Intelligent control method, device and equipment for AI intelligent cylindrical screen

Abstract

The invention provides an intelligent control method, device and equipment for an AI intelligent cylindrical screen, and the method comprises the steps: carrying out the feature extraction of an original image collected by the AI intelligent cylindrical screen, and obtaining a key feature vector; performing frequency analysis on the acquired environment sound signal to obtain a sound characteristic frequency spectrum; performing similarity matching on the key feature vector and data in a pre-established image feature database to obtain an image matching result; comparing and identifying the sound characteristic spectrum with pre-stored data in a sound mode library to obtain a sound mode identification result; according to the image matching result and the sound mode recognition result, a comprehensive decision is made through a preset decision algorithm, and a control instruction of the AI intelligent cylindrical screen is obtained; and performing display content switching and display parameter adjustment on the AI intelligent cylindrical screen based on the control instruction to realize intelligent control of the cylindrical screen. According to the method and the device, the defect that intelligent content display and parameter adjustment cannot be carried out according to the surrounding environment and user requirements is overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly relates to an intelligent control method, device, and equipment for an AI intelligent cylindrical screen. Background Art

[0002] The AI intelligent cylindrical screen is a new type of display device. Most traditional cylindrical screens only have basic display functions and are controlled by a set program. When displaying content, they usually display according to fixed content and parameters and cannot perform intelligent content display and parameter adjustment according to the surrounding environment and user needs.

[0003] In summary, the existing intelligent control technology for cylindrical screens has obvious deficiencies, and there is an urgent need for an intelligent control method to improve the intelligent level and application value of cylindrical screens. Summary of the Invention

[0004] The main object of the present invention is to provide an intelligent control method, device, and equipment for an AI intelligent cylindrical screen, aiming to overcome the defect that the AI intelligent cylindrical screen cannot perform intelligent content display and parameter adjustment according to the surrounding environment and user needs.

[0005] To achieve the above object, the present invention provides an intelligent control method for an AI intelligent cylindrical screen, including the following steps: Extract features from the original image collected by the AI intelligent cylindrical screen to obtain key feature vectors; Perform frequency analysis on the acquired environmental sound signal to obtain a sound feature spectrum; Perform similarity matching between the key feature vectors and the data in the pre-established image feature database to obtain an image matching result; Compare and identify the sound feature spectrum with the data in the pre-stored sound pattern library to obtain a sound pattern recognition result; Based on the image matching result and the sound pattern recognition result, perform comprehensive decision-making through a preset decision algorithm to obtain a control instruction for the AI intelligent cylindrical screen; Based on the control instruction, perform display content switching and display parameter adjustment on the AI intelligent cylindrical screen to achieve intelligent control of the AI intelligent cylindrical screen.

[0006] Further, based on the image matching result and the sound pattern recognition result, performing comprehensive decision-making through a preset decision algorithm to obtain a control instruction for the AI intelligent cylindrical screen includes: Align the feature vector sequence in the image matching result with the corresponding standard feature vector sequence in the image feature database in the time series to obtain the time warping distance of the image matching; Perform state decoding on the voice pattern recognition results based on the hidden Markov model, calculate the transition probability and emission probability of the voice pattern in different states, and obtain the state stability value of the voice pattern recognition; Use a convolutional neural network to perform feature learning on the time warping distance of image matching and the state stability value of voice pattern recognition to obtain a comprehensive feature vector; Input the comprehensive feature vector into a long short-term memory network for multi-step prediction to generate a control instruction for the image matching result and the voice pattern recognition result; the long short-term memory network is pre-trained by historical scenario data, and the historical scenario data includes images, voice features, and corresponding optimal control instructions in different times and environments.

[0007] Furthermore, according to the image matching result and the voice pattern recognition result, perform comprehensive decision-making through a preset decision algorithm to obtain a control instruction for the AI intelligent cylindrical screen, including: Perform multi-dimensional analysis on the image matching result to construct an image matching feature matrix; including analyzing the matching similarity, the position distribution of the matching area in the image, and the spatial geometric relationship of the matching features; Perform dynamic analysis based on time series for the voice pattern recognition result to obtain the change trend and fluctuation frequency of the voice pattern in the time dimension, and form a voice pattern dynamic feature vector; Input the image matching feature matrix and the voice pattern dynamic feature vector into a deep neural network model; the deep neural network model obtains a preliminary result of comprehensive decision-making through multi-layer neuron connection and non-linear activation function operation; Real-time monitor the environmental interference degree in the environment, including the light intensity change rate and the temperature change amount; When the environmental interference degree exceeds a preset threshold, correct the preliminary result of comprehensive decision-making; the correction direction and amplitude are determined according to the environmental interference degree through a pre-set mapping relationship table; Based on the corrected comprehensive decision result, generate a control instruction for the AI intelligent cylindrical screen based on a pre-established control instruction mapping rule.

[0008] Furthermore, according to the image matching result and the voice pattern recognition result, perform comprehensive decision-making through a preset decision algorithm to obtain a control instruction for the AI intelligent cylindrical screen, including: Map the key features in the image matching result and the core attributes in the voice pattern recognition result to qubit states respectively, and use the principle of quantum superposition to interweave them in the qubit state to form a quantum feature superposition state; Obtain a first control instruction based on the quantum feature superposition state; Input the image matching result and the voice pattern recognition result into the Stackelberg game model to obtain a second control instruction. In the Stackelberg game model, set the control instruction generator as the leader and the environmental factors and potential user needs as the follower. During the game process, the control instruction generator outputs a series of candidate control instruction strategies based on the image matching result and the voice pattern recognition result. According to the user demand prediction model constructed by analyzing the environmental factors and past user behaviors, simulate the responses of the follower to each candidate control instruction strategy. Through continuous iteration of the game process, find the optimal control instruction strategy as the second control instruction. Fuse the first control instruction and the second control instruction to generate a control instruction for the AI intelligent cylindrical screen.

[0009] Further, after performing display content switching and display parameter adjustment on the AI intelligent cylindrical screen based on the control instruction to achieve intelligent control of the AI intelligent cylindrical screen, it includes: Construct the mapping relationship between the image matching result, the voice pattern recognition result, and the control instruction. Construct a first array based on the control parameters in the control instruction. Obtain the device parameters of the AI intelligent cylindrical screen and construct a second array. Generate a management key based on the first array and the second array. Encrypt the mapping relationship based on the management key and send it to the management terminal for storage.

[0010] Further, generating a management key based on the first array and the second array includes: Compare the characters at the same positions in the first array and the second array one by one, and extract multiple characters from the first array and the second array according to the comparison results. Based on the extracted multiple characters, perform mutation processing on the encoded characters in the preset encoding table to obtain a mutated encoding table. Among them, the preset encoding table includes an original character column and its corresponding encoded character column. Encode the first array based on the preset encoding table to obtain a first encoding; encode the second array based on the mutated encoding table to obtain a second encoding. Extract the characteristic characters from the first encoding and the second encoding and combine them into the management key.

[0011] Further, generating a management key based on the first array and the second array includes: Form the characters in the first array into a first character sequence. Form each character in the second array into a second character sequence; splice the first character sequence and the second character sequence to obtain a spliced sequence; Generate an attribute parameter sequence based on the array attribute parameters of the first array and the second array, and perform a bitwise exclusive OR operation on the attribute parameter sequence and the spliced sequence to obtain an exclusive OR result; Perform a hash process on the exclusive OR result to obtain a hash value, and intercept the characters at the specified positions from the hash value as the management key.

[0012] The present invention also provides an intelligent control device for an AI intelligent cylindrical screen, including: An extraction unit for extracting features from the original image collected by the AI intelligent cylindrical screen to obtain a key feature vector; An analysis unit for performing frequency analysis on the acquired environmental sound signal to obtain a sound feature spectrum; A matching unit for performing similarity matching between the key feature vector and the data in the pre-established image feature database to obtain an image matching result; An identification unit for comparing and identifying the sound feature spectrum with the data in the pre-stored sound pattern library to obtain a sound pattern identification result; A decision-making unit for making a comprehensive decision through a preset decision algorithm according to the image matching result and the sound pattern identification result to obtain a control instruction for the AI intelligent cylindrical screen; A control unit for switching the display content and adjusting the display parameters of the AI intelligent cylindrical screen based on the control instruction to achieve the intelligent control of the AI intelligent cylindrical screen.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the steps of the method described in any one of the above are implemented.

[0014] The present invention also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method described in any one of the above are implemented.

[0015] The intelligent control method, device, and equipment for the AI intelligent cylindrical screen provided by the present invention include: extracting features from the original images collected by the AI intelligent cylindrical screen to obtain key feature vectors; performing frequency analysis on the acquired environmental sound signals to obtain sound feature spectra; performing similarity matching between the key feature vectors and the data in the pre-established image feature database to obtain an image matching result; comparing and identifying the sound feature spectra with the data in the pre-stored sound pattern library to obtain a sound pattern recognition result; making a comprehensive decision through a preset decision algorithm based on the image matching result and the sound pattern recognition result to obtain a control instruction for the AI intelligent cylindrical screen; and performing display content switching and display parameter adjustment on the AI intelligent cylindrical screen based on the control instruction to achieve intelligent control of the AI intelligent cylindrical screen. In the present invention, the defect that the AI intelligent cylindrical screen cannot perform intelligent content display and parameter adjustment according to the surrounding environment and user needs is overcome. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a schematic diagram of the steps of the intelligent control method for the AI intelligent cylindrical screen in an embodiment of the present invention; Figure 2 is a block diagram of the structure of the intelligent control device for the AI intelligent cylindrical screen in an embodiment of the present invention; Figure 3 is a schematic block diagram of the structure of a computer device in an embodiment of the present invention.

[0017] The implementation, functional features, and advantages of the present invention will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] Referring to Figure 1 , an embodiment of the present invention provides an intelligent control method for an AI intelligent cylindrical screen, including the following steps: Step S1: Extract features from the original images collected by the AI intelligent cylindrical screen to obtain key feature vectors; Step S2: Perform frequency analysis on the acquired environmental sound signals to obtain sound feature spectra; Step S3: Perform similarity matching between the key feature vectors and the data in the pre-established image feature database to obtain an image matching result; Step S4: Compare and identify the sound feature spectra with the data in the pre-stored sound pattern library to obtain a sound pattern recognition result; Step S5: Based on the image matching result and the voice pattern recognition result, perform comprehensive decision-making through a preset decision algorithm to obtain a control instruction for the AI intelligent cylindrical screen; Step S6: Based on the control instruction, perform display content switching and display parameter adjustment on the AI intelligent cylindrical screen to achieve intelligent control of the AI intelligent cylindrical screen.

[0020] In this embodiment, as described in step S1 above, the above-mentioned AI intelligent cylindrical screen is equipped with image acquisition devices such as cameras, which can capture the original images of the surrounding environment in real time. Feature extraction is performed based on an image feature extraction algorithm, such as the convolutional neural network (CNN) algorithm in deep learning. The CNN automatically extracts representative features from the original image by constructing multiple convolutional layers and pooling layers, including information such as object shape, color distribution, and texture details, and converts them into key feature vectors. The key feature vector is a compact and information-rich mathematical representation of the original image, laying the foundation for subsequent matching with the image feature database, reducing the amount of data processing, and improving the recognition efficiency and accuracy.

[0021] As described in step S2 above, the microphone array of the cylindrical screen is responsible for collecting the surrounding environmental sound signals. The sound signal is a continuous signal that changes over time. To deeply understand its characteristics, the fast Fourier transform (FFT), a commonly used frequency analysis method, is adopted. The FFT converts the time-domain sound signal to the frequency domain, revealing the distribution of different frequency components in the sound. By analyzing the energy intensity of the sound at different frequencies, a sound feature spectrum is generated, which can reflect the frequency characteristics of the environmental sound, such as the fundamental frequency range of human voices and the harmonic distribution of music, providing key data support for subsequent comparison and recognition with the sound pattern library, and helping to identify the sound type and potential meaning.

[0022] As described in step S3 above, the above-mentioned image feature database pre-stores a large number of image feature vectors of different scenes and different objects and their corresponding category labels. The key feature vector obtained in step S1 is matched with the data in the database using commonly used algorithms such as the Euclidean distance algorithm and the cosine similarity algorithm. These algorithms calculate the distance or similarity index between the key feature vector and each feature vector in the database to find several feature vectors that are most similar to the input key feature vector, and the corresponding category labels are the image matching results. The image matching result can initially determine the type of the currently captured image scene, such as a shopping mall promotion scene, a conference room scene, etc., providing an image-based basis for subsequent comprehensive decision-making.

[0023] As described in step S4 above, the sound pattern library stores the characteristic spectra of various typical sound patterns and their corresponding sound category information, such as human voices, background music, environmental noises, etc. Comparing the sound characteristic spectrum obtained in step S2 above with the data in the sound pattern library, classification algorithms such as template matching algorithms or support vector machines (SVMs) based on machine learning can be used. Through comparative analysis, the sound pattern category that best matches the input sound characteristic spectrum is found to obtain the sound pattern recognition result. The sound pattern recognition result supplements the information on environmental sounds. Combined with the image matching result, it can provide a more comprehensive understanding of the current environmental conditions.

[0024] As described in step S5 above, the preset decision algorithm is the core logic part of the entire intelligent control method. It integrates the image matching result, the sound pattern recognition result, and the preset rules and strategies. For example, when the image matching result shows a shopping mall promotion scene and the sound pattern recognition result detects a high-volume promotion broadcast sound, the decision algorithm generates corresponding control instructions based on this information and combines the preset rules of preferentially displaying promotion product information, increasing the display brightness and volume. The decision algorithm can be based on expert system rules or can adopt models such as decision trees and neural networks in machine learning. By learning a large amount of historical data, it continuously optimizes the decision-making logic to generate control instructions that better meet the actual needs.

[0025] As described in step S6 above, the control system of the cylindrical screen receives the control instructions generated in step S5. The control instructions specify in detail the specific content of the display content switching, such as switching to a specific product promotion video, conference materials, etc., and the adjusted values of the display parameters, such as brightness, contrast, volume size, etc. The control system precisely controls the display module and audio module of the cylindrical screen according to these instructions to achieve dynamic switching of the display content and real-time adjustment of the display parameters. Through this step, the AI intelligent cylindrical screen can intelligently display appropriate content and adjust to the best display state according to the changes in the surrounding environment and user needs, providing users with a personalized and efficient intelligent interaction experience and fully exerting its application value in various scenarios.

[0026] In one embodiment, according to the image matching result and the sound pattern recognition result, comprehensive decision-making is performed through a preset decision algorithm to obtain the control instructions for the AI intelligent cylindrical screen, including: Align the feature vector sequence in the image matching result with the corresponding standard feature vector sequence in the image feature database in the time series to obtain the time warping distance of the image matching; Perform state decoding on the sound pattern recognition result based on the hidden Markov model, calculate the transition probability and emission probability of the sound pattern in different states, and obtain the state stability value of the sound pattern recognition; Feature learning is performed on the time warping distance of image matching and the state stability value of sound pattern recognition using a convolutional neural network to obtain a comprehensive feature vector; The comprehensive feature vector is input into a long short-term memory network for multi-step prediction to generate control instructions for the image matching result and the sound pattern recognition result; the long short-term memory network is pre-trained by historical scenario data, and the historical scenario data includes images, sound features, and corresponding optimal control instructions in different times and environments.

[0027] In this embodiment, in an actual scenario, the feature vector sequence obtained after the AI intelligent cylindrical screen collects an image and extracts features may have differences in time order and rhythm from the standard feature vector sequence in the image feature database due to factors such as acquisition time and object movement. Therefore, a time series alignment method is adopted, such as the dynamic time warping (DTW) algorithm. This algorithm finds the optimal matching path between two sequences so that they can be aligned in time. For example, when the AI intelligent cylindrical screen collects an image feature vector sequence of a dynamic scene (such as a person walking), and the standard feature vector sequence in the database is a feature sequence of the same scene with a relatively fixed rhythm, the DTW algorithm can align the two in time. The calculated time warping distance reflects the similarity degree and the time difference degree between the feature vector sequence in the image matching result and the standard feature vector sequence. The smaller the distance, the more similar the two are in the time series, and the higher the accuracy of the image matching. This distance provides an important quantitative index for image in the subsequent comprehensive decision-making.

[0028] The above hidden Markov model (HMM) is a powerful statistical model for processing sequence data, especially suitable for analyzing data with time series characteristics such as sound patterns. The sound pattern recognition result is regarded as an observable sequence generated by a series of hidden states. Through the state decoding process of the HMM, it is possible to infer the hidden states that the sound pattern is in at different times. For example, for a segment of environmental sound, the HMM can decompose it into different hidden states, such as "quiet state", "human voice state", "noisy state", etc.

[0029] Calculate the transition probability between different states, that is, the likelihood of transitioning from one state to another; and the emission probability, that is, the probability of generating a specific sound observation value in a certain hidden state. Through the calculation of these probabilities, a stability evaluation of the sound pattern in each state can be obtained, that is, the state stability value. The state stability value can help determine the main state characteristics of the current sound pattern and provide reliable information for sound in the comprehensive decision-making.

[0030] Furthermore, a convolutional neural network is used to perform feature learning on the temporal warping distance of image matching and the state stability value of sound pattern recognition, obtaining a comprehensive feature vector. The above convolutional neural network (CNN) has powerful feature extraction and learning capabilities and can automatically extract important features from the input data. The temporal warping distance of image matching and the state stability value of sound pattern recognition are used as the input data of the CNN. The CNN processes these data through its multiple convolutional layers and pooling layers. The convolutional kernels in the convolutional layers slide over the data to extract local features, and the pooling layers reduce the dimensionality of the features to reduce the computational amount.

[0031] After multiple layers of processing, the CNN can learn the potential correlation features between the image matching and the sound pattern recognition results and fuse these features into a comprehensive feature vector. This comprehensive feature vector contains key information from both the image and the sound, providing a unified feature representation for subsequent generation of control instructions.

[0032] Finally, the comprehensive feature vector is input into a long short-term memory network for multi-step prediction to generate control instructions for the image matching result and the sound pattern recognition result; the long short-term memory network is pre-trained by historical scenario data, and the historical scenario data includes image and sound features and corresponding optimal control instructions under different times and environments. The long short-term memory network can effectively handle the long-term dependence problem in time series data. A large amount of historical scenario data is used to pre-train the LSTM in advance. These data include image and sound features and corresponding optimal control instructions under different times and environments. Through training, the LSTM learns the mapping relationship between the image and sound features and the control instructions.

[0033] The previously obtained comprehensive feature vector is input into the trained LSTM. The LSTM performs multi-step prediction on the input comprehensive feature vector according to the knowledge and patterns it has learned. The prediction result is the control instruction for the current image matching result and the sound pattern recognition result, such as adjusting the display brightness of the AI intelligent cylindrical screen, switching the display content, changing the audio volume, etc. The multi-step prediction ability of the LSTM makes the generation of control instructions more accurate and meet the actual scenario requirements, and can intelligently adjust the working state of the cylindrical screen according to the current image and sound conditions.

[0034] In one embodiment, according to the image matching result and the sound pattern recognition result, a comprehensive decision is made through a preset decision algorithm to obtain the control instruction of the AI intelligent cylindrical screen, including: Perform multi-dimensional analysis on the image matching result to construct an image matching feature matrix; including analyzing the matching similarity, the position distribution of the matching area in the image, and the spatial geometric relationship of the matching features; Perform dynamic analysis based on time series for the voice pattern recognition results, obtain the change trend and fluctuation frequency of the voice pattern in the time dimension, and form a dynamic feature vector of the voice pattern; Input the image matching feature matrix and the dynamic feature vector of the voice pattern into a deep neural network model; the deep neural network model obtains a preliminary comprehensive decision result through operations of multi-layer neuron connections and non-linear activation functions; Real-time monitor the environmental interference degree in the environment, including the light intensity change rate and the temperature change amount; When the environmental interference degree exceeds a preset threshold, correct the preliminary comprehensive decision result; the correction direction and amplitude are determined according to the environmental interference degree through a pre-set mapping relation table; Based on the corrected comprehensive decision result, generate a control instruction for the AI intelligent cylindrical screen according to a pre-established control instruction mapping rule.

[0035] In this embodiment, perform multi-dimensional analysis on the image matching result to construct an image matching feature matrix, deeply analyze the image matching result from multiple angles, comprehensively extract the feature information of the image matching, construct a matrix that can comprehensively reflect the image matching situation, and provide rich and accurate image feature basis for subsequent comprehensive decision-making.

[0036] The matching similarity is an important index to measure the similarity degree between the image matching result and the target image. Through specific similarity calculation methods, such as cosine similarity, structural similarity index (SSIM), etc., a specific value is obtained, and the higher this value is, the more accurate the matching is.

[0037] The position distribution of the matching area in the image is to determine the specific position of the matching area in the image and analyze its distribution law. For example, whether the matching area is concentrated in the center of the image or scattered at the edge, etc. This helps to understand the position characteristics of the key information in the image, and in some scenarios, the matching at different positions has different impacts on the control decision.

[0038] The spatial geometric relationship of the matching features is to study the geometric relationships such as the spatial position, distance, and angle between the matching features. For example, in an image for identifying multiple objects, analyze the relative positions and arrangement ways of these objects. This geometric relationship can provide the spatial layout information of the objects in the image and is of great significance for understanding the image content and scene.

[0039] Integrate the results obtained from the above multi-dimensional analysis to construct a matrix. The rows and columns of the matrix can respectively correspond to different analysis dimensions and specific feature values, so as to represent the comprehensive features of the image matching in a structured manner.

[0040] Sound is a time-varying signal. By performing time series analysis on the results of sound pattern recognition, the dynamic characteristics of sound over time are captured, forming a feature vector that can reflect the changing patterns of sound, providing dynamic information about sound for comprehensive decision-making. Observe the overall changing direction of the sound pattern over a period of time, such as whether the volume of the sound is gradually increasing, decreasing, or remaining stable, and whether the pitch is gradually rising or falling, etc. The changing trend curve can be obtained by fitting the sound characteristic values (such as volume, frequency, etc.). Calculate the fluctuation frequency of the sound pattern over time, that is, the number of changes in the sound characteristic values within a certain period of time. For example, in a piece of music, the tempo can be reflected by the fluctuation frequency. By methods such as spectrum analysis of the sound signal, the fluctuation frequency information can be obtained. Quantify the changing trend and fluctuation frequency and other characteristics of the sound pattern in the time dimension and combine them into a vector. This vector can comprehensively describe the dynamic characteristics of the sound pattern, facilitating subsequent comprehensive processing with image features.

[0041] Utilize the powerful feature learning and non-linear mapping capabilities of deep neural networks to deeply fuse and analyze the features of images and sounds, thereby obtaining a preliminary decision result that comprehensively considers image and sound information. A deep neural network consists of multiple neuron layers, and the neurons in each layer are connected to the neurons in the previous layer through weights. The image matching feature matrix and the dynamic feature vector of the sound pattern are used as inputs and first enter the input layer, and then through layer-by-layer transmission, feature transformation and combination are performed in the hidden layer. Each neuron performs weighted summation according to its input and weights and is processed through a non-linear activation function, and the processing result is passed to the next layer.

[0042] Non-linear activation functions (such as ReLU, Sigmoid, etc.) introduce non-linearity into the neural network, enabling the network to learn complex patterns and relationships. Through the operations of non-linear activation functions, the neural network can perform non-linear transformations on the input features, thereby better capturing the potential connections between image and sound features.

[0043] After calculations and non-linear transformations by multiple layers of neurons, a preliminary result of comprehensive decision-making is finally obtained in the output layer. This result is obtained based on the comprehensive analysis of image and sound features and provides a basis for the generation of subsequent control instructions.

[0044] Environmental factors may interfere with image acquisition and sound recognition, thereby affecting the accuracy of comprehensive decision-making. By real-time monitoring the degree of environmental interference, the impact of environmental changes on the system can be detected in a timely manner, providing a basis for subsequent decision-making correction.

[0045] Rapid changes in light intensity may lead to a decline in the quality of image acquisition, such as overexposure or underexposure, thus affecting the accuracy of image matching. Monitoring the rate of change of light intensity can promptly detect unstable factors in the light environment and provide a reference for subsequent decision-making adjustments.

[0046] Changes in temperature may affect sound propagation and device performance. For example, a sharp change in temperature may cause changes in the speed and quality of sound propagation, affecting the results of sound pattern recognition. Monitoring the amount of temperature change can help determine the degree of interference of the ambient temperature on the system.

[0047] Considering that environmental interference may lead to inaccurate preliminary results of comprehensive decision-making, when the degree of environmental interference exceeds a certain threshold, it is necessary to correct the preliminary results to improve the accuracy and reliability of the decision-making.

[0048] Specifically, compare the rate of change of light intensity and the amount of temperature change monitored in real time with the preset thresholds. If the thresholds are exceeded, it is considered that the degree of environmental interference is large and correction is required. Preset a mapping relation table, which records the correction directions and amplitudes corresponding to different degrees of environmental interference. According to the current degree of environmental interference, look up the corresponding correction information in the mapping relation table. According to the found correction directions and amplitudes, adjust the preliminary results of the comprehensive decision-making. For example, if the rate of change of light intensity is too large and causes inaccurate image matching, it may be necessary to reduce the weight of the image matching result or make certain adjustments to the image matching feature matrix.

[0049] Finally, convert the corrected comprehensive decision-making result into specific control instructions to achieve intelligent control of the AI intelligent cylindrical screen.

[0050] The pre-established control instruction mapping rules define the correspondence between the comprehensive decision-making result and the control instructions. For example, if the comprehensive decision-making result indicates that the current environment is a lively event scene, and both the image and sound features show that the display effect needs to be enhanced, then according to the mapping rules, control instructions such as increasing the screen brightness, switching to dynamic display content, and increasing the volume may be generated. According to the corrected comprehensive decision-making result, look up and convert according to the control instruction mapping rules, and finally generate control instructions suitable for the current scene for the AI intelligent cylindrical screen. These control instructions will be sent to the control system of the cylindrical screen to achieve adjustment of the screen display content and parameters.

[0051] In one embodiment, according to the image matching result and the sound pattern recognition result, through a preset decision-making algorithm for comprehensive decision-making, obtain the control instructions for the AI intelligent cylindrical screen, including: Map the key features in the image matching result and the core attributes in the sound pattern recognition result to qubit states respectively. Utilize the principle of quantum superposition to interweave them in the qubit states, forming a quantum feature superposition state. Obtain a first control instruction based on the quantum feature superposition state. Input the image matching result and the sound pattern recognition result into the Stackelberg game model to obtain a second control instruction. In the Stackelberg game model, set the control instruction generator as the leader and the environmental factors and potential user needs as the follower. During the game process, the control instruction generator outputs a series of candidate control instruction strategies based on the image matching result and the sound pattern recognition result. According to the user demand prediction model constructed by analyzing the environmental factors and past user behavior data, simulate the follower's response to each candidate control instruction strategy. By continuously iterating the game process, find the optimal control instruction strategy as the second control instruction. Fuse the first control instruction and the second control instruction to generate a control instruction for the AI intelligent cylindrical screen.

[0052] In this embodiment, the image matching result contains numerous features, and key features are selected from them, such as representative feature information like the shape, color, texture, etc. of the objects in the image. The core attributes of the sound pattern recognition result may be the frequency, pitch, volume change pattern, etc. of the sound. A qubit is the basic information unit in quantum computing. Different from classical bits that can only be in the state of 0 or 1, a qubit can be in the state of 0, 1, or any superposition state of 0 and 1. Mapping the feature information of the image and sound to qubit states is to transform the traditional feature information into a quantum-level representation. For example, through specific encoding rules, each key feature or core attribute corresponds to a specific state of one or more qubits.

[0053] The principle of quantum superposition allows qubits to be in a superposition of multiple states simultaneously. After mapping the features of the image and sound to qubit states, these qubit states will interweave according to the principle of quantum superposition. Just like multiple waves superposing in space, different qubit states influence and fuse with each other, forming a quantum feature superposition state that contains the comprehensive feature information of the image and sound. This superposition state can express multiple possible feature combinations simultaneously, providing a richer and more complex information basis for subsequent decision-making.

[0054] The quantum feature superposition state contains the comprehensive feature information of the image and sound. Through quantum computing techniques such as quantum measurement, meaningful information is extracted from this superposition state. The measurement process will cause the quantum superposition state to collapse to a definite state, thereby obtaining specific feature values or feature combinations.

[0055] According to pre-set rules and algorithms, the extracted feature information is converted into control instructions. For example, if the measured feature information indicates that the current image is a lively event scene and the sound is in a lively rhythm, then the rules may correspondingly generate control instructions such as increasing the screen brightness, switching to dynamic display content, and increasing the volume. This is the first control instruction. Such control instructions generated based on the quantum feature superposition state can make full use of the parallelism and superposition characteristics of quantum computing, consider more possible feature combination situations, and make the instructions more comprehensive and forward-looking.

[0056] In the Stackelberg game model, there are two roles: the leader and the follower. Here, the control instruction generator is set as the leader, and the environmental factors (such as light intensity, temperature, humidity, etc.) and the potential needs of users are set as the follower. The leader acts first, that is, the control instruction generator outputs a series of candidate control instruction strategies based on the image matching result and the sound pattern recognition result. These candidate strategies are pre-designed according to different scenarios and feature combinations. For example, different combinations of image and sound features correspond to different strategies for switching screen display content, parameter adjustment, etc.

[0057] To simulate the response of the follower (environmental factors and potential user needs) to each candidate control instruction strategy, a user demand prediction model is constructed. The above user demand prediction model is established based on the analysis of environmental factors and past user behavior data. Through this model, it can be predicted how the environmental factors will change and whether the potential needs of users can be met under different candidate control instruction strategies. For example, if the candidate strategy is to increase the screen brightness, the model can predict the current light intensity and the operation habits of users in similar light environments in the past to determine whether users will accept this brightness adjustment.

[0058] Through continuous game iterations, the leader adjusts the candidate control instruction strategies according to the simulated response of the follower. Each iteration will evaluate the advantages and disadvantages of the current strategy, gradually eliminate the unsatisfactory strategies, and finally find an optimal control instruction strategy. This strategy is the second control instruction. This process is similar to the repeated game between the leader and the follower to reach an optimal balance state, so that the control instructions not only consider the current image and sound features but also adapt to environmental factors and meet the potential needs of users.

[0059] The first control instruction is generated based on the quantum feature superposition state, making full use of the advantages of quantum computing and considering more possible combinations of features. The second control instruction is obtained through the Stackelberg game model, comprehensively considering environmental factors and potential user needs. By fusing the two, their respective advantages can be combined to make the generated control instruction more comprehensive, accurate, and intelligent. In this embodiment, a weighted fusion method can be adopted to assign different weights to the first control instruction and the second control instruction according to different application scenarios and requirements. For example, in scenarios with high requirements for real-time performance, the weight of the first control instruction can be appropriately increased; in scenarios with high requirements for user experience and environmental adaptability, the weight of the second control instruction can be increased. Through weighted calculation, the parameters of the two control instructions are fused, and finally, a control instruction suitable for the AI intelligent cylindrical screen is generated to achieve intelligent control of the screen.

[0060] In one embodiment, after performing display content switching and display parameter adjustment on the AI intelligent cylindrical screen based on the control instruction to achieve intelligent control of the AI intelligent cylindrical screen, the following steps are included: Construct the mapping relationship between the image matching result, the sound mode recognition result, and the control instruction; Construct a first array based on the control parameters in the control instruction; Obtain the device parameters of the AI intelligent cylindrical screen and construct a second array; Generate a management key based on the first array and the second array; Encrypt the mapping relationship based on the management key and send it to the management terminal for storage.

[0061] In this embodiment, a corresponding relationship is established among the image matching result, the sound mode recognition result, and the finally generated control instruction for subsequent data traceability, analysis, and verification and optimization of the control instruction generation logic. After achieving intelligent control of the AI intelligent cylindrical screen, record the image matching result (such as the category of the matched image, matching similarity, etc.), the sound mode recognition result (such as the type of sound, characteristic spectrum, etc.), and the corresponding control instruction (including specific information on display content switching and display parameter adjustment) obtained in the current scenario. Then, organize and store this information in a specific format to form a clear mapping relationship. For example, in the form of a database table, the image matching result and the sound mode recognition result can be used as columns of the table, and the control instruction can be used as another column. Each row represents a specific scenario instance and its corresponding information.

[0062] Structurally process the specific parameters in the control instruction for subsequent further calculations and analyses, and also provide a data basis for generating the management key. The control instruction contains various parameters for switching the display content and adjusting the display parameters of the AI intelligent cylindrical screen, such as display brightness, contrast, volume, type of display content, etc. The system extracts these control parameters from the control instruction and arranges them into an array (which can be a one-dimensional array or a multi-dimensional array, depending on the nature and quantity of the parameters) in a certain order. For example, take the brightness parameter as the first element of the array, the contrast parameter as the second element, and so on, to form the first array.

[0063] Obtain the device parameter information of the cylindrical screen itself, construct a device parameter array corresponding to the control parameters for comprehensive analysis and generating the management key, and also help monitor and manage the status and performance of the device. The AI intelligent cylindrical screen has a series of device parameters, such as the screen resolution, refresh rate, pixel density, hardware model, etc. The system obtains the above parameters through the hardware interface of the cylindrical screen or the relevant device management module. Then, also arrange these device parameters into an array in accordance with the preset rules and order, that is, the second array. For example, first arrange the resolution parameters (number of horizontal pixels, number of vertical pixels), then arrange the refresh rate parameter, etc., to form an ordered set of device parameters.

[0064] Generate a unique and highly secure management key to encrypt the mapping relationship between the image matching result, sound pattern recognition result and the control instruction, protect the security and privacy of the data, and prevent the data from being illegally obtained and tampered with. Multiple methods can be used to generate the management key. For example, first preprocess the data in the first array and the second array, such as performing normalization, hashing operations, etc., to reduce data redundancy and improve data security. Then, fuse the processed data, which can be simple concatenation or combination through a specific algorithm, to obtain a new data set. Next, apply an encryption algorithm (such as the symmetric encryption algorithm AES, the asymmetric encryption algorithm RSA, etc.) to this new data set to generate a fixed-length string, and this string is the management key. The generation process of the management key should be reversible so that the encrypted data can be decrypted when needed.

[0065] Ensure the security of the mapping relationships among image matching results, sound pattern recognition results, and control instructions during transmission and storage. Only the management terminal with the correct management key can decrypt and access this data to prevent data leakage and malicious tampering. Use the generated management key to encrypt the previously constructed mapping relationship data (which can be records in a database table or other forms of data storage). The encrypted mapping relationship data becomes a string of ciphertext. Then, send the encrypted mapping relationship data to the management terminal through a secure communication channel. After receiving the data, the management terminal stores it in a secure storage device for subsequent data query, analysis, and system maintenance operations. When accessing this data, the management terminal uses the corresponding management key to decrypt the ciphertext and restore the original mapping relationship data.

[0066] In one embodiment, generating the management key based on the first array and the second array includes: Compare the characters at the same positions in the first array and the second array one by one, and extract multiple characters from the first array and the second array according to the comparison results. Based on the extracted multiple characters, perform mutation processing on the encoded characters in a preset encoding table to obtain a mutated encoding table. The preset encoding table includes an original character column and its corresponding encoded character column in one-to-one mapping. Encode the first array based on the preset encoding table to obtain a first encoding. Encode the second array based on the mutated encoding table to obtain a second encoding. Extract the characteristic characters from the first encoding and the second encoding and combine them into the management key.

[0067] In this embodiment, by comparing the characters at the same positions in the first array and the second array, the characters with specific characteristics are found, providing the basic data for generating the management key later. These characters contain the association information between the control instruction parameters and the device parameters. Specifically, the first array is constructed based on the control parameters in the control instruction, and the second array is constructed from the device parameters of the AI intelligent cylindrical screen. Compare the same positions of the two arrays in order. The comparison rule can be set according to specific requirements. If the comparison results meet a specific relationship (such as equality, the difference is within a certain range, etc.), the corresponding characters are extracted. After comparing one by one, multiple qualified characters are extracted from the two arrays to form a new character set.

[0068] Mutate a preset coding table using the characters extracted from two arrays to make the coding table unique, thereby increasing the security and uniqueness of the subsequently generated management key. The preset coding table is a table containing an original character column and a coded character column that maps to it one by one. For example, the original character column can be common letters, numbers, etc., and the coded character column is the corresponding characters generated according to specific coding rules.

[0069] The mutation process modifies the coded characters in the coding table according to a certain algorithm based on the multiple extracted characters. For example, according to information such as the ASCII code value of the extracted characters and the position of the characters, operations such as replacing and shifting the coded characters at specific positions in the coding table can be performed. After the mutation process, a new coding table, that is, a mutated coding table, is obtained, which is different from the original coding table and incorporates the characteristic information of the first array and the second array.

[0070] Use the preset coding table and the mutated coding table to encode the first array and the second array respectively, convert the original parameter information into a coded form, further hide the information and prepare for generating the management key.

[0071] For the first array, each character in it is traversed, the coded character corresponding to this character is found in the preset coding table, and then these coded characters are combined in sequence to form the first code. For the second array, each character in it is also traversed, but the mutated coding table is used to find the corresponding coded characters, and these coded characters are combined to obtain the second code. Due to the uniqueness of the mutated coding table, the second code and the first code reflect the differences and associations between the two arrays to a certain extent.

[0072] Extract the key characteristic characters from the first code and the second code, and combine them into a management key. This management key contains the comprehensive characteristics of the control instruction parameters and the device parameters, and can be used to encrypt the mapping relationship between the image matching result, the sound pattern recognition result and the control instruction.

[0073] The extraction of characteristic characters can be carried out according to specific rules. For example, select the characters at specific positions in the code, the characters with higher occurrence frequencies, etc. These rules can be designed according to the requirements of security and uniqueness. After extracting the characteristic characters from the first code and the second code respectively, combine them in a preset order to form a string with a fixed length. The above string is the finally generated management key. The generation process of the management key combines the information of the control instruction and the device parameters, and increases its security and uniqueness through coding and feature extraction operations.

[0074] In one embodiment, generating a management key based on the first array and the second array includes: Construct a first character sequence from each character in the first array; Each character in the second array forms a second character sequence; the first character sequence and the second character sequence are concatenated to obtain a concatenated sequence; Based on the array attribute parameters of the first array and the second array, an attribute parameter sequence is generated, and the attribute parameter sequence and the concatenated sequence are subjected to a bitwise exclusive OR operation to obtain an exclusive OR result; The exclusive OR result is subjected to a hash process to obtain a hash value, and the characters at specified positions are intercepted from the hash value as the management key.

[0075] In this embodiment, the first array constructed based on the control parameters in the control instruction is converted into the form of a character sequence, which is convenient for subsequent operations and processing with other sequences, while maintaining the integrity and orderliness of the control parameter information. The first array is composed of the characters related to the control parameters in the control instruction. The system connects these characters in sequence according to the arrangement order of the characters in the first array to form a continuous character sequence, that is, the first character sequence. For example, if the first array contains the characters 'a', 'b', 'c', then the first character sequence is 'abc'.

[0076] The second array constructed based on the AI intelligent cylindrical screen device parameters is converted into a character sequence and concatenated with the first character sequence, integrating the control parameter information and the device parameter information in one sequence, providing a richer data basis for generating the management key later. Similar to the processing method of the first array, the system connects these characters in sequence according to the arrangement order of the characters in the second array to form a second character sequence. Then, the first character sequence and the second character sequence are concatenated, that is, the second character sequence is connected behind the first character sequence to obtain a new and longer character sequence, that is, the concatenated sequence.

[0077] The attribute parameter sequence is generated by using the attribute parameters of the first array and the second array, and the concatenated sequence is further processed through a bitwise exclusive OR operation to increase the complexity and randomness of the data, making the generated management key more secure. The array attribute parameters may include but are not limited to the length of the array, the element type, the value range of the elements, etc. According to these attribute parameters, an attribute parameter sequence is generated according to certain rules. For example, assuming that the length of the first array is 3 and the element type is character type, and the length of the second array is 3 and the element type is numeric type, the system may generate the attribute parameter sequence '3c3n' (where 'c' represents character type and 'n' represents numeric type).

[0078] The bitwise XOR operation is to perform an XOR operation on the characters at the corresponding positions in the concatenated sequence and the attribute parameter sequence (represented in binary form at the computer's underlying level). If the binary bits corresponding to two characters are the same, the result is 0; if they are different, the result is 1. Convert the results of these XOR operations back to character form to obtain the XOR result. Through hashing, convert the XOR result into a hash value of a fixed length to further compress and obfuscate the data, and then intercept the characters at the specified positions from the hash value to generate the management key, ensuring the uniqueness and security of the management key.

[0079] Use a specific hashing algorithm (such as SHA-256, MD5, etc.) to perform hashing on the XOR result, converting the XOR result into a hash value of a fixed length. Hashing algorithms have the properties of being one-way and unique, that is, different inputs will result in different hash values, and it is impossible to reverse-derive the original input from the hash value. From the obtained hash value, intercept the characters at the specified positions according to the pre-set rules. For example, intercept the characters from the 10th to the 20th position of the hash value, and combine these intercepted characters to form the final management key.

[0080] In one embodiment, generating the management key based on the first array and the second array specifically includes the following steps: Convert the control parameters in the first array, such as display brightness and color saturation, into complex number form to form a set of control parameter complex numbers. For example, use the display brightness value as the real part of the complex number and the color saturation value as the imaginary part to construct multiple complex numbers to form a set of control parameter complex numbers.

[0081] For the device parameters in the second array, perform a discrete Fourier transform (DFT) to form a frequency-domain feature vector of the device parameters. Consider data such as the number of horizontal and vertical pixels of the resolution and the refresh rate as discrete time-series signals, and obtain the frequency-domain data through DFT transformation to form a frequency-domain feature vector of the device parameters; Based on the set of control parameter complex numbers, use the polar coordinate representation of complex numbers to convert each complex number into a polar radius and a polar angle. Use the polar radius as the radius and the polar angle as the angle to plot scatter points in the polar coordinate system, and connect the scatter points to form a control parameter curve; For the frequency-domain feature vector of the device parameters, use the frequency as the abscissa and the corresponding frequency-domain amplitude as the ordinate to plot a frequency-spectrum diagram of the device parameters; Sample the control parameter curve to obtain the polar coordinate values of several key sampling points; perform binarization processing on the frequency-spectrum diagram of the device parameters to obtain a frequency-spectrum binary matrix. For example, set the area where the frequency-spectrum amplitude is higher than the threshold to 1 and the area lower than the threshold to 0 to obtain a frequency-spectrum binary matrix.

[0082] The polar coordinate values of the sampling points are fused with the spectral binary matrix to obtain new values. For example, the radial value is multiplied by the element at the corresponding position in the spectral binary matrix, and the angular value is XORed with the matrix element to obtain a series of new values.

[0083] Perform principal component analysis (PCA) on the above new values, extract the main components, and decompose the main components using singular value decomposition (SVD) to obtain a set of singular values.

[0084] Arrange the singular values in a preset order and convert them into binary form. After encryption hashing operation, a fixed-length string is generated, which is the management key. The above management key integrates the graphical curve features of control parameters and device parameters, and is comprehensively processed through unique mathematical properties, with extremely high security and uniqueness.

[0085] Refer to Figure 2 , in another embodiment of the present invention, an intelligent control device for an AI intelligent cylindrical screen is further provided, including: An extraction unit for extracting features from the original image collected by the AI intelligent cylindrical screen to obtain key feature vectors; An analysis unit for performing frequency analysis on the acquired environmental sound signal to obtain a sound feature spectrum; A matching unit for performing similarity matching between the key feature vectors and the data in the pre-established image feature database to obtain an image matching result; An identification unit for comparing and identifying the sound feature spectrum with the data in the pre-stored sound pattern library to obtain a sound pattern recognition result; A decision-making unit for making a comprehensive decision through a preset decision algorithm according to the image matching result and the sound pattern recognition result to obtain a control instruction for the AI intelligent cylindrical screen; A control unit for switching the display content and adjusting the display parameters of the AI intelligent cylindrical screen based on the control instruction to achieve intelligent control of the AI intelligent cylindrical screen.

[0086] In this embodiment, for the specific implementation of each unit in the above device embodiment, please refer to that described in the above method embodiment, and details will not be elaborated here.

[0087] Refer to Figure 3 , in an embodiment of the present invention, a computer device is further provided. The computer device can be a server, and its internal structure can be as Figure 3As shown in the figure. The computer device includes a processor, a memory, a display screen, an input device, a network interface, and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements the above method.

[0088] Those skilled in the art can understand that Figure 3 the structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0089] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0090] In summary, for the intelligent control method, device, and equipment of the AI intelligent cylindrical screen provided in the embodiments of the present invention, it includes: extracting features from the original image collected by the AI intelligent cylindrical screen to obtain a key feature vector; performing frequency analysis on the acquired environmental sound signal to obtain a sound feature spectrum; performing similarity matching between the key feature vector and the data in the pre-established image feature database to obtain an image matching result; comparing and identifying the sound feature spectrum with the data in the pre-stored sound pattern library to obtain a sound pattern recognition result; based on the image matching result and the sound pattern recognition result, performing comprehensive decision-making through a preset decision algorithm to obtain a control instruction for the AI intelligent cylindrical screen; and based on the control instruction, performing display content switching and display parameter adjustment on the AI intelligent cylindrical screen to achieve intelligent control of the AI intelligent cylindrical screen. In the present invention, the defect that the AI intelligent cylindrical screen cannot perform intelligent content display and parameter adjustment according to the surrounding environment and user needs is overcome.

[0091] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided by the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0092] It should be noted that in this article, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, device, article or method comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, device, article or method. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of additional identical elements in the process, device, article or method comprising the element.

[0093] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention accordingly. Any equivalent structure or equivalent process transformation made by using the specification and drawings of the present invention, or directly or indirectly applied in other related technical fields, shall be similarly included in the patent protection scope of the present invention.

Claims

1. An intelligent control method for an AI intelligent cylindrical screen, characterized in that: The following steps are involved: Perform feature extraction on the original image collected by the AI ​​smart cylindrical screen to obtain key feature vectors; Perform frequency analysis on the acquired environmental sound signal to obtain a sound characteristic spectrum; Performing similarity matching between the key feature vector and data in a pre-established image feature database to obtain an image matching result; Comparing and identifying the sound feature spectrum with data in a pre-stored sound pattern library to obtain a sound pattern recognition result; According to the image matching results and the sound pattern recognition results, a comprehensive decision is made through a preset decision algorithm to obtain the control instructions of the AI ​​smart cylindrical screen; Based on the control instructions, the display content of the AI ​​smart cylindrical screen is switched and the display parameters are adjusted to realize intelligent control of the AI ​​smart cylindrical screen.

2. The intelligent control method of the AI ​​intelligent cylindrical screen according to claim 1 is characterized in that: According to the image matching results and the sound pattern recognition results, a comprehensive decision is made through a preset decision algorithm to obtain the control instructions of the AI ​​smart cylindrical screen, including: Align the feature vector sequence in the image matching result with the corresponding standard feature vector sequence in the image feature database in time series to obtain the time-regularized distance of the image matching; Based on the hidden Markov model, the state decoding of the sound pattern recognition results is performed, the transition probability and emission probability of the sound pattern in different states are calculated, and the state stability value of the sound pattern recognition is obtained; The convolutional neural network is used to learn the features of the time-warping distance of image matching and the state stability value of sound pattern recognition to obtain a comprehensive feature vector; The comprehensive feature vector is input into a long short-term memory network to perform multi-step prediction and generate control instructions for the image matching results and sound pattern recognition results; the long short-term memory network is pre-trained by historical scene data, and the historical scene data contains images and sound features at different times and environments and corresponding optimal control instructions.

3. The intelligent control method of the AI ​​intelligent cylindrical screen according to claim 1 is characterized in that: According to the image matching results and the sound pattern recognition results, a comprehensive decision is made through a preset decision algorithm to obtain the control instructions of the AI ​​smart cylindrical screen, including: Performing multi-dimensional analysis on the image matching results to construct an image matching feature matrix, including analyzing matching similarity, position distribution of matching areas in the image, and spatial geometric relationships of matching features; Based on the sound pattern recognition results, dynamic analysis based on time series is performed to obtain the change trend and fluctuation frequency of the sound pattern in the time dimension, and form the dynamic feature vector of the sound pattern; Inputting the image matching feature matrix and the sound pattern dynamic feature vector into a deep neural network model; the deep neural network model obtains a preliminary comprehensive decision result through multi-layer neuron connection and nonlinear activation function operation; Real-time monitoring of the degree of environmental interference in the environment, including the rate of change of light intensity and the amount of temperature change; When the degree of environmental interference exceeds a preset threshold, the preliminary result of the comprehensive decision is corrected; the direction and magnitude of the correction are determined according to the degree of environmental interference through a preset mapping relationship table; According to the revised comprehensive decision results, the control instructions of the AI ​​smart cylindrical screen are generated based on the pre-established control instruction mapping rules.

4. The intelligent control method of the AI ​​intelligent cylindrical screen according to claim 1 is characterized in that: According to the image matching results and the sound pattern recognition results, a comprehensive decision is made through a preset decision algorithm to obtain the control instructions of the AI ​​smart cylindrical screen, including: The key features in the image matching results and the core attributes in the sound pattern recognition results are mapped to quantum bit states respectively, and the quantum bit states are intertwined to form a quantum feature superposition state by using the quantum superposition principle; Obtaining a first control instruction based on the quantum characteristic superposition state; The image matching result and the sound pattern recognition result are input into a Stackelberg game model to obtain a second control instruction; in the Stackelberg game model, the control instruction generator is set as a leader, and the environmental factors and user potential needs are followers; during the game process, the control instruction generator outputs a series of candidate control instruction strategies based on the image matching result and the sound pattern recognition result; according to the user demand prediction model constructed by analyzing the environmental factors and past user behavior data, the response of the follower to each candidate control instruction strategy is simulated; through continuous iteration of the game process, the optimal control instruction strategy is found as the second control instruction; The first control instruction and the second control instruction are integrated to generate the control instruction of the AI ​​smart cylindrical screen.

5. The intelligent control method of the AI ​​intelligent cylindrical screen according to claim 1 is characterized in that: After switching the display content and adjusting the display parameters of the AI ​​smart cylindrical screen based on the control instruction to realize the intelligent control of the AI ​​smart cylindrical screen, the method includes: Constructing a mapping relationship between the image matching result, the sound pattern recognition result and the control instruction; constructing a first array based on the control parameters in the control instruction; Obtaining device parameters of the AI ​​smart cylindrical screen and constructing a second array; generating a management key based on the first array and the second array; The mapping relationship is encrypted based on the management key and then sent to the management terminal for storage.

6. The intelligent control method of the AI ​​intelligent cylindrical screen according to claim 5 is characterized in that: Generating a management key based on the first array and the second array includes: Comparing characters at the same position in the first array and the second array one by one, and extracting a plurality of characters from the first array and the second array according to the comparison result; Based on the extracted multiple characters, the coded characters in the preset coding table are mutated to obtain a mutated coding table; wherein the preset coding table includes an original character column and a one-to-one mapped coded character column; Encoding the first array based on the preset encoding table to obtain a first code; encoding the second array based on the variant encoding table to obtain a second code; Feature characters are extracted from the first code and the second code, and the feature characters are combined into the management key.

7. The intelligent control method of the AI ​​intelligent cylindrical screen according to claim 5 is characterized in that: Generating a management key based on the first array and the second array includes: Each character in the first array is used to form a first character sequence; The characters in the second array are used to form a second character sequence; the first character sequence is concatenated with the second character sequence to obtain a concatenated sequence; Based on the array attribute parameters of the first array and the second array, an attribute parameter sequence is generated, and a bitwise XOR operation is performed on the attribute parameter sequence and the concatenated sequence to obtain an XOR result; The XOR result is hashed to obtain a hash value, and characters at a specified position are intercepted from the hash value as the management key.

8. An intelligent control device for an AI intelligent cylindrical screen, characterized in that: include: An extraction unit is used to extract features from the original image collected by the AI ​​smart cylindrical screen to obtain key feature vectors; An analysis unit, used to perform frequency analysis on the acquired environmental sound signal to obtain a sound characteristic spectrum; A matching unit, used to perform similarity matching between the key feature vector and data in a pre-established image feature database to obtain an image matching result; A recognition unit, used for comparing and recognizing the sound characteristic spectrum with data in a pre-stored sound pattern library to obtain a sound pattern recognition result; A decision-making unit, used to make a comprehensive decision through a preset decision algorithm according to the image matching result and the sound pattern recognition result, and obtain a control instruction for the AI ​​smart cylindrical screen; A control unit is used to switch the display content and adjust the display parameters of the AI ​​smart cylindrical screen based on the control instructions, so as to realize intelligent control of the AI ​​smart cylindrical screen.

9. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

Images