Intelligent lighting control method and system based on AI target detection

Abstract

The invention relates to the technical field of intelligent lighting control, and provides an intelligent lighting control method and system based on AI target detection, and the method comprises the steps: continuously capturing video frames in an environment through a camera, and carrying out the preprocessing of a foreground part of a moving target through image preprocessing; performing target identification on the preprocessed video frames through a target identification model, positioning a target object in the video frames, extracting key point data of the target object through a target detection algorithm to perform action identification, and keeping real-time tracking on the target object in continuous video frames based on a target tracking algorithm; and automatically judging whether the target object enters, leaves or stays in a preset area based on an identification result of the target identification model, and automatically controlling a lighting switch. By integrating target detection, action recognition and real-time tracking, the lighting equipment is automatically adjusted, meanwhile, the lighting setting is intelligently adjusted by learning user habits, and the maintenance cost is reduced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent lighting control technology, and in particular to an intelligent lighting control method and system based on AI target detection. Background Art

[0002] With the rapid development of smart home technology, smart lighting systems are increasingly being used in homes and commercial environments. Traditional smart lighting solutions usually rely on voice control or infrared sensing technology to achieve the function of "lights on when people come and lights off when people leave".

[0003] Among them, traditional voice-controlled lights rely on human voice to detect the presence of users, while infrared sensor lights determine whether a person is in the area by detecting heat changes. Once the person stops making sounds or generating heat, the lights will automatically turn off after a set time. Although these two methods can achieve the basic function of "lights on when people come and lights off when people leave", they have certain limitations. First, voice-controlled lights are easily disturbed by environmental noise, which may lead to misjudgment; second, infrared sensor lights cannot accurately determine the actual position of the human body in the area. Once the human body stops generating heat or making sounds, the lights may be misjudged and turned off, affecting the user experience. In addition, the energy efficiency control of these traditional technologies is not sophisticated enough, and there is often energy waste due to misjudgment or delayed response. Summary of the invention

[0004] The purpose of the present invention is to solve the above technical problems and propose an intelligent lighting control method and system based on AI target detection. By accurately identifying the movement of the target object, the switch of the lighting equipment is automatically controlled to achieve the energy-saving effect of turning on the lights when people come and turning off the lights when people leave. The purpose of the present invention can be achieved by the following technical solutions: The present invention provides an intelligent lighting control method based on AI target detection, comprising: The camera continuously captures video frames in the environment and performs image preprocessing to identify the foreground of moving objects; The pre-processed video frames are identified by the target recognition model and the target objects in the video frames are located. At the same time, the key point data of the target objects are extracted by the target detection algorithm for action recognition. The target objects are tracked in real time in the continuous video frames based on the target tracking algorithm. Based on the recognition results of the target recognition model, it automatically determines whether the target object enters, leaves or stays in the preset area and automatically controls the lighting switch.

[0005] Furthermore, the camera continuously captures video frames in the environment, and uses image preprocessing technology to process the foreground part of the moving target, including, The noise in the video frame is removed by filtering method, and then the background subtraction technique is applied to extract the foreground mask of the dynamic target from the static background. At the same time, the foreground mask is converted into a binary image, and the video frame is normalized to standardize its brightness and contrast.

[0006] Furthermore, the preprocessed video frame is subjected to target recognition through a target recognition model, and the target object in the video frame is located, including: Perform contour detection on video frames and obtain contour lists and hierarchical structure information.

[0007] Calculate the area of ​​each contour in the contour list and compare it with the set threshold, filter out the noise contours that are smaller than the preset threshold, and calculate the bounding box information of the retained valid contours; Based on the bounding box information of the valid contour, a rectangular box is drawn in the original video frame to calibrate the position of the target object.

[0008] Furthermore, the key point data of the target object is extracted through the target detection algorithm for action recognition, including: Inside the rectangular box, identify and mark the key points of the target object, including the positions of the head, shoulders, arms, hands, and legs; The key points and the corresponding key point connecting lines are marked inside the rectangular box in the video frame through the graphics processing library. At the same time, the position changes of the key points in consecutive frames are compared and the action classification is performed based on the time series analysis method.

[0009] Furthermore, the target object is tracked in real time in continuous video frames based on the target tracking algorithm, including: Determine a first video frame containing a target object through a target recognition model, determine an initial state of the target object, and set an initial error covariance and a measurement noise covariance of a filter; The state transfer matrix is ​​used to predict the predicted state of the target object, and the predicted state is the position and velocity of the next time step. The Kalman gain is calculated by using the position data extracted from continuous video frames through the target recognition model as the measurement input in the update process to update the state estimation of the target object.

[0010] Further, the first video frame containing the target object is determined by the target recognition model, the initial state of the target object is determined, and the initial error covariance and measurement noise covariance of the filter are set, including, Set the initial state of the target object, ,in, is the initial position of the target object, is the initial velocity; Set the initial error covariance and measurement noise covariance of the filter; The initial error covariance is: , in, and Uncertain measures of the initial position and initial velocity of the target object; The measurement noise covariance is: ; in, and are the noise variances in the x and y directions respectively; Furthermore, the state transfer matrix is ​​used to predict the predicted state of the target object, and the predicted state is the position and velocity of the next time step, including, The state transfer matrix is: , in, is the sampling interval; The predicted status is: Where Q is the process noise covariance.

[0011] Furthermore, the Kalman gain is calculated by using the position data extracted from the continuous video frames by the target recognition model as the measurement input in the updating process to update the state estimation of the target object, including: Set up the observation matrix: ; The Kalman gain is calculated as: , in, is the Kalman gain, K is the time step, H is the observation matrix, is the error covariance, R is the measurement noise covariance; Update time step states and error covariance; in, is the identity matrix, is the actual measured value.

[0012] Furthermore, based on the recognition result of the target recognition model, it is automatically determined whether the target object enters, leaves or stays in the preset area and the lighting switch is automatically controlled, including: When the target object enters the preset area for the first time, the lighting equipment is automatically turned on according to the ambient light conditions and time period; When the target object stays in the preset area, adjust the lighting intensity or direction according to its activity mode; When the target object leaves the preset area, the lighting is automatically turned off.

[0013] Based on the same inventive concept, the present invention also provides an intelligent lighting control system based on AI target detection, which adopts the above-mentioned intelligent lighting control method, including: The data acquisition module is used to continuously capture video frames in the environment through the camera and perform foreground processing of moving targets through image preprocessing; The data processing module is used to identify the target through the target recognition model after the preprocessed video frame, and locate the target object in the video frame. At the same time, the key point data of the target object is extracted through the target detection algorithm for action recognition, and the target object is tracked in real time in the continuous video frame based on the target tracking algorithm; The lighting control module is used to automatically determine whether the target object enters, leaves or stays in a preset area based on the recognition result of the target recognition model and automatically control the lighting switch.

[0014] Compared with the prior art, the present invention has at least one of the following technical effects: By integrating target detection, motion recognition and real-time tracking, it realizes automatic adjustment of lighting equipment and optimizes lighting according to user behavior and environmental conditions, significantly saving energy and improving user experience. At the same time, it adjusts lighting settings in an intelligent way by learning user habits, reducing maintenance costs.

[0015] (1) The present invention effectively reduces unnecessary energy consumption by precisely controlling the on and off of lighting equipment and activating lighting only when the presence of a human body is detected.

[0016] (2) Integrated target detection, motion recognition and tracking technologies can understand user behavior patterns and automatically adjust lighting settings accordingly, reducing the number of times users need to manually operate switches. At the same time, through the use of intelligent control, the service life of lighting equipment can be extended and the replacement frequency can be reduced, thereby reducing long-term maintenance and replacement costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solution of the embodiment of the present invention, the following is a brief introduction to the drawings required for describing the embodiment: Figure 1 The figure is a flowchart of the steps of the intelligent lighting control method based on AI target detection of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be understood that the specific implementation methods described herein are only used to explain the present invention and are not used to limit the present invention.

[0019] In traditional lighting systems, infrared sensing and voice control technologies often fail to accurately determine the presence of human bodies due to misjudgment and delayed response, causing lights to be erroneously turned off or on at inappropriate times. This not only causes inconvenience to users, but also reduces the practicality of the system because it cannot respond to people's dynamic behaviors in real time. In addition, existing lighting systems mostly use fixed-time control and cannot adjust the light intensity according to actual lighting needs, resulting in insufficient energy efficiency optimization. At the same time, these systems usually lack the ability to adapt to environmental changes and user behaviors, and cannot perform intelligent self-learning and optimization, which limits their efficiency and convenience in changing application scenarios.

[0020] In response to these problems, the present invention proposes an intelligent lighting method based on artificial intelligence technology. The inventors have designed an intelligent module that can accurately identify and track human movement by deeply thinking about and applying deep learning and real-time data analysis technology, effectively solving the problems of misjudgment and response delay. The system dynamically analyzes human behavior and position information, intelligently controls the switch and brightness of lighting equipment, and maximizes energy efficiency. At the same time, it also has adaptive optimization and multi-device linkage functions, reflecting greater application prospects and higher practical value. The specific implementation methods of the present invention are as follows: The present invention provides an intelligent lighting control method based on AI target detection, which is implemented by a target recognition model, a lighting device control component and a lighting device, wherein the target recognition model is responsible for identifying and collecting data of a target object, i.e., a person, and making a judgment, and the lighting device control module controls the on and off state of the lighting device according to the judgment result of the target recognition model. The specific operation steps include: The camera continuously captures video frames in the environment and performs image preprocessing to identify the foreground of moving objects; The pre-processed video frames are identified by the target recognition model and the target objects in the video frames are located. At the same time, the key point data of the target objects are extracted by the target detection algorithm for action recognition. The target objects are tracked in real time in the continuous video frames based on the target tracking algorithm. Based on the recognition results of the target recognition model, it automatically determines whether the target object enters, leaves or stays in the preset area and automatically controls the lighting switch.

[0021] Specifically, human movement data is collected in real time through AI technology, and the data is transmitted to the lighting equipment control component; the intelligent target recognition model processes and analyzes human movement data through a deep learning algorithm to determine whether the human body enters or leaves the lighting area; when the target recognition model determines that the human body enters the lighting area, the lighting equipment control module turns on the lighting equipment; when it is determined that the human body leaves the lighting area, the lighting equipment control component turns off the lighting equipment. In this way, by accurately identifying human movement, the present invention can avoid the energy waste caused by misjudgment of traditional voice-controlled lights and infrared sensor lights; adjust the brightness and switching time of the lighting equipment according to actual needs to further achieve energy-saving effects. At the same time, the lighting equipment control component adopts an intelligent control chip to achieve precise control of the lighting equipment, and supports a variety of communication protocols, such as Wi-Fi, Bluetooth, etc., which is convenient for linkage with other smart home devices. Remotely view and control the status of lighting equipment through mobile terminals such as mobile phone APP and PC web pages.

[0022] It also automatically adjusts the light brightness according to the ambient light intensity, keeping the indoor light comfortable and energy-saving, avoiding discomfort caused by too strong or too weak light to the human body, and reducing energy consumption. At the same time, it presets multiple lighting scene modes, such as reading mode, movie viewing mode, etc. Users can switch with one click according to their needs. Each scene mode will correspond to different light brightness, color and color temperature to create the most suitable lighting atmosphere.

[0023] Furthermore, the camera continuously captures video frames in the environment, and uses image preprocessing technology to process the foreground part of the moving target, including, The noise in the video frame is removed by filtering method, and then the background subtraction technique is applied to extract the foreground mask of the dynamic target from the static background. At the same time, the foreground mask is converted into a binary image, and the video frame is normalized to standardize its brightness and contrast.

[0024] Furthermore, the preprocessed video frame is subjected to target recognition through a target recognition model, and the target object in the video frame is located, including: Perform contour detection on video frames and obtain contour lists and hierarchical structure information.

[0025] Calculate the area of ​​each contour in the contour list and compare it with the set threshold, filter out the noise contours that are smaller than the preset threshold, and calculate the bounding box information of the retained valid contours; Based on the bounding box information of the valid contour, a rectangular box is drawn in the original video frame to calibrate the position of the target object.

[0026] Furthermore, the key point data of the target object is extracted through the target detection algorithm for action recognition, including: Inside the rectangular box, identify and mark the key points of the target object, including the positions of the head, shoulders, arms, hands, and legs; The key points and the corresponding key point connecting lines are marked inside the rectangular box in the video frame through the graphics processing library. At the same time, the position changes of the key points in consecutive frames are compared and the action classification is performed based on the time series analysis method.

[0027] Furthermore, the target object is tracked in real time in continuous video frames based on the target tracking algorithm, including: Determine a first video frame containing a target object through a target recognition model, determine an initial state of the target object, and set an initial error covariance and a measurement noise covariance of a filter; The state transfer matrix is ​​used to predict the predicted state of the target object, and the predicted state is the position and velocity of the next time step. The Kalman gain is calculated by using the position data extracted from continuous video frames through the target recognition model as the measurement input in the update process to update the state estimation of the target object.

[0028] Further, the first video frame containing the target object is determined by the target recognition model, the initial state of the target object is determined, and the initial error covariance and measurement noise covariance of the filter are set, including, Set the initial state of the target object, ,in, is the initial position of the target object, is the initial velocity; Set the initial error covariance and measurement noise covariance of the filter; The initial error covariance is: , in, and Uncertain measures of the initial position and initial velocity of the target object; The measurement noise covariance is: ; in, and are the noise variances in the x and y directions respectively; Furthermore, the state transfer matrix is ​​used to predict the predicted state of the target object, and the predicted state is the position and velocity of the next time step, including, The state transfer matrix is: , in, is the sampling interval; The predicted status is: Where Q is the process noise covariance.

[0029] Furthermore, the Kalman gain is calculated by using the position data extracted from the continuous video frames by the target recognition model as the measurement input in the updating process to update the state estimation of the target object, including: Set up the observation matrix: ; The Kalman gain is calculated as: , in, is the Kalman gain, K is the time step, H is the observation matrix, is the error covariance, R is the measurement noise covariance; Update time step states and error covariance; in, is the identity matrix, is the actual measured value.

[0030] It should be specifically noted that when developing this intelligent lighting control method, the inventors comprehensively considered different target tracking algorithms to ensure that the system can accurately track the dynamic behavior of users in a preset area. In addition to using Kalman filters to adapt to linear and high-speed dynamic changes, the inventors also considered using mean shift algorithms to optimize target tracking. The mean shift algorithm, with its non-parametric and density-based characteristics, is suitable for handling nonlinear motion and complex path problems in lighting control.

[0031] Initialize the tracking position and initialize the tracking window of the mean shift algorithm based on the bounding box information provided by the target recognition model. The tracking window is based on the rectangular box of the initially calibrated target object and is set around the center position of the target; Set up the feature space. Since the image has been binarized, the shape and edge features of the target area are used as the basis for tracking. Calculate the feature density distribution within the current window and find the area with the highest density. The window then moves to the centroid of this area. This process is repeated in each frame to track the dynamic movement of the target. Update the position of the tracking window according to the newly calculated centroid position.

[0032] The real-time tracking results provided by the mean shift algorithm are used to control the lighting system to ensure that the target area is always properly illuminated. At the same time, by analyzing the target's movement patterns and action changes, the system can further optimize the lighting intensity and direction to adapt to the user's activity needs. Furthermore, based on the recognition result of the target recognition model, it is automatically determined whether the target object enters, leaves or stays in the preset area and the lighting switch is automatically controlled, including: (1) When the target object enters the preset area for the first time, the lighting equipment is automatically turned on according to the ambient light conditions and time period; (2) When the target object stays in the preset area, the lighting intensity or direction is adjusted according to its activity mode; It should be specifically stated that the voice-controlled light will only light up when a person makes a sound. The disadvantage is that when a person is in a fixed place, he needs to keep making sounds to keep the light on; while the infrared detection / radar sensor light: when the human body moves, it can sense the person and turn on the light. The disadvantage is that in a fixed place, it needs to keep moving to keep the light on, which is also inconvenient in an environment that requires long-term sitting. Compared with Yuci, the AI ​​vision of this application detects whether a person is in a fixed place. Through visual deep learning, it can recognize the outline, body shape and movement of a person, and can keep the light on without the person making a sound or moving. (3) When the target object leaves the preset area, the lighting equipment will automatically turn off, further saving energy and extending the service life of the equipment.

[0033] Based on the same inventive concept, the present invention also provides an intelligent lighting control system based on AI target detection, which adopts the above-mentioned intelligent lighting control method, including: The data acquisition module is used to continuously capture video frames in the environment through the camera and perform foreground processing of moving targets through image preprocessing; The data processing module is used to identify the target through the target recognition model after the preprocessed video frame, and locate the target object in the video frame. At the same time, the key point data of the target object is extracted through the target detection algorithm for action recognition, and the target object is tracked in real time in the continuous video frame based on the target tracking algorithm; The lighting control module is used to automatically determine whether the target object enters, leaves or stays in a preset area based on the recognition result of the target recognition model and automatically control the lighting switch.

[0034] Although the present invention has been disclosed as above in the form of a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art may make possible changes and modifications to the technical solution of the present invention by using the methods and technical contents disclosed above without departing from the spirit and scope of the present invention. Therefore, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall fall within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent lighting control method based on AI target detection, characterized in that: include The camera continuously captures video frames in the environment and performs image preprocessing to identify the foreground of moving objects; The pre-processed video frames are subjected to target recognition through a target recognition model, and the target object in the video frame is located. At the same time, the key point data of the target object is extracted through a target detection algorithm for action recognition, and the target object is tracked in real time in the continuous video frames based on a target tracking algorithm; Based on the recognition result of the target recognition model, it is automatically determined whether the target object enters, leaves or stays in a preset area and the lighting switch is automatically controlled.

2. The intelligent lighting control method according to claim 1, characterized in that: The camera continuously captures video frames in the environment, and uses image preprocessing technology to process the foreground part of the moving target, further including: The noise in the video frame is removed by filtering, and then the foreground mask of the dynamic target is extracted from the static background by applying the background subtraction technique. The foreground mask is converted into a binary image, and the video frame is normalized to standardize its brightness and contrast.

3. The intelligent lighting control method according to claim 1, characterized in that: The pre-processed video frame is subjected to target recognition through a target recognition model, and the target object in the video frame is located, further comprising: Performing contour detection on the video frame to obtain a contour list and hierarchical structure information; Calculate the area of ​​each contour in the contour list and compare it with a set threshold, filter out noise contours that are smaller than the preset threshold, and calculate bounding box information of the retained valid contours; A rectangular box is drawn in the original video frame based on the bounding box information of the valid contour to mark the position of the target object.

4. The intelligent lighting control method according to claim 3, characterized in that: Extracting key point data of the target object by a target detection algorithm to perform action recognition further includes: Inside the rectangular frame, identifying and marking key points of the target object, the key points including the positions of the head, shoulders, arms, hands and legs; The key points and the corresponding key point connecting lines are marked inside the rectangular frame in the video frame by a graphics processing library, and the position changes of the key points in consecutive frames are compared and action classification is performed based on a time series analysis method.

5. The intelligent lighting control method according to claim 4, characterized in that: Maintaining real-time tracking of the target object in the continuous video frames based on the target tracking algorithm further includes: Determine a first video frame containing the target object through the target recognition model, determine an initial state of the target object, and set an initial error covariance and a measurement noise covariance of a filter; A state transfer matrix is ​​applied to predict the predicted state of the target object, wherein the predicted state is the position and the speed of the next time step; the Kalman gain is calculated using the position data extracted from the continuous video frames through the target recognition model as the measurement input in the updating process, and is used to update the state estimation of the target object.

6. The intelligent lighting control method according to claim 5, characterized in that: Determining a first video frame containing the target object through the target recognition model, determining an initial state of the target object, and setting an initial error covariance and a measurement noise covariance of a filter, further comprising, setting the initial state of the target object, ,in, is the initial position of the target object, is the initial velocity; Setting an initial error covariance and a measurement noise covariance of the filter; The initial error covariance is: ， in, and a measure of uncertainty of the initial position and the initial velocity of the target object; The measurement noise covariance is: ； in, and are the noise variances in the x and y directions, respectively.

7. The intelligent lighting control method according to claim 6, characterized in that: Applying the state transfer matrix to predict the predicted state of the target object, the predicted state being the position and the speed at the next time step, further comprising, The state transfer matrix is: ， in, is the sampling interval; The predicted status is: Wherein, Q is the process noise covariance.

8. The intelligent lighting control method according to claim 7, characterized in that: The Kalman gain is calculated from the position data extracted by the target recognition model from the continuous video frames as the measurement input in the updating process, and is used to update the state estimation of the target object, further comprising: Set up the observation matrix: ; The Kalman gain is calculated as: , in, is the Kalman gain, K is the time step, H is the observation matrix, is the error covariance, R is the measurement noise covariance; updating the time step state and the error covariance; in, is the identity matrix, is the actual measured value.

9. The intelligent lighting control method according to claim 8, characterized in that: Automatically judging whether the target object enters, leaves or stays in a preset area based on the recognition result of the target recognition model and automatically controlling the lighting switch, further comprising: When the target object enters the preset area for the first time, the lighting device is automatically turned on according to the ambient light conditions and time period; When the target object stays in the preset area, adjusting the lighting intensity or direction according to its activity mode; When the target object leaves the preset area, the lighting is automatically turned off.

10. An intelligent lighting control system based on AI target detection, using any one of the intelligent lighting control methods described in claims 1 to 9, characterized in that: include, The data acquisition module is used to continuously capture video frames in the environment through the camera and perform foreground processing of moving targets through image preprocessing; A data processing module, used to perform target recognition on the pre-processed video frame through a target recognition model, locate the target object in the video frame, extract key point data of the target object through a target detection algorithm for action recognition, and keep real-time tracking of the target object in the continuous video frame based on a target tracking algorithm; The lighting control module is used to automatically determine whether the target object enters, leaves or stays in a preset area based on the recognition result of the target recognition model and automatically control the lighting switch.

Images