An intelligent AI holographic projection method and system based on artificial intelligence

Abstract

The present invention proposes an intelligent AI holographic projection method and system based on artificial intelligence. The method includes: identifying users within the projection area, extracting features from the two-dimensional images and three-dimensional spatial information of the users, and integrating them into user feature vectors; performing precise hierarchical classification on the users according to the user feature vectors in combination with two-dimensional image features, distance features of the users in the three-dimensional space, and the line-of-sight directions of the users; generating and distributing personalized content for the users according to the user demand matrix to meet the specific needs of the users in the holographic projection environment; designing a view angle adaptive projection matrix and a content synchronization mechanism to perform personalized holographic real-time projection of the personalized content. The present invention provides a new holographic projection system solution, which can provide personalized display and adaptive projection effects according to the needs of individual users in a multi-user scenario, has high flexibility and adaptability, and ensures that each user can obtain good visual effects and personalized experiences at any position.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence, and particularly relates to an intelligent AI holographic projection method and system based on artificial intelligence. Background Art

[0002] In modern fields such as exhibition displays, education and training, and medical imaging, holographic projection technology is widely used because it can create three-dimensional and visual spatial images. This technology presents three-dimensional content to users with a visual stereoscopic effect through hierarchical processing and projection of images, creating a highly immersive experience. However, with the diversification of user needs, especially in multi-user interaction scenarios in public places, traditional holographic projection technology is facing severe challenges. Existing holographic projection systems mainly perform well in single-user or small-scale user experiences, and can provide good visual effects under a single content or single perspective. However, in multi-user scenarios (such as museums, exhibitions, etc.), the limitations of traditional holographic projection systems become increasingly obvious.

[0003] First of all, existing holographic projection systems generally lack the ability to identify multiple users and provide personalized responses, making it difficult to provide personalized interaction experiences for multiple users in the same scenario. When multiple users are in the same projection area, the system often can only display the same content, which is difficult to meet the needs of different users. For example, some users may want to see more detailed content, while others only want to quickly obtain an overview.

[0004] Since the display content of holographic projection cannot change dynamically according to different user needs, it is easy to cause content conflicts and information interference among users, affecting the experience effect in multi-user scenarios. Secondly, in terms of perspective adaptation, traditional holographic systems lack the ability to adapt to multiple perspectives. Different users stand at different positions in the projection area, and the viewing effects may be very different. Some users may see clear holographic images, while others can only see blurred edges or distorted images due to angle limitations. This single-perspective projection method is limited in a multi-user environment and it is difficult to achieve the best visual experience for each user.

[0005] Finally, existing holographic systems mostly have single-level interactions in user interaction, and it is difficult to dynamically adjust the projection content according to the behaviors and interests of different users, resulting in a single and rigid experience. For example, although some systems can recognize users' gestures or positions, they lack hierarchical management in multi-user scenarios and cannot effectively distinguish the needs and interaction levels of different users, resulting in the content display being difficult to meet the requirements of multi-user interaction. Summary of the Invention

[0006] The objective of the present invention is to propose an intelligent AI holographic projection method and system based on artificial intelligence. Through technological innovations such as multi-user recognition, personalized content generation, and perspective adaptive adjustment, it realizes the provision of personalized immersive experiences for multiple users within the same projection area.

[0007] To achieve the above objective, in the first aspect of the present invention, an intelligent AI holographic projection method based on artificial intelligence is provided. The method includes:

[0008] S1. Identify users within the projection area, extract features from the two-dimensional images and three-dimensional spatial information of the users, and integrate them into user feature vectors; the user feature vectors include two-dimensional image features, distance features of the users in the three-dimensional space, and the line-of-sight directions of the users.

[0009] S2. Based on the user feature vectors combined with two-dimensional image features, distance features of the users in the three-dimensional space, and the line-of-sight directions of the users, conduct precise hierarchical classification of the users, specifically including:

[0010] First, conduct preliminary stratification:

[0011] Set an interaction distance threshold D th , if the distance feature is greater than the interaction distance threshold D th then classify user i into the long-distance viewing layer, and if the distance feature is less than the interaction distance threshold D th then classify user i into the close-distance interaction layer;

[0012] Calculate the interest score of each user based on the two-dimensional image features, distance features of the users in the three-dimensional space, and the line-of-sight directions of the users. At the same time, design an interest score threshold I th , if the interest score is greater than the interest score threshold I th then mark this user as the high-interest interaction layer and allocate detailed content to it. If the interest score is less than the interest score threshold I th then mark this user as the low-interest interaction layer, and the content display will be mainly in an overview;

[0013] Integrate the results of the preliminary stratification and interest score classification to construct a multi-level user demand matrix M to support the hierarchical management of personalized content display;

[0014] S3. Generate and distribute personalized content for users according to the user demand matrix M to meet the specific needs of users in the holographic projection environment, specifically including:

[0015] Generate corresponding content matrices C for different user layers according to the user demand matrix M;

[0016] Design a distribution frequency matrix L to control the content update frequency of the content matrices C corresponding to different user layers;

[0017] S4. Design a perspective adaptive projection matrix and a content synchronization mechanism to perform personalized holographic real-time projection of personalized content.

[0018] Furthermore, the extraction of features from the user's two-dimensional image and three-dimensional space information specifically includes:

[0019] Arrange cameras and distance sensors in the projection area to collect the two-dimensional image and three-dimensional space information of each user, which is expressed as follows:

[0020] U i ={P i , D i , V i}

[0021] Among them, U i represents the detection feature vector of each user, P i represents the two-dimensional image feature of the user, which is obtained in real time by an image detection model; D i represents the distance feature of the user in the three-dimensional space, that is, the relative position with the projection system, which is measured by a distance sensor; V i represents the line-of-sight direction of the user, which is determined by detecting the angle and orientation of the face through the camera.

[0022] Furthermore, the image detection model is a convolutional neural network. The two-dimensional image feature processes the two-dimensional image feature P i of the user to identify the facial features of the user in the projection area, and then calculates the line-of-sight direction V i of the user. The pose information θ i of the image is analyzed through a multi-layer neural network, which is expressed as follows:

[0023] V i =sin(θ i ) + cos(θ i )

[0024] Among them, θ i represents the facial orientation angle of the user, which is obtained by using the three-dimensional point-to-point pose estimation method. The line-of-sight direction V i is calculated through a combination formula of sine and cosine to ensure the calculation accuracy at different angles and distances;

[0025] After detecting the user's facial information and line-of-sight direction, combine the distance sensor data to calculate the distance feature D i of the user in the three-dimensional space in real time, that is, the distance from the user to the projection device, which is expressed as follows:

[0026]

[0027] Among them, (xi , y i , z i ) represents the spatial coordinates of the user, (x c , y c , z c ) represents the spatial coordinates of the projection center.

[0028] Furthermore, the interest score is calculated by combining the user's fixation time, facial orientation angle, and distance feature in three-dimensional space, as shown below:

[0029] I i = w T T i + w θ |θ i | + λ·f(D i )

[0030] where, I i represents the user's interest score, which is used to judge the user's personalized content needs; T i is the time the user fixates on the content, reflecting the user's attention to the content; θ i is the user's facial orientation angle; D i is the distance feature in three-dimensional space. The distance regularization term is controlled by the attenuation function of the distance feature D i in three-dimensional space, and α is the distance attenuation coefficient; w , w T , and λ are weight coefficients, which are used to adjust the influence of T θ , θ i , and D i on the final interest score, and their values are verified through experiments; i The user demand matrix M is expressed as follows:

[0031]

[0032]

[0033] where, U 近，高 is the set of user features with high interest at close range; U 近，低 is the set of user features with low interest at close range; U 远，高 is the set of user features with high interest at long range; U 远，低 is the set of user features with low interest at long range

[0034] Furthermore, most of the attenuation functions reduce the interest of users at long range as the distance increases, so as to reasonably control resource allocation in a multi-user scenario.

[0035] Furthermore, generating a corresponding content matrix C for different user layers according to the user demand matrix M specifically includes:

[0036] For each user layer U ij , the element C of the content generation matrix ij is calculated by the content generation function φ:

[0037] C ij = φ(M ij )

[0038] where C ij represents the personalized content display of the corresponding layer U ij ; φ is the content generation function, which inputs the demand matrix M ij and returns the customized content;

[0039] The design of the distribution frequency matrix L controls the content update frequency of the content matrix C corresponding to different user layers, as shown below:

[0040] L ij = δ·(1 + η·I ij )

[0041] where L ij represents the content update frequency of layer U ij ; δ is the basic distribution frequency, set as the global frame number update standard; η is the interest gain coefficient, which controls the influence of interest on the content update frequency.

[0042] Furthermore, a regularization term for the interest adjustment item is designed in the content matrix C to adaptively adjust the content details, as shown below:

[0043] C ij = (R·exp(-β·D ij ))·(1 + γ·I ij )

[0044] where R is the basic resolution of the content, which determines the clarity of the content; D ij is the average distance of the user set U ij , which is used as a regularization term to control the attenuation of content details; β is the distance weight coefficient, which controls the influence of distance on the content display effect; I ij is the average interest of the users in this layer, which is used to increase the detail level of users with high interest; γ is the amplification coefficient of interest.

[0045] Furthermore, the perspective adaptive projection matrix is shown below:

[0046]

[0047] where T ijDenoted as the user layer U ij The generated projection matrix that controls the brightness, contrast, and transparency of the content; θ ij is the angle by which the user i's viewing angle deviates from the center; α ij and β ij are parameters for adjusting the brightness and contrast, and adjust the display effect in combination with the user's distance and line of sight direction; D ij is the user's distance, and λ is the distance regularization coefficient, used to ensure that users at different distances obtain balanced brightness and contrast;

[0048] The content synchronization mechanism is expressed as follows:

[0049] S ij = C ij · T ij · (1 + η · I ij )

[0050] where S ij represents the synchronization content signal provided by the user layer U ij ; C ij is the personalized content that defines the basic display of the content; T ij is the view angle adaptive projection matrix used to adjust the brightness and contrast of the content; I ij is the user's interest score, which reflects the user's demand intensity for the current content; η is the interest enhancement coefficient used to increase the content display density for high-interest users.

[0051] Furthermore, in S4, an error correction mechanism is also introduced. By detecting the deviation between T ij and the ideal projection parameters in real time, automatic correction is triggered, which is expressed as follows:

[0052]

[0053] where E ij represents the view angle error between the current projection matrix and the ideal projection parameters ; is the projection parameter of the current frame; is the ideal projection value calculated from the user's real-time view angle data;

[0054] When the error E ij exceeds the threshold ∈, an update will be automatically triggered to recalculate the projection matrix T ij , and it will be synchronized to S ij in real time to ensure the consistency of the content from the user's perspective.

[0055] In the second aspect of the present invention, an intelligent AI holographic projection system based on artificial intelligence is provided. The system includes:

[0056] A user projection data acquisition module, which is used to identify users within the projection area, extract features from the two-dimensional images and three-dimensional space information of the users, and integrate them into user feature vectors; the user feature vectors include two-dimensional image features, distance features of the users in the three-dimensional space, and the line-of-sight directions of the users.

[0057] A user classification module, which is used to accurately classify users into different levels according to the user feature vectors in combination with two-dimensional image features, distance features of the users in the three-dimensional space, and the line-of-sight directions of the users. Specifically, it includes:

[0058] First, perform preliminary stratification:

[0059] Set an interaction distance threshold D th , if the distance feature is greater than the interaction distance threshold D th then classify user i into the long-distance viewing layer, if the distance feature is less than the interaction distance threshold D th then classify user i into the close-distance interaction layer;

[0060] Calculate the interest score of each user according to the two-dimensional image features, distance features of the users in the three-dimensional space, and the line-of-sight directions of the users. At the same time, design an interest score threshold I th , if the interest score is greater than the interest score threshold I th then mark this user as the high-interest interaction layer and allocate detailed content to it. If the interest score is less than the interest score threshold I th then mark this user as the low-interest interaction layer, and the content display is mainly in overview;

[0061] Integrate the results of preliminary stratification and interest score classification to construct a multi-level user demand matrix M, which is used to support the hierarchical management of personalized content display;

[0062] A user demand generation module, which is used to generate and distribute personalized content for users according to the user demand matrix M to meet the specific needs of users in the holographic projection environment. Specifically, it includes:

[0063] Generate a corresponding content matrix C for different user layers according to the user demand matrix M;

[0064] Design a distribution frequency matrix L to control the content update frequency of the content matrix C corresponding to different user layers;

[0065] A holographic projection synchronization module, which is used to design a perspective adaptive projection matrix and a content synchronization mechanism to perform personalized holographic real-time projection of personalized content.

[0066] The beneficial technical effects of the present invention are at least as follows:

[0067] The invention adopts a multi-layer AI architecture, and through cameras and sensors, it monitors and identifies the position, behavior, and line of sight direction of each user in real time. Based on the AI analysis results, the system assigns users to different interaction layers (such as a close-range interaction layer and a long-range viewing layer), thus realizing precise hierarchical management of multiple users. This hierarchical mechanism can not only meet the personalized needs of each user but also effectively avoid the problems of content conflict and visual interference among users.

[0068] The AI content generation module of the present invention generates personalized holographic content in real time according to the behavior, interests, and interaction levels of users. For example, for users in the close range, the system can provide content displays with rich details, such as enlarged views or hierarchical anatomical information, while long-range users receive simplified overview content. This innovation not only enhances the diversity of the user experience but also effectively avoids the problem that multi-user content in traditional holographic projection cannot be personalized, ensuring that each user obtains targeted content displays.

[0069] Based on the position and line of sight direction of the user, through an AI-driven projection adjustment mechanism, the present invention automatically optimizes the angle, brightness, and contrast of the projection content, enabling each user to enjoy a clear and distortion-free visual experience no matter from which direction they view. This perspective optimization mechanism overcomes the visual limitations brought by single-perspective displays in traditional systems and is particularly suitable for application in complex multi-user scenarios, ensuring that the viewing experience of each user reaches the best effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] The present invention is further described with the aid of the drawings, but the embodiments in the drawings do not constitute any limitation to the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained according to the following drawings.

[0071] Figure 1 It is a flowchart of a method for intelligent AI holographic projection based on artificial intelligence according to the present invention.

[0072] Figure 2 It is a framework diagram of a system for intelligent AI holographic projection based on artificial intelligence according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0073] The embodiments of the present invention are described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements with the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present invention and should not be construed as a limitation to the present invention.

[0074] Such as Figure 1As shown in the figure, an intelligent AI holographic projection method based on artificial intelligence provided by an embodiment of the present invention includes:

[0075] S1. Identify the user within the projection area, extract features from the user's two-dimensional image and three-dimensional space information, and integrate them into a user feature vector; the user feature vector includes two-dimensional image features, the distance feature of the user in the three-dimensional space, and the line-of-sight direction of the user.

[0076] S2. Based on the user feature vector combined with the two-dimensional image features, the distance feature of the user in the three-dimensional space, and the line-of-sight direction of the user, perform precise hierarchical classification on the user, specifically including:

[0077] First, perform preliminary stratification:

[0078] Set the interaction distance threshold D th , if the distance feature is greater than the interaction distance threshold D th then classify user i as the long-distance viewing layer, if the distance feature is less than the interaction distance threshold D th then classify user i as the close-distance interaction layer; that is, if D i ≤D th , then classify user i as the close-distance interaction layer; if D i >D th , then classify user i as the long-distance viewing layer. The logic of this distance classification is to ensure that users closer to the projection area can obtain a higher-precision content display while reducing the visual interference of long-distance users. After classification, a preliminary user set U 近 ={U i |D i ≤D th} and U 远 ={U i |D i >D th} are formed;

[0079] Calculate the interest score of each user according to the two-dimensional image features, the distance feature of the user in the three-dimensional space, and the line-of-sight direction of the user. At the same time, design the interest score threshold I th , if the interest score is greater than the interest score threshold I th then mark this user as the high-interest interaction layer and allocate detailed content to it. If the interest score is less than the interest score threshold I th then mark this user as the low-interest interaction layer, and the content display is mainly in overview;

[0080] Integrate the results of the preliminary stratification and the interest score classification to construct a multi-level user demand matrix M for supporting the hierarchical management of personalized content display;

[0081] S3. Generate and distribute personalized content for users according to the user demand matrix M to meet the specific needs of users in the holographic projection environment, specifically including:

[0082] Generate corresponding content matrices C for different user layers according to the user demand matrix M;

[0083] Design a distribution frequency matrix L to control the content update frequency of the content matrices C corresponding to different user layers;

[0084] S4. Design a perspective adaptive projection matrix and a content synchronization mechanism to perform personalized holographic real-time projection of the personalized content.

[0085] Furthermore, the extraction of features from the two-dimensional images and three-dimensional space information of users specifically includes:

[0086] Arrange cameras and distance sensors in the projection area to collect the two-dimensional images and three-dimensional space information of each user, expressed as follows:

[0087] U i ={P i ,D i ,V i}

[0088] where U i represents the detection feature vector of each user, P i represents the two-dimensional image features of the user (such as face, gesture), which are obtained in real time by the image detection model; D i represents the distance feature of the user in the three-dimensional space, that is, the relative position with the projection system, which is measured by the distance sensor; V i represents the line-of-sight direction of the user, which is determined by detecting the angle and orientation of the face through the camera.

[0089] Furthermore, the image detection model is a convolutional neural network, and the two-dimensional image feature processes the two-dimensional image feature P i of the user to identify the face features of the user in the projection area, and then calculates the line-of-sight direction V j , and analyzes the pose information θ i of the image through a multi-layer neural network, expressed as follows:

[0090] V i =sin(θ i )+cos(θ i )

[0091] where θ i represents the face orientation angle of the user, which is obtained by using the three-dimensional point-to-point pose estimation method, and calculates the line-of-sight direction V i through a combination formula of sine and cosine to ensure the calculation accuracy at different angles and distances;

[0092] After detecting the user's facial information and line-of-sight direction, the distance feature D of the user in the three-dimensional space is calculated in real time by combining the data of the distance sensor i , that is, the distance from the user to the projection device, which is expressed as follows:

[0093]

[0094] where (x i , y i , z i ) represents the spatial coordinates of the user, and (x c , y c , z c ) represents the spatial coordinates of the projection center.

[0095] Furthermore, the above features are integrated to generate the user feature vector U i ={P i , D i , V i}, where each feature has been calibrated according to the three-dimensional space information. This feature vector characterizes the real-time spatial position, posture and line-of-sight direction of each user, and ensures clear and non-overlapping feature output for each user in a multi-user scenario through the weighted average method, avoiding confusion between multiple users. Finally, all U i vectors are used as the final output of this step, providing basic data for subsequent layering and content generation.

[0096] Furthermore, the interest score is calculated by combining the user's fixation time, face orientation angle and distance feature in the three-dimensional space, which is expressed as follows:

[0097] I i =w T T i +w θ |θ i |+λ·f(D i )

[0098] where I i represents the user's interest score, which is used to further judge the user's personalized content needs; T i is the time (unit: second) that the user fixes on this content, reflecting the user's attention to the content; θ i is the rotation angle (unit: degree) of the user's head, obtained through the posture recognition module, measuring the content viewing angle of the user; D i is the distance feature of the user, and the distance regularization term is controlled through the attenuation function i of D , and α is the distance attenuation coefficient; w T , w θAnd λ are weight coefficients used to adjust T i , θ i and D i 's influence on the final interest degree score. The values are verified through experiments and set to w T = 0.6, w e = 0.3, λ = 0.1.

[0099] Among them, the design of f(D i ) innovatively introduces the user distance as a regularization term, making the interest degree of distant users gradually decrease as the distance increases, so as to reasonably control resource allocation in a multi-user scenario. After completing the calculation of I i , compare I i with the set threshold I th :

[0100] If I i ≥I th , then mark this user as a high-interest interaction layer, and the system will allocate more detailed content to it.

[0101] If I i <I th , then mark it as a low-interest interaction layer, and the content display is mainly overview-based to reduce system resource occupancy.

[0102] Integrate the results of the above two classification steps to construct a multi-level user demand matrix M for supporting the hierarchical management of personalized content display. Among them, the user demand matrix M is expressed as follows:

[0103]

[0104] Among them, U 近，高 is the set of user characteristics of close-range and high-interest users; U 近，低 is the set of user characteristics of close-range but low-interest users; U 远，高 is the set of user characteristics of distant but high-interest users; U 远，低 is the set of user characteristics of distant and low-interest users;

[0105] The output multi-level matrix M contains complete user stratification information, which is used as the basic input for the next step of personalized content generation. The matrix M provides support for the system to refine content allocation in a multi-user scenario, ensuring that the content generation in subsequent steps can accurately match the personalized needs of each level of users.

[0106] Furthermore, most of the attenuation functions make the interest degree of distant users decrease as the distance increases, so as to reasonably control resource allocation in a multi-user scenario.

[0107] Furthermore, generating the corresponding content matrix C for different user layers according to the user demand matrix M specifically includes:

[0108] For each user layer U ij , the element C of the content generation matrix ij is calculated by the content generation function φ:

[0109] C ij = φ(M ij )

[0110] where C ij represents the personalized content display of the corresponding layer U ij ; φ is the content generation function, which takes the demand matrix M ij as input and returns the customized content;

[0111] where the content generation function φ adopts a parameterized template generation strategy to adjust the content details, resolution, and display brightness according to the different needs of user layers. For example, for the U 近，高 level, the content generation is more inclined to high-resolution, high-brightness, and high-detail displays to meet the needs of users with high interest at close range; while for the U 远，低 layer, low-brightness and low-resolution content are generated to ensure the reasonable utilization of system resources.

[0112] After generating the content matrix C, to ensure the smoothness and real-time nature of content display in a multi-user scenario, the system designs a distribution frequency matrix L. The matrix L controls the content update frequency of users in different layers to ensure that high-interest users receive higher-frequency updates, while low-interest users reduce the content refresh frequency, thereby optimizing resource utilization.

[0113] Furthermore, the design of the distribution frequency matrix L controls the content update frequency of the content matrix C corresponding to different user layers, which is expressed as follows:

[0114] L ij = δ·(1 + η·I ij )

[0115] where L ij represents the content update frequency of layer U ij (unit: frames per second); δ is the basic distribution frequency, which is set as the global frame update standard; η is the interest gain coefficient, which controls the influence of interest on the content update frequency.

[0116] where, through the above formula, L ij and I ijIt is directly proportional, so that the content update frequency of high-interest users is higher, ensuring the timeliness of personalized content; at the same time, the content update frequency of low-interest users is reduced, reducing the system load. This design ensures the efficient distribution and synchronization of content in the holographic projection scenario.

[0117] Furthermore, a regularization term for interest adjustment is designed in the content matrix C to adaptively represent content details, as follows:

[0118] C ij =(R·exp(-β·D ij ))·(1+γ·I ij )

[0119] Where R is the basic resolution of the content, which determines the clarity of the content; D ij is the average distance of the user set U ij , serving as a regularization term to control the attenuation of content details; β is the distance weight coefficient, controlling the impact of distance on the content display effect; I ij is the average interest of the users at this layer, used to increase the detail level of high-interest users; γ is the amplification coefficient of interest.

[0120] Among them, through the attenuation term exp(-β·D ij ) of D ij , the system reduces the detail level of content for users at a long distance; at the same time, the interest adjustment term (1+γ·I ij ) makes the content display of high-interest users more detailed. This generation formula innovatively combines the distance and interest regularization terms, matching the detail level of the content with the interaction needs of users, thus effectively solving the challenge of unbalanced content display in multi-user scenarios.

[0121] Furthermore, based on the position and line-of-sight information V i in the user's feature vector U i ={Pi, D i , V i , the present invention constructs an innovative perspective adaptive projection matrix T ij . T ij can automatically adjust the projection parameters according to the user's perspective and distance to ensure the display effect for multiple users at different perspectives. The perspective adaptive projection matrix is expressed as follows:

[0122]

[0123] Where T ij represents the projection matrix generated for the user layer U ij , controlling the brightness, contrast, and transparency of the content; θ ij is the angle by which the perspective of user i deviates from the center; α ij and βij Adjust the parameters of brightness and contrast, and adjust the display effect in combination with the user's distance and line of sight direction; D ij is the user distance, λ is the distance regularization coefficient, used to ensure that users at different distances obtain balanced brightness and contrast;

[0124] The construction of this projection matrix performs a combined correction of cosine and sine according to the perspective deviation of different users, and at the same time adds an inverse distance adjustment term Ensure that users at different distances obtain balanced brightness and contrast.

[0125] The content synchronization mechanism is described as follows:

[0126] S ij = C ij ·T ij ·(1 + η·I ij )

[0127] Among them, S ij represents the synchronization content signal provided by the user layer U ij ; C ij is personalized content, which defines the basic display of the content; T ij is the perspective adaptive projection matrix, used to adjust the brightness and contrast of the content; I ij is the user's interest score, which reflects the user's demand intensity for the current content; η is the interest enhancement coefficient, used to increase the content display density of high-interest users.

[0128] Among them, in this formula, the distribution signal S ij is controlled by T ij and I ij doubly, not only realizing personalized content distribution, but also ensuring the smoothness of content update when the user moves.

[0129] Furthermore, in S4, an error correction mechanism is also introduced. By detecting the deviation between T ij and the ideal projection parameters in real time, automatic correction is triggered, which is described as follows:

[0130]

[0131] Among them, E ij represents the perspective error between the current projection matrix and the ideal projection parameters ; is the projection parameter of the current frame; is the ideal projection value, calculated from the user's real-time perspective data;

[0132] When the error E ijWhen it exceeds the threshold ∈, the update will be automatically triggered to recalculate the projection matrix T ij and synchronize it to S in real time ij , ensuring the consistency of the content from the user's perspective. Error correction ensures the consistent experience of multiple users in the dynamic perspective.

[0133] As Figure 2 shown, the embodiment of the present invention also provides an intelligent AI holographic projection system based on artificial intelligence. The system includes:

[0134] A user projection data acquisition module 3011, configured to identify users within the projection area, extract features from the two-dimensional images and three-dimensional spatial information of the users, and integrate them into a user feature vector; the user feature vector includes two-dimensional image features, distance features of the user in the three-dimensional space, and the line-of-sight direction of the user;

[0135] A user classification module 3012, configured to perform precise hierarchical classification on users according to the user feature vector in combination with two-dimensional image features, distance features of the user in the three-dimensional space, and the line-of-sight direction of the user. Specifically, it includes:

[0136] First, perform preliminary stratification:

[0137] Set an interaction distance threshold D th , if the distance feature is greater than the interaction distance threshold D th then user i is classified into the long-distance viewing layer, and if the distance feature is less than the interaction distance threshold D th then user i is classified into the short-distance interaction layer;

[0138] Calculate the interest score of each user according to the two-dimensional image features, distance features of the user in the three-dimensional space, and the line-of-sight direction of the user. At the same time, design an interest score threshold I th , if the interest score is greater than the interest score threshold I th then mark the user as a high-interest interaction layer and allocate detailed content to it. If the interest score is less than the interest score threshold I th then mark the user as a low-interest interaction layer, and the content display is mainly in overview;

[0139] Integrate the results of the preliminary stratification and interest score classification to construct a multi-level user demand matrix M for supporting the hierarchical management of personalized content display;

[0140] A user demand generation module 3013, configured to generate and distribute personalized content for users according to the user demand matrix M to meet the specific needs of users in the holographic projection environment. Specifically, it includes:

[0141] Generate a corresponding content matrix C for different user layers according to the user demand matrix M;

[0142] Design a distribution frequency matrix L to control the content update frequency of the content matrix C corresponding to different user layers.

[0143] A holographic projection synchronization module 3014 is used to design a perspective adaptive projection matrix and a content synchronization mechanism to perform personalized holographic real-time projection of personalized content.

[0144] Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0145] In several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection may be through some interfaces, and the indirect coupling or communication connection of the devices or units may be in electrical, mechanical, or other forms.

[0146] If the above functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs that can store program codes.

[0147] Although the embodiments of the present invention have been shown and described, those skilled in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirits of the present invention. The scope of the present invention is defined by the claims and their equivalents.

Claims

1. An intelligent AI holographic projection method based on artificial intelligence, characterized in that, The method includes: S1. Identify the users within the projection area, extract features from the 2D images and 3D spatial information of the users, and integrate them into a user feature vector; the user feature vector includes 2D image features, distance features of the users in the 3D space, and the line-of-sight directions of the users. S2. Based on the user feature vector, combined with 2D image features, distance features of the users in the 3D space, and the line-of-sight directions of the users, perform precise hierarchical classification on the users, specifically including: First, conduct preliminary stratification: Set the interaction distance threshold D th , if the distance feature is greater than the interaction distance threshold D th then classify user i into the long-distance viewing layer, and if the distance feature is less than the interaction distance threshold D th then classify user i into the short-distance interaction layer; Calculate the interest score for each user based on the two-dimensional image features, the distance features of the user in the three-dimensional space, and the user's line of sight direction. At the same time, design an interest score threshold I th , if the interest score is greater than the interest score threshold I th then mark the user as a high-interest interaction layer and allocate detailed content to them. If the interest score is less than the interest score threshold I th then mark the user as a low-interest interaction layer, and the content display will be mainly in overview; Integrate the results of the preliminary stratification and the interest score classification to construct a multi-level user demand matrix M for supporting the hierarchical management of personalized content display. S3. Generate and distribute personalized content for the users according to the user demand matrix M to meet the specific needs of the users in the holographic projection environment, specifically including: Generate a corresponding content matrix C for different user layers according to the user demand matrix M. Design a distribution frequency matrix L to control the content update frequency of the content matrix C corresponding to different user layers. S4. Design a perspective adaptive projection matrix and a content synchronization mechanism to perform personalized holographic real-time projection on the personalized content.

2. The intelligent AI holographic projection method based on artificial intelligence according to claim 1, characterized in that, The extraction of features from the 2D images and 3D spatial information of the users specifically includes: Arrange cameras and distance sensors in the projection area to collect the 2D images and 3D spatial information of each user, which is expressed as follows: U i = {P i , D i , V i} Among them, U i represents the detection feature vector of each user, and P i represents the two-dimensional image feature of the user, which is obtained in real time by the image detection model; D i represents the distance feature of the user in the three-dimensional space, that is, the relative position with the projection system, which is measured by the distance sensor; V i represents the line-of-sight direction of the user, which is determined by detecting the angle and orientation of the face through the camera.

3. The intelligent AI holographic projection method based on artificial intelligence according to claim 2, wherein The image detection model is a convolutional neural network, and the two-dimensional image feature processes the two-dimensional image feature P of the user i Identify the facial features of the user in the projection area, and then calculate the user's line-of-sight direction V i , analyze the pose information θ of the image through a multi-layer neural network i , which is expressed as follows: V i = sin(θ i ) + cos(θ i ) Among them, θ i represents the facial orientation angle of the user, which is obtained by using the three-dimensional point-to-point pose estimation method, and the line-of-sight direction V i is calculated through a combined formula of sine and cosine to ensure the calculation accuracy at different angles and distances; After detecting the user's facial information and line of sight direction, the distance feature D of the user in the three-dimensional space is calculated in real time in combination with the distance sensor data i , that is, the distance from the user to the projection device, which is expressed as follows: Among them, (x i , y i , z i ) represents the spatial coordinates of the user, and (x c , y c , z c ) represents the spatial coordinates of the projection center.

4. An intelligent AI holographic projection method based on artificial intelligence according to claim 1, characterized in that, The interest score is calculated by combining the user's fixation time, face orientation angle, and distance features in the 3D space, which is expressed as follows: I i = w T T i + w θ |θ i | + λ·f(D i ) Among them, I i represents the user's interest degree score, which is used to judge the user's personalized content needs; T i is the time that the user gazes at the content, reflecting the user's attention to the content; θ i is the user's facial orientation angle; D i is the distance feature in the three-dimensional space. Through the attenuation function i of the distance feature D in the three-dimensional space, the distance regularization term control is introduced, and α is the distance attenuation coefficient; w T , w θ and λ are weight coefficients, which are used to adjust T i , θ i and D i 's influence on the final interest degree score, and the values are verified through experiments; The user demand matrix M is expressed as follows: Among them, U 近,高 is a set of user characteristics with a short distance and high interest; U 近,低 is a set of user characteristics with a short distance but low interest; U 远,高 is a set of user characteristics with a long distance but high interest; U 远,低 is a set of user characteristics with a long distance and low interest.

5. A method for intelligent AI holographic projection based on artificial intelligence according to claim 4, characterized in that, Most of the attenuation functions reduce the interest of remote users as the distance increases, so as to reasonably control resource allocation in a multi-user scenario.

6. An intelligent AI holographic projection method based on artificial intelligence according to claim 1, characterized in that, The generation of a corresponding content matrix C for different user layers according to the user demand matrix M specifically includes: For each user layer U ij , an element C of the content generation matrix ij is calculated by the content generation function φ: C ij = φ(M ij ) Among them, C ij represents the personalized content display of the corresponding layer U ij ; φ is a content generation function that takes the demand matrix M ij as input and returns customized content; The design of the distribution frequency matrix L to control the content update frequency of the content matrix C corresponding to different user layers is expressed as follows: L ij = δ·(1 + η·I ij ) Among them, L ij represents the content update frequency of layer U ij ; δ is the basic distribution frequency, which is set as the global frame number update standard; η is the interest gain coefficient, which controls the influence of interest on the content update frequency.

7. An intelligent AI holographic projection method based on artificial intelligence according to claim 6, characterized in that Design an interest adjustment term regularization term in the content matrix C to adaptively adjust the content details, which is expressed as follows: C ij = (R·exp(-β·D ij ))·(1 + γ·I ij ) Among them, R is the basic resolution of the content, which determines the clarity of the content; D ij is the average distance of the user set U ij , which is used as a regularization term to control the attenuation of content details; β is the distance weight coefficient, which controls the influence of distance on the content display effect; I ij is the average interest of users at this layer, which is used to increase the detail level of users with high interest; γ is the amplification coefficient of interest.

8. An intelligent AI holographic projection method based on artificial intelligence according to claim 1, characterized in that, The perspective adaptive projection matrix is expressed as follows: Among them, T ij represents the projection matrix generated by the user layer U ij to control the brightness, contrast, and transparency of the content; θ ij is the angle by which the user i's viewing angle deviates from the center; α ij and β ij are parameters for adjusting the brightness and contrast, and adjust the display effect in combination with the user's distance and line of sight direction; D ij is the user's distance, and λ is the distance regularization coefficient, which is used to ensure that users at different distances obtain balanced brightness and contrast; The content synchronization mechanism is expressed as follows: S ij = C ij · T ij · (1 + η · I ij ) Among them, S ij represents the synchronization content signal provided by the user layer U ij ; C ij is personalized content, defining the basic display of the content; T ij is the perspective adaptive projection matrix, used to adjust the brightness and contrast of the content; I ij is the user's interest score, reflecting the user's demand intensity for the current content; η is the interest enhancement coefficient, used to increase the content display density of high-interest users.

9. An intelligent AI holographic projection method based on artificial intelligence according to claim 8, characterized in that, In S4, an error correction mechanism is also introduced to trigger automatic correction by detecting the deviation between the real-time detected T ij and the ideal projection parameters, as shown below: Among them, E ij represents the viewing angle error between the current projection matrix and the ideal projection parameters; is the projection parameter of the current frame; is the ideal projection value, calculated from the user's real-time viewing angle data; When the error E ij exceeds the threshold ∈, an update will be automatically triggered to recalculate the projection matrix T ij and synchronize it to S in real time ij to ensure the consistency of the content from the user's perspective.

10. An intelligent AI holographic projection system based on artificial intelligence, characterized in that, The system includes: A user projection data acquisition module, which is used to identify the users within the projection area, extract features from the 2D images and 3D spatial information of the users, and integrate them into a user feature vector; the user feature vector includes 2D image features, distance features of the users in the 3D space, and the line-of-sight directions of the users. A user classification module, which is used to perform precise hierarchical classification on the users based on the user feature vector, combined with 2D image features, distance features of the users in the 3D space, and the line-of-sight directions of the users, specifically including: First, conduct preliminary stratification: Set the interaction distance threshold D th If the distance feature is greater than the interaction distance threshold D th Then user i is classified into the long-distance viewing layer. If the distance feature is less than the interaction distance threshold D th Then user i is classified into the short-distance interaction layer; Calculate the interest score for each user based on the two-dimensional image features, the distance features of the user in the three-dimensional space, and the user's line of sight direction. At the same time, design an interest score threshold I th , if the interest score is greater than the interest score threshold I th then mark this user as a high-interest interaction layer and allocate detailed content to them. If the interest score is less than the interest score threshold I th then mark this user as a low-interest interaction layer, and the content display will be mainly in overview; Integrate the results of the preliminary stratification and the interest score classification to construct a multi-level user demand matrix M for supporting the hierarchical management of personalized content display. A user demand generation module, which is used to generate and distribute personalized content for the users according to the user demand matrix M to meet the specific needs of the users in the holographic projection environment, specifically including: Generate a corresponding content matrix C for different user layers according to the user demand matrix M. Design a distribution frequency matrix L to control the content update frequency of the content matrix C corresponding to different user layers. A holographic projection synchronization module, which is used to design a perspective adaptive projection matrix and a content synchronization mechanism to perform personalized holographic real-time projection of personalized content.

Images