position relationship of facial feature points, such as the distance and angle from the corners of the eyes and mouth to the tip of the nose, the rotation angle of the head is calculated using geometric principles, or the fitting results of the facial contour, such as the major and minor axis directions of the ellipse, are used to estimate the tilt direction of the head. Machine learning algorithm: A regression model is trained, such as support vector regression (SVR) or random forest regression, using the detected feature points or facial contour as input and the head inclination angle as output for training. During the testing phase, the newly captured facial image is input into the trained model to predict the head inclination angle.

[0066] Inclination angle judgment and adjustment: The calculated inclination angle is compared with the preset appropriate inclination angle range. If the inclination angle is not within the appropriate range, the height adjustment module is triggered to adjust the height of the robot accordingly.

[0067] Height adjustment module: Based on the judgment result of the first analysis unit, the height of the robot is adjusted accordingly by controlling the lifting mechanism of the robot (lifting cylinder 301) so that it maintains an appropriate angle with the user's line of sight.

[0068] The second analysis unit analyzes the user's facial expression through an expression recognition algorithm to determine whether the user has an expression such as squinting and cannot see the screen clearly. In this embodiment, the second analysis unit recognizes and analyzes the user's facial expression as follows:

[0069] Feature extraction: Using image processing techniques such as edge detection and contour extraction, key facial features of the user are extracted from the pre-processed image. Feature information includes the shape, size, position of the eyes, and the skin texture around the eyes.

[0070] Squint detection: Based on the extracted feature information, the eye area is analyzed using machine learning algorithms or deep learning algorithms. Specific algorithms include the following:

[0071] 1) Geometric feature-based method: By calculating geometric features such as the width and height of the eye area and the degree of eye opening, it can be used to determine whether the user is squinting.

[0072] 2) Template matching-based method: A template image of the squinting state is predefined, and the extracted eye area is matched with the template. The degree of match is used to determine whether the user is squinting.

[0073] 3) Deep learning-based methods: Using deep learning models such as convolutional neural networks (CNNs) to train on a large number of squinting and non-squinting images, the model can automatically learn the characteristic representation of squinting and accurately identify the squinting state in practical applications;

[0074] Judgment and Feedback: Based on the results of squinting detection, if the user is judged to have an expression that makes it difficult to see the screen, the corresponding feedback mechanism is triggered. Specifically, if the user is squinting (i.e., making it difficult to see the screen), the distance adjustment module is triggered a second time, causing the robot to move autonomously. During this movement, the robot monitors changes in the user's expression in real time until the robot moves to the closest distance within the safe observation range. If the problem is still not resolved, the screen adjustment module is triggered to resolve the problem by adjusting the screen brightness, contrast, font size, etc.

[0075] Screen adjustment module: responsible for adjusting the brightness, contrast, and font size of the screen to solve the problem that users cannot see the screen clearly and improve readability; in this embodiment, the brightness, contrast, and font size displayed on the screen are adjusted specifically by controlling the software of the screen display.

[0076] The third analysis unit analyzes the user's facial expression using an expression recognition algorithm to determine whether the user has an expression of confusion or embarrassment, such as a frown. In this embodiment, the third analysis unit analyzes the user's facial expression as follows:

[0077] Feature extraction: Using image processing techniques such as edge detection, contour extraction, and texture analysis, key facial features are extracted from preprocessed images, particularly the characteristic changes in areas such as eyebrows, eyes, and mouth, including the degree of eyebrow curvature, the degree of eye opening, and the shape of the mouth.

[0078] Expression classification and recognition: Based on the extracted feature information, machine learning algorithms (such as support vector machines (SVM) and K-nearest neighbor (KNN)) or deep learning algorithms (such as convolutional neural networks (CNN) and recurrent neural networks (RNN)) are used to classify and recognize expressions. Specifically, the algorithms are trained and learned from a large number of labeled facial expression images to automatically learn the characteristic representations of different expressions and accurately recognize specific expressions in practical applications.

[0079] Frown detection and judgment: Based on expression classification and recognition, this system pays special attention to expressions of confusion and embarrassment, such as frowns. Based on the algorithm's output, if the user is judged to have a frown or other expression, a corresponding feedback mechanism is triggered.

[0080] Response adjustment module: Based on the results of frown detection, the current page on the screen is paused for a longer period of time, and a more patient and detailed explanation or suggestion is given to respond to the user's emotional state.

[0081] In summary, after the robot is awakened, it first reminds the user to look at the screen in a relaxed and comfortable state through the voice prompt module, and triggers the infrared ranging module to measure the distance. Then, the distance judgment module determines whether the distance between the user and the robot is appropriate based on the ranging results. If the distance is not within the appropriate range, the distance adjustment module is triggered, allowing the robot to move and adjust the distance autonomously to ensure that the user is within the appropriate distance range for observing the screen. After the distance is adjusted to the appropriate distance, the image acquisition module will use the camera to capture real-time images of the user's facial expressions and analyze the images using the central control analysis module. The first analysis unit uses an algorithm to analyze the user's head tilt when viewing the screen to determine whether the robot's height is appropriate, and the height adjustment module makes corresponding adjustments. At the same time, the second analysis unit uses algorithm analysis to determine whether the user is squinting (i.e., the expression of not being able to see the screen clearly). If so, the distance adjustment module is triggered for the second time to make the robot move autonomously with the user, and during the movement, the user's expression changes are detected in real time until the robot moves to the closest distance within the safe observation range. If the problem is still not solved, the screen adjustment module is triggered to solve the problem by adjusting the brightness, contrast, and font size of the screen. In addition, the third analysis unit uses algorithm analysis to determine whether the user is frowning (i.e., an expression of confusion and embarrassment). If so, the response adjustment module is triggered to trigger the delayed stop control of the current page of the screen, and generate extended explanation information to be superimposed on the stop page, such as giving more patient and detailed explanations or suggestions to respond to the user's emotional state.

[0082] like Figure 5-Figure 7 As shown, the robot includes a body 1 and an electric-controlled slide rail 7. The top of the body 1 is fixedly connected to a head 2, and the head 2 includes a display screen 201, a camera 202, a speaker 203 and a microphone 204. The front of the display screen 201 is provided with a camera 202, and the front of the display screen 201 is provided with a speaker 203, and the top of the display screen 201 is fixedly connected to the microphone 204. The bottom of the body 1 is embedded with a leg 3, and the leg 3 includes a lifting cylinder 301 and an electric-controlled roller 302. The bottom end of the lifting cylinder 301 is connected to the body 1. The bottom of the body 1 is embedded in the structure, and the bottom end of the lifting cylinder 301 is fixedly connected to the electric control roller 302. The front of the fuselage 1 is provided with an operation compartment 4, and the interior of the operation compartment 4 is provided with an operation keyboard 5, and the inner top wall of the operation compartment 4 is provided with a disinfection component 6, which includes a sprayer 601, a shunt pipe 602, an atomizing nozzle 603, a fixing frame 604 and an alcohol storage box 605. The sprayer 601 is provided just above the operation keyboard 5, and the top of the sprayer 601 is provided with a shunt pipe 602, and the bottom of the shunt pipe 602 is provided with a shunt pipe 602. An atomizing nozzle 603 is provided at a distance, a fixing frame 604 is welded and fixed to the inner wall of the operating chamber 4, and an alcohol storage tank 605 is clamped and installed between the fixing frame 604 and the inner wall of the operating chamber 4, and the alcohol storage tank 605 is connected to the sprayer 601 through a hose, an electric control slide 7 is fixedly installed on the front of the fuselage 1, and the top of the electric control slide 7 driven inside is equipped with a transparent cover 8, the top of the electric control slide 7 is flush with the bottom of the operating chamber 4, and the electric control slide 7 and the transparent cover 8 constitute a sliding structure, and the left and right outer walls of the fuselage 1 are fixedly connected The hand 9 includes an air supply pump 901, an air inlet pipe 902, an air extraction pump 903, and an air extraction pipe 904. The air supply pump 901 is fixedly mounted on the left outer wall of the fuselage 1, and the air outlet of the air supply pump 901 is fixedly connected to the air inlet pipe 902, and one end of the air inlet pipe 902 passes through the left wall of the fuselage 1 and extends to the operating compartment 4. The air extraction pump 903 is fixedly mounted on the right outer wall of the fuselage 1, and the air inlet of the air extraction pump 903 is fixedly connected to the air extraction pipe 904, and one end of the air extraction pipe 904 passes through the right wall of the fuselage 1 and extends to the operating compartment 4.

[0083] The specific operation is as follows: when the user wakes up the robot by touching the display screen 201, the transparent cover 8 moves under the control of the inner drive of the electric control slide rail 7, exposing the operation compartment 4 so that the operator can touch the operation keyboard 5 for interactive operations. After the power marketing service is completed, the transparent cover 8 returns to its original position, so that the operation compartment 4 is reclosed, and then the disinfection component 6 is automatically started. The sprayer 601 sprays the alcohol in the alcohol storage box 605 onto the operation keyboard 5 through the atomizing nozzle 603 to sterilize it. After the alcohol evaporates, the air supply pump 901 and the air exhaust pump 903 operate simultaneously to allow fresh air to enter the operation compartment 4, and discharge the moisture and the pungent smell of alcohol caused by the volatilization of alcohol in the operation compartment 4.

[0084] The embodiments of the present invention are presented for purposes of illustration and description and are not intended to be exhaustive or to limit the invention to the disclosed forms. Many modifications and variations will be apparent to those skilled in the art. The embodiments are chosen and described in order to better illustrate the principles of the invention and its practical application and to enable those skilled in the art to understand the invention and design various embodiments with various modifications as suited for specific applications.

Claims

1. An intelligent robot for power marketing based on large model technology, characterized by: include: A control system, comprising: An image acquisition module is configured to capture a user's facial expression image in real time through a camera and transmit the image to a central control analysis module; Central control analysis module, including: The first analysis unit is used to analyze the head tilt angle of the user viewing the screen based on the large model algorithm and determine the height deviation between the robot and the user's line of sight; The second analysis unit is used to detect whether the user has a squinting expression through an expression recognition algorithm; a third analysis unit, configured to detect whether the user has a frowning expression through an expression recognition algorithm; a distance adjustment module, connected to the first analysis unit, controlling the movement mechanism to perform forward and backward or left and right displacement according to the height deviation, so that the robot screen maintains a preset distance range from the user; a height adjustment module, which drives the lifting mechanism to adjust the vertical height of the robot screen according to the height deviation output by the first analysis unit, so that the robot screen maintains a suitable angle with the user's line of sight; a screen adjustment module, responsive to the squinting detection signal from the second analysis unit, to dynamically adjust the brightness, contrast, and font size of the screen; The response adjustment module triggers the delayed parking control of the current page of the screen after receiving the frown detection signal from the third analysis unit, and generates extended explanation information to be superimposed and displayed on the parking page.

2. The power marketing intelligent robot based on large model technology according to claim 1 is characterized in that: The control system further comprises: Voice prompt module: used to remind users to relax and look at the screen through voice when they interact with the robot; the voice prompt module uses text-to-speech technology to convert preset text information into voice output; Infrared ranging module: used to measure the distance between the user and the robot; the infrared ranging module emits infrared light through an infrared sensor, receives the reflected light, and calculates the distance based on the time difference or phase difference; Distance determination module: Based on the data provided by the infrared ranging module, it determines whether the distance between the user and the robot is within an appropriate range; the distance determination module uses a PID control algorithm to adjust the robot's movement parameters by comparing the measured distance with a preset threshold.

3. The power marketing intelligent robot based on large model technology according to claim 1 is characterized in that: The first analysis unit analyzes the head tilt angle of the user viewing the screen based on the large model algorithm to determine the height deviation between the robot and the user's line of sight, including: Feature point detection: Using Haar features + Adaboost classifier, key feature points of the user's head are detected from the user's facial image. The feature points include facial features such as the corners of the eyes, corners of the mouth, and the tip of the nose. Facial contour fitting: After detecting feature points, these points are used to fit the contour of the user's face. The fitting methods include ellipse fitting and polygon fitting; Head pose estimation: After obtaining the facial contour or feature points, a geometric algorithm or machine learning algorithm is used to estimate the head pose, that is, to calculate the head tilt angle. The geometric algorithm uses geometric principles to calculate the head rotation angle based on the relative position relationship of facial feature points, or uses the fitting result of the facial contour to estimate the head tilt direction. Inclination angle judgment and adjustment: The calculated inclination angle is compared with the preset appropriate inclination angle range. If the inclination angle is not within the appropriate range, the height adjustment module is triggered to adjust the height of the robot accordingly.

4. The power marketing intelligent robot based on large model technology according to claim 3 is characterized in that: The machine learning algorithm trains a regression model, takes the detected feature points or facial contours as input, and the head tilt angle as output. During the testing phase, a newly captured facial image is input into the trained model to predict the head tilt angle.

5. The power marketing intelligent robot based on large model technology according to claim 1 is characterized in that: The second analysis unit detects whether the user has a squinting expression by using an expression recognition algorithm, including: Feature extraction: Using image processing technology, key facial features of the user are extracted from the pre-processed image. The feature information includes the shape, size, position of the eyes, and the skin texture around the eyes. Squint detection: Based on the extracted feature information, the eye area is analyzed using machine learning algorithms or deep learning algorithms. Specific algorithms include the following: 1) Geometric feature-based method: By calculating geometric features such as the width and height of the eye area and the degree of eye opening, it can be used to determine whether the user is squinting. 2) Template matching-based method: A template image of the squinting state is predefined, and the extracted eye area is matched with the template. The degree of match is used to determine whether the user is squinting. 3) Deep learning-based methods: A deep learning model is trained on a large number of squinting and non-squinting images, enabling the model to automatically learn the characteristic representation of squinting and accurately identify the squinting state in practical applications; Judgment and feedback: Based on the results of squinting detection, if it is determined that the user has an expression such as squinting and cannot see the screen clearly, the corresponding feedback mechanism is triggered.

6. The power marketing intelligent robot based on large model technology according to claim 5 is characterized in that: The feedback mechanism is specifically as follows: if the user squints, the distance adjustment module is triggered for the second time to make the robot move autonomously with the user, and during the movement, the user's expression changes are detected in real time until the robot moves to the closest distance within the safe observation range. If the problem is still not solved, the screen adjustment module is triggered to solve the problem by adjusting the screen brightness, contrast, font size, etc.

7. The power marketing intelligent robot based on large model technology according to claim 1 is characterized in that: The third analysis unit detects whether the user has a frowning expression by using an expression recognition algorithm, including: Feature extraction: Using image processing technology, key facial features are extracted from the pre-processed image, especially the characteristic changes in the eyebrows, eyes, and mouth areas, including the degree of eyebrow curvature, the degree of eye opening and closing, and the shape of the mouth; Expression classification and recognition: Based on the extracted feature information, machine learning algorithms or deep learning algorithms are used to classify and recognize expressions. Specifically, the algorithms are trained and learned from a large number of labeled facial expression images to automatically learn the characteristic representations of different expressions and accurately recognize specific expressions in practical applications. Frown detection and judgment: Based on expression classification and recognition, special attention is paid to frowning expressions. According to the results output by the algorithm, if it is determined that the user has an expression such as frowning, the corresponding feedback mechanism is triggered.

8. The power marketing intelligent robot based on large model technology according to claim 1 is characterized in that: The robot includes a body and an electric-controlled slide rail. The top of the body is fixedly connected to a head, and the head includes a display screen, a camera, a speaker and a microphone. The front of the display screen is provided with a camera, and the front of the display screen is provided with a speaker, and the top of the display screen is fixedly connected to a microphone. The bottom of the body is embedded with legs, and the legs include a lifting cylinder and an electric-controlled roller. The bottom end of the lifting cylinder and the bottom of the body form an embedded structure, and the bottom end of the lifting cylinder is fixedly connected to the electric-controlled roller. An operation compartment is provided on the front of the body, and an operation keyboard is provided inside the operation compartment, and a disinfection component is provided on the inner top wall of the operation compartment. The electric-controlled slide rail is fixedly installed on the front of the body, and the top driven by the inner side of the electric-controlled slide rail is equipped with a transparent cover. The top of the electric-controlled slide rail is flush with the floor of the operation compartment, and the electric-controlled slide rail and the transparent cover plate form a sliding structure. Hands are fixedly connected to the left and right outer walls of the body.

9. The power marketing intelligent robot based on large model technology according to claim 8 is characterized in that: The disinfection component includes a sprayer, a diverter pipe, an atomizing nozzle, a fixing frame and an alcohol storage box. The sprayer is arranged directly above the operating keyboard, and a diverter pipe is provided on the top of the sprayer, and atomizing nozzles are equidistantly provided on the bottom of the diverter pipe. The fixing frame is welded and fixed to the inner wall of the operating chamber, and an alcohol storage box is clamped and installed between the fixing frame and the inner wall of the operating chamber, and the alcohol storage box is connected to the sprayer through a hose.

10. The power marketing intelligent robot based on large model technology according to claim 9 is characterized in that: The hand includes an air supply pump, an air intake pipe, an air exhaust pump and an air exhaust pipe. The air supply pump is fixedly installed on the left outer wall of the fuselage, and the air outlet of the air supply pump is fixedly connected to the air intake pipe, and one end of the air intake pipe passes through the left side wall of the fuselage and extends to the operating compartment. The air exhaust pump is fixedly installed on the right outer wall of the fuselage, and the air intake of the air exhaust pump is fixedly connected to the air exhaust pipe, and one end of the air exhaust pipe passes through the right side wall of the fuselage and extends to the operating compartment.

Images