Health lighting adaptive adjustment system based on light environment information

Abstract

The present invention discloses a health lighting adaptive adjustment system based on light environment information, which relates to the technical field of light adjustment and realizes the intelligent control of lighting according to the user's mood. The present invention inputs the three-dimensional personnel movement trajectory and the scene ambient light change texture model into the scene three-dimensional image model, and then establishes a number of behavior event decisions. According to the real-time three-dimensional personnel action trajectory and the real-time facial image model of each target person extracted from the real-time scene graph model, a corresponding real-time scene ambient light change texture model is generated according to the real-time ambient light change data. The real-time three-dimensional personnel action trajectory is matched with each behavior event decision, and according to the matching result, it is judged whether the light intensity in the real-time scene ambient light change texture model meets the standard, and then the corresponding event lighting decision is generated and executed.

Description

Technical Field

[0001] The present invention relates to the technical field of lighting adjustment, and specifically to a healthy lighting adaptive adjustment system based on light environment information. Background Art

[0002] Healthy lighting generally refers to optimizing people's physical and mental health through lighting. Such a lighting system can simulate the changes of natural light, such as the color temperature and brightness changes during the day, so as to adjust the human biological clock. The design of healthy lighting takes into account the physiological needs of humans, can improve attention, mood and productivity, and at the same time improve sleep quality.

[0003] Existing defects of existing healthy lighting adaptive adjustment technologies:

[0004] Accuracy and perception differences: Although technologies can adjust the color and brightness of lights according to time and environment, there are differences in the biological clocks and visual perceptions of individuals. Some people may be more sensitive to specific spectra, so it may not fully meet the needs of everyone.

[0005] Insufficient research on long-term effects and health impacts: Although healthy lighting systems can improve attention and mood in the short term, further scientific research and evidence are still needed for the impacts of long-term use. Especially for the long-term effects on sleep quality and chronic health problems, more tracking and research are required.

[0006] Therefore, how to accurately adjust the lighting brightness of the lighting lamp according to the visual perception and mood of the target person is a difficulty in the existing technology. For this reason, a healthy lighting adaptive adjustment system based on light environment information is provided. Summary of the Invention

[0007] In order to solve the above technical problems, the purpose of the present invention is to provide a healthy lighting adaptive adjustment system based on light environment information.

[0008] In order to achieve the above purpose, the present invention provides the following technical solutions:

[0009] A healthy lighting adaptive adjustment system based on light environment information, including a cloud computing platform, which is communicatively connected to a historical lighting data acquisition module, a lighting environment mapping relationship analysis module, and a lighting decision management module;

[0010] The lighting data acquisition module is used to install a number of groups of lighting sensors and cameras in the scene area, so as to collect the historical behavior video data of the target person in the scene area and the corresponding historical ambient light change data, and collect the real-time behavior video data and real-time ambient light change data of the scene area;

[0011] The illumination environment mapping relationship analysis module is used to establish a number of three-dimensional personnel movement trajectories and a scene three-dimensional image model based on historical behavior video data, and at the same time establish a scene environment light change texture model according to historical ambient light change data. The three-dimensional personnel movement trajectories and the scene environment light change texture model are input into the scene three-dimensional image model, and then a number of behavior event decisions are established.

[0012] The illumination decision management module is used to extract the real-time three-dimensional personnel action trajectories and real-time facial image models of each target person from the real-time scene graph model, generate a corresponding real-time scene environment light change texture model according to the real-time ambient light change data, match the real-time three-dimensional personnel action trajectories with each behavior event decision, and judge whether the illumination intensity in the real-time scene environment light change texture model meets the standard according to the matching result, and then generate a corresponding event lighting decision and execute it.

[0013] Further, the acquisition process of the ambient light change data and the behavior video data includes:

[0014] There is an overlapping relationship in the data acquisition ranges of the light sensors and cameras at adjacent spatial positions, that is, for any group of light sensors and cameras, the overlapping range between the data acquisition ranges of each group of light sensors and cameras at their adjacent spatial positions is equal to the data acquisition range of this group of light sensors and cameras;

[0015] Set a data acquisition period, and then each group of light sensors and cameras collect historical ambient light change data and historical behavior video data within their data ranges in each data acquisition period, as well as collect real-time behavior video data and real-time ambient light change data in the real-time data acquisition period, and label the numbers of the corresponding groups of light sensors and cameras for each ambient light change data and behavior video data.

[0016] Further, the establishment process of the three-dimensional personnel movement trajectory includes:

[0017] The illumination environment mapping relationship analysis module divides the historical behavior video data collected by each group of cameras into several pieces of historical behavior image data frame by frame, and uses image blurring technology to segment the target person image and the light source image from each piece of historical behavior image data;

[0018] Then the staff uploads the corresponding target person information according to the target person image, and the target person information includes gender, age, and identity number;

[0019] Set an action node at the head, torso, and limbs of each target person image respectively, and use the action node at the torso as the central node, and connect the action nodes at other parts to the central node to obtain a motion structure tree;

[0020] Sequentially splice the motion structure trees in the historical behavior image data of each historical behavior video data in chronological order to obtain the two-dimensional personnel action trajectories of each historical behavior video data;

[0021] Mutually map the corresponding two-dimensional personnel action trajectories to obtain three-dimensional personnel motion trajectories, establish a three-dimensional coordinate system, overlap and map the three-dimensional personnel action trajectories of different historical behavior video data onto the same three-dimensional coordinate system, divide into several equally long time segments according to the time length of the data acquisition period, and then divide each personnel action trajectory into several local action trajectories according to the time segments;

[0022] Compare the local action trajectories of historical behavior video data in the same time segment under different data acquisition periods, and then obtain the similarity η of each three-dimensional personnel action trajectory;

[0023] Set a similarity threshold. If the similarity η of the three-dimensional personnel action trajectories corresponding to two historical behavior video data under different data acquisition periods is less than or equal to the similarity threshold, randomly eliminate the three-dimensional personnel action trajectory corresponding to one historical behavior video data, otherwise do nothing;

[0024] The staff annotates the corresponding behavior event names for each retained three-dimensional personnel action trajectory, and then obtains several behavior events.

[0025] Furthermore, the process of obtaining the similarity η includes:

[0026] Divide several equally spaced time nodes within the time segment, and then divide each local action trajectory into several local action nodes according to the distribution of each time node, and obtain the coordinate points of each local action node on the three-dimensional coordinate system;

[0027] For any local action trajectory, generate corresponding local action vectors from two local action nodes on its adjacent time nodes, and then obtain the angle θ between the local action vectors of each local action trajectory at the same pair of time nodes. The calculation formula for the angle θ is: Where and respectively represent the local action vectors in two local action trajectories;

[0028] Set an angle threshold. If the number of angles between two local action vectors is greater than or equal to the angle threshold, determine that the pair of local action vectors is similar, otherwise determine that they are not similar;

[0029] Count the number of similar local action vectors between each local action trajectory within each time segment, and then calculate the similarity η between each three-dimensional personnel action trajectory. The calculation formula for the similarity η is: where N represents the total number of time segments, and Num i represents the number of local action vectors similar between two local action trajectories within the i-th time segment.

[0030] Furthermore, the process of establishing the scene three-dimensional image model includes:

[0031] The lighting environment relationship analysis module establishes corresponding facial image models based on the target person images in each historical behavior video data. At the same time, it establishes corresponding local three-dimensional models of the scene based on each historical behavior video data under the same data acquisition period, and sets the light source position in the corresponding local three-dimensional model of the scene according to the light source images extracted from the historical behavior image data. The light source position can be a device such as an LED lamp;

[0032] According to the relationship that there is an intersection in the data acquisition ranges of each camera within the scene area, the local three-dimensional models of the scene are sequentially spliced to obtain the scene three-dimensional image model.

[0033] Furthermore, the process of establishing the scene ambient light change texture model based on the historical ambient light change data includes:

[0034] The historical ambient light change data includes the light intensities at different positions in the corresponding scene area. A number of ambient light textures are generated around the corresponding light source position, and then the scene ambient light change texture model is obtained, where the light intensities of each ambient light texture are successively within each light intensity threshold interval;

[0035] Facial expression models corresponding to each mood are obtained through the Internet, and then the characteristic facial feature combinations are extracted from each facial expression model. At the same time, each mood is divided into normal, general, and discomfort, and classification labels are assigned to each characteristic facial feature combination according to the classification results;

[0036] The facial expression models and the scene ambient light change texture models corresponding to each group of optical sensors and cameras under the same data acquisition period are matched, and the matching results are associated with the corresponding three-dimensional human action trajectories;

[0037] An event timeline is established according to the time length of the data acquisition period, and the relevant facial expression models, scene ambient light change texture models, and three-dimensional human action trajectories are mapped onto the event timeline.

[0038] Furthermore, the process of establishing the behavior event decision includes:

[0039] Set age ranges, and then, under the ambient light change texture models of various scenarios, obtain the historical facial expressions of the corresponding facial expression models when the same three-dimensional human motion trajectory changes over time through the ambient light textures in different ambient light change texture models of scenarios. Match each historical facial expression with each combination of characteristic facial features. If the combination of characteristic facial features is similar to the historical facial expression, it is determined that the two match; otherwise, no operation is performed.

[0040] Mark the classification labels of the corresponding combinations of characteristic facial features on the historical facial expressions according to the matching results. Then, according to the illumination intensity threshold ranges where the ambient light textures corresponding to the historical facial expressions are located, obtain the moods of target persons of different age ranges and genders when performing the corresponding three-dimensional human motion trajectories in each illumination intensity threshold range.

[0041] The illumination environment mapping relationship analysis module integrates the three-dimensional human motion trajectories of each behavior event and the moods corresponding to each illumination intensity threshold range to obtain the corresponding illumination decisions for behavior events, and marks the corresponding age ranges and genders.

[0042] Furthermore, the generation process of the event illumination decision includes:

[0043] The illumination decision management module generates a corresponding real-time scenario ambient light change texture model according to the real-time ambient light change data. At the same time, extract the real-time three-dimensional human motion trajectories and real-time facial image models of each target person from the real-time scenario graph model, and obtain the ages and genders of the target persons according to the real-time facial image models.

[0044] Map the real-time scenario ambient light change texture model to the corresponding light source positions in the scenario three-dimensional image model, and match the ages, genders, and real-time three-dimensional human motion trajectories of the target persons with each illumination decision for behavior events.

[0045] Perform a secondary match between each illumination intensity threshold range of the illumination decision for behavior events and the real-time ambient light texture in the real-time scenario ambient light change texture model. If the illumination intensity of the real-time ambient light texture where the target person is located is within the illumination intensity threshold range indicating normal in the illumination decision for behavior events, no operation is performed.

[0046] Otherwise, generate an event illumination decision according to the difference between the illumination intensity of the real-time ambient light texture where the target person is located and the illumination intensity threshold range indicating normal in the illumination decision for behavior events.

[0047] The event illumination decision includes the light source position where the target person is located and the illumination intensity adjustment value for the light source position.

[0048] Compared with the prior art, the beneficial effects of the present invention are:

[0049] The present invention inputs the three-dimensional human motion trajectory and the scene ambient light change texture model into the scene three-dimensional image model, and then establishes a number of behavior event decisions. According to the real-time three-dimensional human motion trajectory and the real-time facial image model of each target person extracted from the real-time scene graph model, a corresponding real-time scene ambient light change texture model is generated according to the real-time ambient light change data. The real-time three-dimensional human motion trajectory is matched with each behavior event decision, and it is judged whether the light intensity in the real-time scene ambient light change texture model meets the standard according to the matching result, and then a corresponding event lighting decision is generated and executed, realizing accurate real-time adjustment of the lighting brightness of the lighting lamp according to the visual perception and emotion of the target person. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention.

[0051] Figure 1 It is the schematic diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments of the present invention fall within the scope protected by the present invention.

[0053] As Figure 1 shown, the healthy lighting adaptive adjustment system based on light environment information includes a cloud computing platform, and the cloud computing platform is communicatively connected to a historical light data acquisition module, a light environment mapping relationship analysis module, and a light decision management module;

[0054] The light data acquisition module is used to install a number of groups of light sensors and cameras in the scene area, and then collect the historical behavior video data of the target person in the scene area and the corresponding historical ambient light change data, as well as collect the real-time behavior video data and real-time ambient light change data of the field area, and then send the historical behavior video data and historical ambient light change data to the light environment mapping relationship analysis module, and send the real-time behavior video data and real-time ambient light change data to the light decision management module;

[0055] The illumination environment mapping relationship analysis module is used to establish several three-dimensional personnel movement trajectories and scene three-dimensional image models based on historical behavior video data, and at the same time establish a scene environment light change texture model based on historical environmental light change data. The three-dimensional personnel movement trajectories and the scene environment light change texture model are input into the scene three-dimensional image model, and then several behavior event decisions are established.

[0056] The illumination decision management module is used to extract the real-time three-dimensional personnel action trajectories and real-time facial image models of each target person from the real-time scene graph model, generate a corresponding real-time scene environment light change texture model according to the real-time environmental light change data, match the real-time three-dimensional personnel action trajectories with each behavior event decision, and judge whether the illumination intensity in the real-time scene environment light change texture model meets the standard according to the matching result, and then generate a corresponding event illumination decision and execute it.

[0057] Further, the working principle of the present invention is illustrated by the following embodiments:

[0058] Install n groups of illumination sensors and cameras in the scene area, and set numbers a 1 、a 2 、a 3 、……、a n for each illumination sensor and camera, where n is a natural number greater than 0;

[0059] It should be noted that there is an intersection relationship in the data acquisition ranges of the illumination sensors and cameras at adjacent spatial positions, that is, for any group of illumination sensors and cameras, the intersection range between the data acquisition ranges of each group of illumination sensors and cameras at its adjacent spatial positions is equal to the data acquisition range of this group of illumination sensors and cameras;

[0060] Set the data acquisition period, and then each group of illumination sensors and cameras collect historical environmental light change data and historical behavior video data within their data ranges in each data acquisition period, as well as collect real-time behavior video data and real-time environmental light change data in the real-time data acquisition period, and label the corresponding numbers of the illumination sensors and cameras for each environmental light change data and behavior video data;

[0061] Then the illumination data acquisition module sends the historical behavior video data and historical environmental light change data to the illumination environment mapping relationship analysis module, and sends the real-time behavior video data and real-time environmental light change data to the illumination decision management module.

[0062] Further, the illumination environment mapping relationship analysis module divides the historical behavior video data collected by each group of cameras into several pieces of historical behavior image data frame by frame, and uses image blurring technology to segment the target person image and the light source image from each piece of historical behavior image data;

[0063] Furthermore, the staff uploads the corresponding target personnel information according to the target personnel image, and the target personnel information includes gender, age, and identity number;

[0064] An action node is set at the head, torso, and limbs of each target personnel image respectively, and taking the action node of the torso part as the central node, the action nodes of other parts are connected to the central node to obtain a motion structure tree;

[0065] The motion structure trees in the historical behavior image data of each historical behavior video data are sequentially spliced according to the time sequence, and then the two-dimensional personnel action trajectories of each historical behavior video data are obtained;

[0066] Since there is an overlapping relationship in the data acquisition ranges of the cameras at adjacent spatial positions, then in the same data acquisition period, there must be the same part between the two-dimensional personnel action trajectories generated from the historical behavior video data collected by the cameras with overlapping data acquisition ranges. Then, the corresponding two-dimensional personnel action trajectories are mutually mapped, and then the corresponding three-dimensional personnel motion trajectories are obtained;

[0067] A three-dimensional coordinate system is established, and the three-dimensional personnel action trajectories of different historical behavior video data are overlapped and mapped onto the same three-dimensional coordinate system. According to the time length of the data acquisition period, several equal-length time segments are divided, and then according to the time segments, each personnel action trajectory is divided into several local action trajectories;

[0068] The local action trajectories of the historical behavior video data in the same time segment under different data acquisition periods are compared with each other, and then the similarity η of each three-dimensional personnel action trajectory is obtained. The process of obtaining the similarity η includes:

[0069] Several time nodes with equal intervals are divided within the time segment, and then according to the distribution of each time node, each local action trajectory is divided into several local action nodes, and the coordinate positions of each local action node on the three-dimensional coordinate system are obtained;

[0070] For any local action trajectory, a corresponding local action vector is generated by two local action nodes at its adjacent time nodes, and then the included angle θ between the local action vectors of each local action trajectory at the same pair of time nodes is obtained. The calculation formula of the included angle θ is: Where and respectively represent the local action vectors in two local action trajectories;

[0071] A threshold for the included angle degree is set. If the number of included angles between two local action vectors is greater than or equal to the threshold for the included angle degree, it is determined that the pair of local action vectors is similar, otherwise it is determined to be dissimilar;

[0072] Count the number of similar local action vectors between local action trajectories in each time period, and then calculate the similarity η between each three-dimensional human action trajectory. The calculation formula for the similarity η is as follows: where N represents the total number of time periods, and Num i represents the number of similar local action vectors between two local action trajectories in the i-th time period;

[0073] Set a similarity threshold. If the similarity η of the three-dimensional human action trajectories corresponding to two historical behavior video data in different data collection cycles is less than or equal to the similarity threshold, randomly eliminate the three-dimensional human action trajectory corresponding to one of the historical behavior video data; otherwise, do nothing.

[0074] The staff annotates the corresponding behavior event names for each remaining three-dimensional human action trajectory, and then obtains a number of behavior events.

[0075] Furthermore, the light environment relationship analysis module establishes a corresponding facial image model based on the target person images in each historical behavior video data, and at the same time establishes a corresponding local three-dimensional scene model based on each historical behavior video data in the same data collection cycle. According to the light source image extracted from the historical behavior image data, the light source position is set in the corresponding local three-dimensional scene model. The light source position can be a device such as an LED lamp.

[0076] According to the relationship that there is an intersection in the data collection ranges of each camera in the scene area, the local three-dimensional scene models are sequentially spliced to obtain a three-dimensional scene image model.

[0077] Furthermore, input the three-dimensional human action trajectories of each behavior event into the three-dimensional scene image model. At the same time, according to all the historical behavior video data associated with each three-dimensional human action trajectory, connect the corresponding facial image models with the three-dimensional human action trajectories respectively. At the light source position corresponding to the three-dimensional scene image model, a scene ambient light change texture model is established for the historical ambient light change data related to each historical behavior video data.

[0078] Since the light intensity of the ambient light decreases as the propagation distance increases, set multiple light intensity threshold intervals. Then the historical ambient light change data includes the light intensity at different positions in the corresponding scene area, and several ambient light textures are generated around the corresponding light source position, and then a scene ambient light change texture model is obtained. The light intensity of each ambient light texture is successively in each light intensity threshold interval.

[0079] Obtain facial expression models corresponding to various moods through the Internet, and then extract characteristic facial feature combinations from each facial expression model. At the same time, classify each mood into normal, general, and discomfort, and label classification tags for each characteristic facial feature combination according to the classification results. For example, the mood can be happy, irritable, etc.;

[0080] Match the facial expression models and the scene ambient light change texture models corresponding to each group of optical sensors and cameras in the same data acquisition cycle, and associate the matching results with the corresponding three-dimensional human motion trajectories;

[0081] Establish an event timeline according to the time length of the data acquisition cycle, and map each relevant facial expression model, scene ambient light change texture model, and three-dimensional human motion trajectory onto the event timeline;

[0082] Set an age range. Then, under each scene ambient light change texture model, when obtaining the historical facial expressions of the corresponding facial expression model as the same three-dimensional human motion trajectory changes over time through the ambient light textures in different scene ambient light change texture models, match each historical facial expression with each characteristic facial feature combination. If the characteristic facial feature combination is similar to the historical facial expression, it is determined that the two match; otherwise, no operation is performed;

[0083] Label the classification tags of the corresponding characteristic facial feature combinations on the historical facial expressions according to the matching results. Then, according to the illumination intensity threshold range where the ambient light texture corresponding to each historical facial expression is located, obtain the moods corresponding to each illumination intensity threshold range for target personnel of different age ranges and genders when performing the corresponding three-dimensional human motion trajectories.

[0084] Furthermore, the illumination environment mapping relationship analysis module integrates the three-dimensional human motion trajectories of each behavior event and the moods corresponding to each illumination intensity threshold range to obtain the corresponding behavior event illumination decisions, and labels the corresponding age range and gender;

[0085] Then, the illumination environment mapping relationship analysis module sends all behavior event illumination decisions to the illumination decision management module. The illumination decision management module generates the corresponding real-time scene ambient light change texture model according to the real-time ambient light change data. At the same time, extract the real-time three-dimensional human motion trajectories and real-time facial image models of each target person from the real-time scene graph model, and obtain the age and gender of the target person according to the real-time facial image model;

[0086] Map the real-time scene ambient light change texture model onto the corresponding light source positions in the scene three-dimensional image model, and match the age, gender, and real-time three-dimensional human motion trajectories of the target person with each behavior event illumination decision;

[0087] If the similarity rate between the real-time 3D human motion trajectory and the 3D human motion trajectory is above 90%, and the age and gender match the lighting decision of the behavior event, then the real-time 3D human motion trajectory matches the corresponding behavior event lighting decision; otherwise, it is judged as unmatched.

[0088] Match each lighting intensity threshold interval of the behavior event lighting decision with the real-time ambient light texture in the real-time scene ambient light change texture model. If the lighting intensity of the real-time ambient light texture where the target person is located is within the lighting intensity threshold interval indicating normal in the behavior event lighting decision, then do nothing.

[0089] Otherwise, generate an event lighting decision based on the difference between the lighting intensity of the real-time ambient light texture where the target person is located and the lighting intensity threshold interval indicating normal in the behavior event lighting decision.

[0090] The event lighting decision includes the light source position where the target person is located and the lighting intensity adjustment value for the light source position.

[0091] The lighting decision management module adjusts the lighting intensity of the corresponding light source position according to the event lighting decision until the target person finishes executing the current behavior event.

[0092] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A healthy lighting adaptive adjustment system based on light environment information, including a cloud computing platform, characterized in that: The cloud computing platform is communicatively connected with a historical lighting data acquisition module, a lighting environment mapping relationship analysis module, and a lighting decision management module; The illumination data acquisition module is used to install several groups of illumination sensors and cameras in the scene area, thereby collecting historical behavior video data of the target person in the scene area and corresponding historical ambient light change data, as well as collecting real-time behavior video data and real-time ambient light change data of the scene area; The illumination environment mapping relationship analysis module is used to establish a number of three-dimensional personnel motion trajectories and scene three-dimensional image models based on historical behavior video data, and at the same time establish a scene ambient light change texture model based on historical ambient light change data, and input the three-dimensional personnel motion trajectory and scene ambient light change texture model into the scene three-dimensional image model, thereby establishing a number of behavior event decisions; The lighting decision management module is used to extract the real-time three-dimensional personnel motion trajectory and real-time facial image model of each target person from the real-time scene graph model, generate a corresponding real-time scene ambient light change texture model according to the real-time ambient light change data, match the real-time three-dimensional personnel motion trajectory with each behavior event decision, and judge whether the light intensity in the real-time scene ambient light change texture model meets the standard according to the matching result, and then generate and execute the corresponding event lighting decision; The process of establishing the three-dimensional personnel motion trajectory includes: The historical behavior video data collected by each group of cameras is divided into several historical behavior image data by frame, and the target person image and light source image are segmented from each historical behavior image data by image blur technology. The staff uploads the corresponding target person information according to the target person image, and the target person information includes gender and age; An action node is set at the head, torso and limbs of each target person image, and each action node is connected to obtain a motion structure tree; The motion structure tree in the historical behavior image data of each historical behavior video data is sequentially spliced ​​in chronological order to obtain a two-dimensional personnel motion trajectory of each historical behavior video data; The corresponding two-dimensional personnel motion trajectories are mapped to each other to obtain three-dimensional personnel motion trajectories, a three-dimensional coordinate system is established, and the three-dimensional personnel motion trajectories of different historical behavior video data are overlapped and mapped on the same three-dimensional coordinate system. According to the time length of the data collection cycle, several equal-length time segments are divided, and then each personnel motion trajectory is divided into several local motion trajectories according to the time segment; Compare the local motion trajectories of the historical behavior video data in different data collection cycles in the same time segment, and then obtain the similarity η of each three-dimensional person's motion trajectory; Set a similarity threshold. If the similarity η of the three-dimensional personnel motion trajectories corresponding to two historical behavior video data in different data collection cycles is less than or equal to the similarity threshold, then randomly remove the three-dimensional personnel motion trajectory corresponding to one historical behavior video data, otherwise do nothing; The process of obtaining the similarity η includes: A number of time nodes are divided in the time segment, and then each local action trajectory is divided into a number of local action nodes according to the distribution of each time node, and the coordinate point position of each local action node in the three-dimensional coordinate system is obtained; For any local action trajectory, the corresponding local action vector is generated by the two local action nodes at its adjacent time nodes, and then the angle θ between the local action vectors of each local action trajectory at the same pair of time nodes is obtained, where the calculation formula of the angle θ is: in and Respectively represent the local action vectors in two local action trajectories; Set an angle degree threshold. If the angle between two local motion vectors is greater than or equal to the angle degree threshold, the pair of local motion vectors are considered similar, otherwise they are considered dissimilar. The similarity of local motion vectors between local motion trajectories in each time segment is counted, and then the similarity η between the motion trajectories of three-dimensional personnel is calculated. The calculation formula of the similarity η is: Where N represents the total number of time segments, i represents the number of similarities between the local motion vectors of two local motion trajectories in the i-th time segment; The process of establishing the scene three-dimensional image model includes: Establish a corresponding facial image model according to the target person image in each historical behavior video data, and at the same time establish a corresponding scene local 3D model according to each historical behavior video data in the same data collection period, and set the light source position in the corresponding scene local 3D model according to the light source image extracted from the historical behavior image data; According to the relationship that the data collection ranges of the cameras in the scene area overlap, the local 3D models of the scenes are sequentially spliced ​​to obtain a 3D image model of the scene; The establishment process of the behavioral event decision includes: Set the age range, and then obtain the historical facial expressions of the facial expression model when the same 3D person's motion trajectory changes over time and passes through the ambient light texture in the ambient light change texture model of different scenes under the ambient light change texture model of each scene, and match each historical facial expression with each characteristic facial feature combination. If the characteristic facial feature combination is similar to the historical facial expression, it is determined that the two are matched, otherwise no operation is performed; According to the matching results, the classification labels of the corresponding characteristic facial features are marked on the historical facial expressions, and then according to the light intensity threshold intervals of the ambient light textures corresponding to each historical facial expression, the moods corresponding to the target persons of different age ranges and genders in each light intensity threshold interval when they perform the corresponding three-dimensional personnel action trajectory are obtained; The lighting environment mapping relationship analysis module integrates the three-dimensional motion trajectory of each behavioral event and the mood corresponding to each lighting intensity threshold range, obtains the corresponding lighting decision for the behavioral event, and marks the corresponding age range and gender; The generation process of the event lighting decision includes: Generate a corresponding real-time scene ambient light change texture model based on the real-time ambient light change data, and extract the real-time three-dimensional motion trajectory and real-time facial image model of each target person from the real-time scene graph model, and obtain the age and gender of the target person based on the real-time facial image model; Map the real-time scene ambient light change texture model to the corresponding light source position in the scene 3D image model, and match the target person’s age, gender, and real-time 3D person’s motion trajectory with the lighting decisions of each behavior event; According to the matching result, the light intensity of the real-time ambient light texture where the target person is located is obtained, and the difference between the light intensity threshold range representing the normal light intensity in the behavior event lighting decision is obtained, and then the event lighting decision is generated and executed.

2. The healthy lighting adaptive adjustment system based on light environment information according to claim 1 is characterized in that: The process of collecting the ambient light change data and the behavior video data includes: For any group of illumination sensors and cameras, the intersection range between the data collection ranges of each group of illumination sensors and cameras at adjacent spatial positions is equal to the data collection range of the group of illumination sensors and cameras; A data collection cycle is set, and each group of light sensors and cameras collects historical ambient light change data and historical behavior video data within its data range in each data collection cycle, as well as real-time behavior video data and real-time ambient light change data in a real-time data collection cycle, and each ambient light change data and behavior video data is labeled with a corresponding number.

3. The healthy lighting adaptive adjustment system based on light environment information according to claim 2 is characterized in that: The process of establishing a scene ambient light change texture model based on historical ambient light change data includes: The historical ambient light change data includes the light intensity at different positions of the corresponding scene area, and a plurality of ambient light textures are generated around the corresponding light source position, thereby obtaining a scene ambient light change texture model, wherein the light intensity of each ambient light texture is in each light intensity threshold interval in turn; Obtain facial expression models corresponding to each mood through the Internet, and then extract characteristic facial features from each facial expression model. At the same time, classify each mood into normal, general, and uncomfortable, and label each characteristic facial features according to the classification results. The facial expression models and scene ambient light change texture models corresponding to each group of optical sensors and cameras in the same data collection cycle are matched, and the matching results are associated with the corresponding three-dimensional personnel motion trajectories. An event timeline is established according to the length of the data collection cycle, and each related facial expression model, scene ambient light change texture model and three-dimensional personnel motion trajectory is mapped on the event timeline.

Images