Environmental gas monitoring method and system based on deep learning

Abstract

The invention discloses an environment gas monitoring method and system based on deep learning, and the method comprises the steps: constructing an environment monitoring mobile robot, obtaining the obstacle image data of a target environment, carrying out the data enhancement processing, and obtaining the enhanced wood material obstacle image data; a TAA strategy is introduced to train the YOLOv8 target detection model, and an improved YOLOv8 target detection model is constructed; performing identification processing on the enhanced wood material obstacle image data to obtain wood furniture information data of the target environment; and controlling the environment monitoring mobile robot to globally move to perform gas monitoring on the target environment so as to obtain a target environment harmful gas monitoring result. According to the invention, the position of the wooden furniture can be accurately identified, the concentration of harmful gas can be effectively monitored in real time, and the safety and health of the home environment are ensured. The environment gas monitoring method and system based on deep learning can be widely applied to the technical field of environment gas monitoring.

Description

Technical Field

[0001] The present invention relates to the field of environmental gas monitoring technology, and in particular to an environmental gas monitoring method and system based on deep learning. Background Art

[0002] In modern homes and offices, wooden furniture is popular for its beauty and durability. However, these furniture may release harmful gases such as formaldehyde and TVOC (total volatile organic compounds), posing a potential threat to the health of occupants. Although traditional air quality monitoring equipment, such as handheld detectors or fixed monitoring stations, can measure indoor air quality, they cannot accurately identify pollution sources or continuously monitor changes in air quality in the entire environment due to the lack of automation and positioning functions. In related technologies, most detection equipment does not have automatic navigation and positioning functions, resulting in incomplete monitoring results and the inability to track the source and concentration changes of harmful gases in real time. In addition, these devices often rely on manual operation, are inefficient, and cannot adapt to complex indoor environments, especially in scenes with dense furniture, where it is difficult to accurately detect harmful gases released from the surface of furniture. Summary of the invention

[0003] In order to solve the above technical problems, the purpose of the present invention is to provide an environmental gas monitoring method and system based on deep learning, which can accurately identify the location of wooden furniture and effectively monitor the concentration of harmful gases in real time to ensure the safety and health of the home environment.

[0004] The first technical solution adopted by the present invention is: a method for monitoring environmental gases based on deep learning, comprising the following steps:

[0005] An environmental monitoring mobile robot is constructed to obtain obstacle image data of the target environment and perform data enhancement processing to obtain enhanced wooden obstacle image data;

[0006] The TAA strategy is introduced to train the YOLOv8 target detection model and an improved YOLOv8 target detection model is constructed;

[0007] Based on the improved YOLOv8 target detection model, the enhanced wooden material obstacle image data is recognized and processed to obtain the wooden furniture information data of the target environment;

[0008] According to the wooden furniture information data of the target environment, the environmental monitoring mobile robot is controlled to move globally to monitor the gas in the target environment, and the harmful gas monitoring results of the target environment are obtained.

[0009] Furthermore, the environmental monitoring mobile robot specifically includes a fisheye camera, an ultrasonic distance sensor and an environmental air quality detector, wherein:

[0010] The fisheye camera is used to obtain obstacle information data of the target environment;

[0011] The ultrasonic distance sensor is used to obtain the distance value between the environment monitoring mobile robot and the obstacle;

[0012] The environmental air quality detector is used to monitor the gas quality in the target environment.

[0013] Furthermore, the step of constructing the environment monitoring mobile robot to obtain obstacle image data of the target environment and perform data enhancement processing to obtain enhanced wooden material obstacle image data specifically includes:

[0014] Constructing an environmental monitoring mobile robot and a target scene, wherein the target scene includes a plurality of sampling points;

[0015] The fisheye camera of the environment monitoring mobile robot collects the sampling points in the target scene from multiple angles to obtain obstacle image data of the target environment;

[0016] Performing data classification processing on obstacle image data of the target environment to obtain wooden material obstacle image data and non-wood material obstacle image data;

[0017] The wooden material obstacle image data is subjected to digital enhancement processing to obtain enhanced wooden material obstacle image data.

[0018] Furthermore, the step of performing digital enhancement processing on the wooden material obstacle image data to obtain enhanced wooden material obstacle image data specifically includes:

[0019] Performing digital increase and cropping processing on the wooden material obstacle image data to obtain cropped wooden material obstacle image data;

[0020] Scaling the cropped wooden obstacle image data to obtain scaled wooden obstacle image data;

[0021] Performing exposure enhancement processing on the scaled wooden material obstacle image data to obtain exposure enhanced wooden material obstacle image data;

[0022] The exposure enhanced wood material obstacle image data is randomly rotated and flipped to obtain enhanced wood material obstacle image data.

[0023] Furthermore, the step of introducing the TAA strategy to train the YOLOv8 target detection model and constructing an improved YOLOv8 target detection model specifically includes:

[0024] Pre-train the YOLOv8 target detection model using the pre-training data set to obtain a pre-trained YOLOv8 target detection model;

[0025] Based on the pre-trained YOLOv8 target detection model, the position loss function and classification loss function are introduced to build an optimized YOLOv8 target detection model;

[0026] The TAA strategy is introduced to dynamically adjust the ratio of positive and negative samples of the optimized YOLOv8 target detection model to obtain the improved YOLOv8 target detection model.

[0027] Furthermore, the expression of the TAA strategy is specifically as follows:

[0028] t=s α *u β

[0029] In the above formula, t represents the weighted score, s represents the prediction score of the corresponding category of the prediction box, u represents the intersection-and-union ratio between the prediction box and the true box, and α and β represent hyperparameters.

[0030] Furthermore, the step of performing recognition processing on the enhanced wooden material obstacle image data based on the improved YOLOv8 target detection model to obtain wooden furniture information data of the target environment specifically includes:

[0031] The enhanced wooden obstacle image data is input into the improved YOLOv8 target detection model;

[0032] Based on the backbone network module of the improved YOLOv8 target detection model, feature extraction processing is performed on the enhanced wooden obstacle image data to obtain wooden obstacle image feature data;

[0033] Based on the neck network module of the improved YOLOv8 target detection model, feature fusion processing is performed on the image feature data of wooden obstacles to obtain fused image feature data of wooden obstacles;

[0034] Based on the head network module of the improved YOLOv8 target detection model, the fused wooden material obstacle image feature data is classified and detected to obtain the wooden furniture information data of the target environment.

[0035] Further, the step of controlling the environment monitoring mobile robot to perform global movement to monitor the gas in the target environment according to the wooden furniture information data of the target environment to obtain the harmful gas monitoring result of the target environment specifically includes:

[0036] According to the wooden furniture information data of the target environment, the environmental air quality detector of the environmental monitoring mobile robot is controlled to collect the target environmental gas to obtain the harmful gas monitoring result of the current sampling point;

[0037] The ultrasonic distance sensor of the environmental monitoring mobile robot is used to obtain the distance value between the environmental monitoring mobile robot and the obstacle, perform global movement planning, and control the environmental monitoring mobile robot to move to the next sampling point to collect the target environmental gas;

[0038] Until all sampling points are traversed, the monitoring results of harmful gases in the target environment are obtained.

[0039] The second technical solution adopted by the present invention is: an environmental gas monitoring system based on deep learning, comprising:

[0040] The first module is used to construct an environmental monitoring mobile robot to obtain obstacle image data of the target environment and perform data enhancement processing to obtain enhanced wooden material obstacle image data;

[0041] The second module is used to introduce the TAA strategy to train the YOLOv8 target detection model and build an improved YOLOv8 target detection model;

[0042] The third module is used to identify and process the enhanced wooden material obstacle image data based on the improved YOLOv8 target detection model to obtain the wooden furniture information data of the target environment;

[0043] The fourth module is used to control the environmental monitoring mobile robot to move globally to monitor the gas in the target environment according to the wooden furniture information data of the target environment, and obtain the harmful gas monitoring results of the target environment.

[0044] The beneficial effects of the method and system of the present invention are as follows: the present invention acquires obstacle image data of the target environment by constructing an environmental monitoring mobile robot and performs data enhancement processing, thereby increasing the diversity of the target environment data, further introduces a TAA strategy to train the YOLOv8 target detection model, adopts a Task-Aligned Assigner (TAA) strategy, dynamically adjusts the ratio of positive and negative samples, and makes the model pay more attention to high-quality prediction frames. Based on the improved YOLOv8 target detection model, the enhanced wooden material obstacle image data is recognized and processed, which can realize accurate detection and classification of targets in the image. Finally, according to the wooden furniture information data of the target environment, the environmental monitoring mobile robot is controlled to move globally to monitor the gas in the target environment, and the concentration of harmful gases is monitored in real time and effectively to ensure the safety and health of the home environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1It is a flowchart of the steps of an environmental gas monitoring method based on deep learning of the present invention;

[0046] Figure 2 It is a structural block diagram of an environmental gas monitoring system based on deep learning of the present invention;

[0047] Figure 3 is a schematic diagram of wood identification results provided by a specific embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of formaldehyde data for detecting wooden furniture provided by a specific embodiment of the present invention;

[0049] Figure 5 It is a schematic diagram of formaldehyde data for detecting wood products in an office environment provided by a specific embodiment of the present invention. DETAILED DESCRIPTION

[0050] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The step numbers in the following embodiments are only provided for the convenience of explanation and description, and the order between the steps is not limited in any way. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0051] Reference Figure 1 The present invention provides an environmental gas monitoring method based on deep learning, which comprises the following steps:

[0052] S100, constructing an environment monitoring mobile robot to obtain obstacle image data of a target environment and perform data enhancement processing to obtain enhanced wooden material obstacle image data;

[0053] First of all, it should be noted that the environmental monitoring mobile robot specifically includes a fisheye camera, an ultrasonic distance sensor and an environmental air quality detector, wherein the fisheye camera is used to obtain obstacle information data of the target environment; the ultrasonic distance sensor is used to obtain the distance value between the environmental monitoring mobile robot and the obstacle; the environmental air quality detector is used to monitor the gas quality in the target environment.

[0054] In this embodiment, the robot is equipped with a fisheye camera and an ultrasonic distance sensor, the former is used to capture images with a wide viewing angle, and the latter is used for environmental perception and obstacle detection. MasterPi is equipped with a 120-degree wide-angle telephoto camera, with a viewing angle of up to 180° automatic surround and over 180° automatic pitch. With the flexible movement of the car, the visual range can reach 360° all-round without blind spots, ensuring that the robot can obtain a panoramic image of the indoor environment and provide high-quality image data for subsequent wooden furniture recognition; the luminous ultrasonic sensor equipped with an IIC interface not only does not require too much CPU resources to achieve detection and obstacle avoidance, but also can adjust the automatic obstacle avoidance distance and the luminous color of the ultrasonic wave in real time through the mobile phone APP, so that the robot can effectively perceive obstacles in the surrounding environment, achieve autonomous obstacle avoidance, and ensure safe movement in complex environments; it integrates a highly sensitive environmental air quality detector, which can monitor the concentration of harmful gases such as formaldehyde, TVOC, C6H6 in real time, and display temperature and humidity data at the same time. It ensures that the robot can monitor and report the concentration of harmful gases in the home environment in real time, discover potential health risks in a timely manner, and provide users with a safer and healthier living environment.

[0055] Furthermore, it is also necessary to explain that an environmental monitoring mobile robot and a target scene are constructed, wherein the target scene includes a plurality of sampling points; the sampling points in the target scene are collected from multiple angles based on the fisheye camera of the environmental monitoring mobile robot to obtain obstacle image data of the target environment; data classification processing is performed on the obstacle image data of the target environment to obtain wooden material obstacle image data and non-wooden material obstacle image data; and digital enhancement processing is performed on the wooden material obstacle image data to obtain enhanced wooden material obstacle image data.

[0056] In this embodiment, in the process of constructing the furniture data set, the embodiment of the present invention adopts a random sampling method to ensure the diversity and representativeness of the data. First, random sampling is performed, and furniture is randomly selected at each sampling point for data collection. This method helps to avoid subjective bias and ensure that the data set can truly reflect the diversity of furniture on the market. In order to ensure the comprehensiveness of the data, sampling is performed in different environments (such as living rooms, bedrooms, offices, etc.), covering furniture of different styles and materials, and then multi-angle data collection is performed. Multi-angle data collection is performed on each piece of furniture, including multiple perspectives such as the front, side, back and top. This ensures that the model can accurately identify furniture from different angles in practical applications, and uses a high-resolution camera for shooting to ensure image quality so that the model can extract more detailed features. Finally, data classification is performed, and the data set is divided into wood and non-wood categories. The wood data set contains images of various wooden furniture, while the non-wood data set contains other types of furniture (such as metal, plastic, etc.). In this way, the model can be trained to more accurately distinguish wooden furniture from other types of furniture.

[0057] Furthermore, the YOLOv8 target detection algorithm is used in the model. In order to further improve the performance of the model, the embodiment of the present invention also adopts a data enhancement method. First, data cropping is performed to randomly crop a part of the image to simulate the situation of furniture in different positions and sizes, and then data scaling is performed to enlarge or reduce the image to simulate the visual effect of furniture at different distances. Data exposure enhancement is further performed to adjust the brightness and contrast of the image to simulate scenes under different lighting conditions, and then data rotation is performed to randomly rotate the image to simulate the visual effect of furniture at different angles. Finally, data flipping is performed to flip the image horizontally or vertically to increase the diversity of the data.

[0058] S200, introduce the TAA strategy to train the YOLOv8 target detection model and build an improved YOLOv8 target detection model;

[0059] Specifically, the YOLOv8 target detection model is pre-trained with a pre-training data set to obtain a pre-trained YOLOv8 target detection model; based on the pre-trained YOLOv8 target detection model, a position loss function and a classification loss function are introduced to construct an optimized YOLOv8 target detection model; the TAA strategy is introduced to dynamically adjust the ratio of positive and negative samples of the optimized YOLOv8 target detection model to obtain an improved YOLOv8 target detection model.

[0060] In this embodiment, the YOLOv8 model is first pre-trained on a large dataset (such as COCO) to learn the features of general objects. First, the model is fine-tuned on the collected furniture dataset so that the model can more accurately identify wooden furniture, and then the loss function is optimized, using a comprehensive loss function, including position loss (box_loss), classification loss (cls_loss), etc., to ensure the accuracy of the model in position and category, and finally task alignment is performed, using the Task-AlignedAssigner (TAA) strategy to dynamically adjust the ratio of positive and negative samples so that the model pays more attention to high-quality prediction boxes.

[0061] Further model evaluation and optimization are carried out. First, the model's performance is tested. Multiple days of testing are carried out in different indoor environments to evaluate the robot's performance and stability in complex environments. Then data analysis is carried out to record the detection results in different spaces, different heights, and different concentrations of harmful gases. The accuracy and stability of the robot are analyzed. Finally, the model is optimized and improved based on the experimental results and user feedback to improve the recognition accuracy and speed.

[0062] S300, based on the improved YOLOv8 target detection model, the enhanced wooden material obstacle image data is identified and processed to obtain wooden furniture information data of the target environment;

[0063] Specifically, the enhanced wooden material obstacle image data is input into the improved YOLOv8 target detection model; based on the backbone network module of the improved YOLOv8 target detection model, feature extraction processing is performed on the enhanced wooden material obstacle image data to obtain wooden material obstacle image feature data; based on the neck network module of the improved YOLOv8 target detection model, feature fusion processing is performed on the wooden material obstacle image feature data to obtain fused wooden material obstacle image feature data; based on the head network module of the improved YOLOv8 target detection model, classification detection is performed on the fused wooden material obstacle image feature data to obtain wooden furniture information data of the target environment, such as Figure 3 shown.

[0064] First, the structure of the improved YOLOv8 target detection model in the embodiment of the present invention is described. The network structure of YOLOv8 is mainly composed of three parts: Backbone, Neck and Head. These three parts work together to realize the function of target detection.

[0065] Backbone is responsible for extracting features of the input image. YOLOv8's Backbone draws on the idea of ​​the CSP (compact and separate) module, but replaces the C3 module in YOLOv5 with the C2f module to achieve further lightweight while maintaining the idea of ​​CSP. In addition, YOLOv8 also uses the SPPF (SPPFast) module to enhance feature extraction capabilities. In Backbone, convolutional modules (Conv) and C2f modules are used alternately to extract features through convolution operations, and feature fusion and dimensionality reduction are performed through the Bottleneck structure in the C2f module. Compared with the C3 module, the C2f module has fewer parameters and better feature extraction capabilities, which helps to improve the performance and efficiency of the model.

[0066] The Neck part is responsible for fusing the feature maps of different scales extracted by Backbone to enhance the feature representation capability. The Neck of YOLOv8 adopts the PANet (Path Aggregation Network) structure, which is a bidirectional pathway network that introduces a bottom-up path, making it easier for the underlying information to be transmitted to the top of the high-level layer. In PANet, the feature maps output by Backbone are subjected to upsampling, channel fusion and other operations to form feature maps of multiple scales. These feature maps are then sent to DetectHead for target detection and classification.

[0067] The head part is responsible for the final target detection and classification tasks. YOLOv8 uses a Decoupled Head structure similar to YOLOX, which separates the regression branch and the prediction branch, making the network training and reasoning more efficient. In the Decoupled Head, the regression branch uses the DFL (Distribution Focal Loss) strategy for coordinate prediction, transforming the coordinates into a distribution to solve the problem of blurred boundaries. At the same time, the classification branch uses global average pooling to classify each feature map.

[0068] The network structure diagram of YOLOv8 shows its complete process from input image to output detection result. Through the collaborative work of Backbone, Neck and Head, YOLOv8 can accurately detect and classify objects in images. Its innovations include using C2f module instead of C3 module, using SPPF module to enhance feature extraction capability, and introducing DecoupledHead structure to improve training and reasoning efficiency. These improvements make YOLOv8 have higher performance and flexibility in the field of object detection.

[0069] In addition, the improved YOLOv8 target detection model algorithm of the embodiment of the present invention needs to be explained. First, it should be noted that YOLOv8 automatically adjusts the shape it uses to predict the position of the object, so that it can more accurately match the shape of the actual object. In order to further reduce false detections, by optimizing these shapes, YOLOv8 can reduce false positives, that is, mistaking the background for the object.

[0070] Further including the comprehensive loss function, YOLOv8 uses a special formula to measure the accuracy of the prediction, which takes into account the correctness of the position and category of the object, as well as the regularization strategy: YOLOv8 also uses some techniques to avoid overfitting, that is, making the model too dependent on the training data and not being well applied to new data. Among them, Task-Aligned Assigner (TAA) is a strategy for dynamically adjusting the distribution ratio of positive and negative samples. Its core idea is to dynamically select positive samples based on the performance of classification and regression during training. This method is called "task alignment" because it tries to make the selection of positive samples more in line with the needs of the final task.

[0071] The principle of task alignment is further explained:

[0072] TAA uses a weighted score to measure the degree of alignment between each predicted box and the ground truth (GT). The weighted score is determined by the classification score and the intersection over union (IoU) between the predicted box and the GT box. The formula for calculating the weighted score is:

[0073] t=s α *u β

[0074] In the above formula, t represents the weighted score, s represents the prediction score of the corresponding category of the prediction box, u represents the intersection-and-union ratio between the prediction box and the true box, and α and β represent hyperparameters.

[0075] The weighted score combines the classification score and the IoU value, reflects the degree of task alignment, and sets the hyperparameters. α and β are used as hyperparameters to adjust the relative importance of the classification score and IoU. In this description, the embodiment of the present invention does not directly mention these two hyperparameters, but directly calculates the weighted score through the formula.

[0076] Specific steps:

[0077] 1) Calculate the alignment score:

[0078] Based on the classification score of the predicted box and the IoU between the predicted box and the GT box, the alignment score t of each predicted box is calculated.

[0079] 2) Screening candidate positive samples:

[0080] Determine which prediction boxes’ center points are inside the GT box, and only these prediction boxes are eligible to become positive samples.

[0081] 3) Select positive samples:

[0082] From the candidate positive samples, the top K prediction boxes are selected as positive samples according to the alignment score t.

[0083] The remaining predicted boxes are regarded as negative samples.

[0084] Through such a strategy, TAA can help the model pay more attention to high-quality prediction boxes, that is, those with higher classification scores and better positioning, thereby improving the overall performance of the model.

[0085] Finally, transfer learning and pre-training models are performed. YOLOv8 is first trained on a large dataset to learn how to recognize various types of objects and then fine-tuned. Then, for specific tasks (such as identifying furniture), YOLOv8 will perform additional training on a small amount of related data to improve its performance on the task.

[0086] S400, controlling the environment monitoring mobile robot to move globally to monitor the gas in the target environment according to the wooden furniture information data of the target environment, and obtaining the monitoring result of harmful gas in the target environment.

[0087] Specifically, according to the wooden furniture information data of the target environment, the environmental air quality detector of the environmental monitoring mobile robot is controlled to collect the target environmental gas to obtain the harmful gas monitoring result of the current sampling point; the distance value between the environmental monitoring mobile robot and the obstacle is obtained through the ultrasonic distance sensor of the environmental monitoring mobile robot, and global movement planning is performed to control the environmental monitoring mobile robot to move to the next sampling point to collect the target environmental gas; until all sampling points are traversed, the harmful gas monitoring result of the target environment is obtained.

[0088] In this embodiment, multi-sensor fusion is first performed, combining a fisheye camera and an ultrasonic distance sensor to obtain comprehensive environmental information and improve the robot's ability to perceive the environment; then motion control optimization is performed, the motion control algorithm is optimized, smooth turning, acceleration and deceleration and other actions are achieved, and the robot's stability and safety in complex environments are improved.

[0089] More specifically:

[0090] 1) Route planning: Calculate the best path from the current location to the target location based on the environmental information provided by the YOLO model and ultrasonic sensors, avoiding obstacles such as tables and chairs.

[0091] 2) Action decision: Based on the results of path calculation, the robot decides on its next action, such as going straight, turning, accelerating or decelerating. At the same time, drivers and shock absorbers are installed in the power system to further reduce the impact and vibration of the machine. The driver provides more precise motor control and response performance, while the shock absorber can reduce the vibration transmitted to the surrounding environment during the movement of the machine to optimize the smoothness of the action execution.

[0092] 3) Obstacle avoidance strategy: During the movement, if the YOLO model or ultrasonic sensor detects a new obstacle, the path planning is updated in real time and an obstacle avoidance strategy is adopted to ensure the robot’s safe passage.

[0093] 4) Feedback control, through the feedback control system, PID controller, adjust the speed and direction of the robot to ensure that it travels stably along the predetermined path and responds to changes in the external environment in a timely manner.

[0094] Finally, the method of the present invention is summarized and described, which is specifically as follows:

[0095] 1) Movement and positioning: The robot uses a fisheye camera to obtain images of the surrounding environment and uses an ultrasonic distance sensor to sense obstacles, thereby planning a movement path and achieving autonomous navigation and obstacle avoidance. The robot will move autonomously in the indoor environment, looking for and approaching wooden furniture.

[0096] 2) Wooden furniture recognition: When the robot approaches wooden furniture, it uses a fisheye camera to take pictures of the furniture. These pictures are fed into the YOLOv8 algorithm for processing to identify wooden furniture in the image.

[0097] 3) Gas concentration detection: Once wooden furniture is identified, the robot will stop and use the integrated high-sensitivity air quality detector to detect the concentration of harmful gases near the furniture, such as formaldehyde, TVOC, etc.

[0098] 4) Data processing and output: The detected data will be transmitted to the data processing center in real time or directly sent to the user's mobile device via a wireless network. Users can view the identification results of wooden furniture and the concentration data of harmful gases through the application.

[0099] 5) Repeat the steps. After completing the inspection of the current furniture, the robot will continue to move to the next target position and repeat the above process until the inspection of the entire room or set area is completed.

[0100] 6) Remote control: Users can also remotely control the robot through mobile devices, such as adjusting the monitoring area or replanning the path.

[0101] Furthermore, the embodiments of the present invention are verified by experimental simulation:

[0102] First, home environment monitoring and furniture recognition are carried out. In this embodiment, the intelligent robot is deployed in a typical home environment, including a living room, bedroom, and kitchen. The robot is equipped with a fisheye camera and an ultrasonic distance sensor to ensure that it can obtain comprehensive visual information and perceive surrounding obstacles. At the same time, the robot is also integrated with a high-precision air quality detector for real-time monitoring of the concentration of harmful gases such as formaldehyde, TVOC, and C6H6, as well as environmental parameters such as temperature and humidity. The robot begins to move autonomously on a predetermined path, using a fisheye camera to capture images, while the ultrasonic sensor continuously scans the front to detect any obstacles. The captured images are processed by the deep learning model YOLO to identify wooden furniture in the image. At the same time, the harmful gas concentration data collected by the air quality detector is also analyzed in real time, and the processed data is transmitted to the data center in real time via a wireless network. Users can view the recognition results of wooden furniture and the concentration data of harmful gases through the application on the mobile device, such as Figure 4 As shown, in addition, when an abnormal situation is detected, the user can manually control the robot through the application, such as adjusting the monitoring area or replanning the path.

[0103] Then, the office environment is monitored. In the office environment, the intelligent robot is set to regular patrol mode to monitor the air quality in the employee's work area. During the patrol process, the robot uses fisheye cameras and ultrasonic distance sensors to avoid collisions with objects such as desks, chairs, and filing cabinets. At the same time, it uses a high-precision air quality detector to continuously monitor the concentration of harmful gases such as formaldehyde and TVOC. The robot moves autonomously in the office along a predetermined route while collecting environmental data. It uses a deep learning algorithm to identify wooden furniture in the office and monitors the concentration of harmful gases through an air quality detector. The monitoring data is uploaded to the server in real time. Office managers can view the monitoring report through computer software to understand the health of the office environment, such as Figure 5 As shown, if necessary, managers can use software to adjust the robot's movement path or immediately dispatch the robot to a specific area for detailed inspection.

[0104] In summary, the wooden furniture recognition accuracy of the embodiment of the present invention is: in a series of tests, the robot's accuracy in recognizing wooden furniture reached more than 95%, and the harmful gas concentration monitoring accuracy, the robot's harmful gas concentration monitoring error under different environmental conditions was less than ±5%, indicating that it performed well in monitoring accuracy.

[0105] Reference Figure 2 , an environmental gas monitoring system based on deep learning, comprising:

[0106] The first module 201 is used to construct an environment monitoring mobile robot to obtain obstacle image data of a target environment and perform data enhancement processing to obtain enhanced wooden material obstacle image data;

[0107] The second module 202 is used to introduce the TAA strategy to train the YOLOv8 target detection model and build an improved YOLOv8 target detection model;

[0108] The third module 203 is used to identify and process the enhanced wooden material obstacle image data based on the improved YOLOv8 target detection model to obtain wooden furniture information data of the target environment;

[0109] The fourth module 204 is used to control the environmental monitoring mobile robot to perform global movement to monitor the gas in the target environment according to the wooden furniture information data of the target environment, and obtain the harmful gas monitoring results of the target environment.

[0110] In summary, the robot constructed by the embodiment of the present invention can not only shuttle freely in the indoor environment, but also accurately identify the type and location of wooden furniture. This intelligent robot is also equipped with advanced environmental monitoring sensors, which can detect the concentration of harmful gases in indoor air in real time, such as formaldehyde, benzene, etc. When an abnormally high level of harmful gases is detected, the robot will automatically issue an early warning to notify the user and remind him to take corresponding measures. Through continuous monitoring and data analysis, the device can generate detailed air quality reports to help users understand the changing trends of the indoor environment. This integrated solution not only improves the efficiency and accuracy of detection, but also reduces the dependence on manual operation, especially in the case of covering a large area and complex environment. Users can not only obtain air quality data in real time, but also improve indoor air quality through the suggestions provided by the robot. This innovative design provides higher security for families and offices that pursue a healthy living environment, creating a safer and healthier living and working space for users.

[0111] The contents of the above method embodiments are all applicable to the present system embodiments. The functions specifically implemented by the present system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0112] The above is a specific description of the preferred implementation of the present invention, but the invention is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A method for monitoring environmental gases based on deep learning, characterized in that: The following steps are involved: An environmental monitoring mobile robot is constructed to obtain obstacle image data of the target environment and perform data enhancement processing to obtain enhanced wooden obstacle image data; The TAA strategy is introduced to train the YOLOv8 target detection model and an improved YOLOv8 target detection model is constructed; Based on the improved YOLOv8 target detection model, the enhanced wooden material obstacle image data is recognized and processed to obtain the wooden furniture information data of the target environment; According to the wooden furniture information data of the target environment, the environmental monitoring mobile robot is controlled to move globally to monitor the gas in the target environment, and the harmful gas monitoring results of the target environment are obtained.

2. According to claim 1, a method for monitoring environmental gases based on deep learning is characterized in that: The environmental monitoring mobile robot specifically includes a fisheye camera, an ultrasonic distance sensor and an environmental air quality detector, wherein: The fisheye camera is used to obtain obstacle information data of the target environment; The ultrasonic distance sensor is used to obtain the distance value between the environment monitoring mobile robot and the obstacle; The environmental air quality detector is used to monitor the gas quality in the target environment.

3. According to claim 2, a method for monitoring environmental gases based on deep learning, characterized in that: The step of constructing an environment monitoring mobile robot to obtain obstacle image data of a target environment and perform data enhancement processing to obtain enhanced wooden material obstacle image data specifically includes: Constructing an environmental monitoring mobile robot and a target scene, wherein the target scene includes a plurality of sampling points; The fisheye camera of the environment monitoring mobile robot collects sample points in the target scene from multiple angles to obtain obstacle image data of the target environment; Performing data classification processing on obstacle image data of the target environment to obtain wooden material obstacle image data and non-wood material obstacle image data; The wood material obstacle image data is digitally enhanced to obtain enhanced wood material obstacle image data.

4. According to claim 3, a method for monitoring environmental gases based on deep learning is characterized in that: The step of performing digital enhancement processing on the wooden material obstacle image data to obtain enhanced wooden material obstacle image data specifically includes: Performing digital increase and cropping processing on the wooden material obstacle image data to obtain cropped wooden material obstacle image data; Scaling the cropped wooden obstacle image data to obtain scaled wooden obstacle image data; Performing exposure enhancement processing on the scaled wooden material obstacle image data to obtain exposure enhanced wooden material obstacle image data; The exposure enhanced wood material obstacle image data is randomly rotated and flipped to obtain enhanced wood material obstacle image data.

5. The method for monitoring environmental gas based on deep learning according to claim 4, characterized in that: The step of introducing the TAA strategy to train the YOLOv8 target detection model and constructing an improved YOLOv8 target detection model specifically includes: Pre-train the YOLOv8 target detection model using the pre-training data set to obtain a pre-trained YOLOv8 target detection model; Based on the pre-trained YOLOv8 target detection model, the position loss function and classification loss function are introduced to build an optimized YOLOv8 target detection model; The TAA strategy is introduced to dynamically adjust the ratio of positive and negative samples of the optimized YOLOv8 target detection model to obtain the improved YOLOv8 target detection model.

6. According to claim 5, a method for monitoring environmental gases based on deep learning is characterized in that: The expression of the TAA strategy is specifically as follows: t=s α *u β In the above formula, t represents the weighted score, s represents the prediction score of the corresponding category of the prediction box, u represents the intersection-and-union ratio between the prediction box and the true box, and α and β represent hyperparameters.

7. The method for monitoring environmental gases based on deep learning according to claim 6, characterized in that: The step of identifying and processing the enhanced wooden material obstacle image data based on the improved YOLOv8 target detection model to obtain wooden furniture information data of the target environment specifically includes: The enhanced wooden obstacle image data is input into the improved YOLOv8 target detection model; Based on the backbone network module of the improved YOLOv8 target detection model, feature extraction processing is performed on the enhanced wooden obstacle image data to obtain wooden obstacle image feature data; Based on the neck network module of the improved YOLOv8 target detection model, feature fusion processing is performed on the image feature data of wooden obstacles to obtain fused image feature data of wooden obstacles; Based on the head network module of the improved YOLOv8 target detection model, the fused wooden material obstacle image feature data is classified and detected to obtain the wooden furniture information data of the target environment.

8. The method for monitoring environmental gases based on deep learning according to claim 7, characterized in that: The step of controlling the environment monitoring mobile robot to perform global movement to monitor the gas in the target environment according to the wooden furniture information data of the target environment to obtain the harmful gas monitoring result of the target environment specifically includes: According to the wooden furniture information data of the target environment, the environmental air quality detector of the environmental monitoring mobile robot is controlled to collect the target environmental gas to obtain the harmful gas monitoring result of the current sampling point; The ultrasonic distance sensor of the environmental monitoring mobile robot is used to obtain the distance value between the environmental monitoring mobile robot and the obstacle, perform global movement planning, and control the environmental monitoring mobile robot to move to the next sampling point to collect the target environmental gas; Until all sampling points are traversed, the monitoring results of harmful gases in the target environment are obtained.

9. An environmental gas monitoring system based on deep learning, characterized in that: Includes the following modules: The first module is used to construct an environmental monitoring mobile robot to obtain obstacle image data of the target environment and perform data enhancement processing to obtain enhanced wooden material obstacle image data; The second module is used to introduce the TAA strategy to train the YOLOv8 target detection model and build an improved YOLOv8 target detection model; The third module is used to identify and process the enhanced wooden material obstacle image data based on the improved YOLOv8 target detection model to obtain the wooden furniture information data of the target environment; The fourth module is used to control the environmental monitoring mobile robot to move globally to monitor the gas in the target environment according to the wooden furniture information data of the target environment, and obtain the harmful gas monitoring results of the target environment.

Images