People and vehicle safety monitoring system of unmanned vehicle washing machine

Abstract

The invention belongs to the technical field of unmanned car washer safety, and discloses a man-car safety monitoring system of an unmanned car washer. The system is composed of a sensor module, a camera module, a controller module and an alarm module. The sensor module comprises a human body infrared sensor, a vehicle position sensor and the like and can detect personnel approaching and vehicle states. The camera module comprises a high-definition camera and an infrared camera and is used for real-time monitoring. Personnel behavior monitoring utilizes an image recognition technology and a deep learning algorithm to predict behavior intentions. The vehicle safety monitoring comprises entry detection, door and window detection, vehicle type identification and water pressure adjustment, damage detection and the like. The system has a remote monitoring function, and data are transmitted to a remote monitoring center, so that management personnel and vehicle owners can know the operation state and alarm information conveniently. According to the system, the safety and reliability of the unmanned car washing machine are effectively improved, a powerful guarantee is provided for unmanned car washing service, and intelligent and safe development of the car washing industry is promoted.

Description

Technical Field

[0001] The present invention relates to the technical field of safety of unmanned car washers, and in particular to a human and vehicle safety monitoring system for unmanned car washers. Background Art

[0002] An unmanned car washing machine is a device that uses automation technology and intelligent control systems to achieve vehicle washing. It can automatically identify the vehicle model and complete the steps of car washing such as spraying, waxing, brushing, rinsing and air drying through an automated process. This type of car washing machine is efficient, energy-saving and environmentally friendly. It can provide 24-hour uninterrupted service, with low operating costs and high return on investment. It is becoming an important development direction for the car washing service industry. With the continuous advancement of science and technology, unmanned car washing machines are gradually emerging in the field of car washing. Unmanned car washing machines have been favored by more and more consumers due to their advantages such as high efficiency, convenience and labor cost saving. However, the current unmanned car washing machines still have some shortcomings in the safety monitoring of people and vehicles.

[0003] First, existing unmanned car washers usually rely on a single type of sensor or camera to monitor human behavior. The data source is relatively single and cannot fully and accurately capture human behavior information. This leads to inaccurate predictions of human behavior intentions, making it difficult to take effective measures to prevent dangerous behavior intentions in a timely manner.

[0004] Secondly, the existing unmanned car wash machines have deficiencies in real-time response and safety assurance. In terms of personnel safety, there is a lack of effective real-time monitoring mechanism, and when dangerous behaviors occur, the alarm cannot be quickly issued and the car wash machine cannot be stopped, which poses a safety hazard.

[0005] Finally, existing unmanned car washers lag behind in model updates. They cannot collect new data in a timely manner to update and optimize the model, resulting in the gradual decline of model performance over time and the inability to adapt to changes in human behavior and the environment. Summary of the invention

[0006] The main purpose of the present invention is to provide a human and vehicle safety monitoring system for an unmanned car washing machine, which can effectively solve the problems mentioned in the background technology.

[0007] To achieve the above object, the technical solution adopted by the present invention is: A human and vehicle safety monitoring system for an unmanned car washing machine, the human and vehicle safety monitoring system comprising: Sensor module: Human infrared sensors installed at the entrance and exit of the car washing machine are used to detect whether someone is close to the car washing machine; vehicle position sensors installed inside the car washing machine are used to detect the position and movement status of the vehicle; water pressure sensors installed on the water gun of the car washing machine are used to detect the water pressure of the water gun; temperature sensors installed inside the car washing machine are used to detect the temperature inside the car washing machine; proximity sensors installed in various places inside the car washing machine are used to detect the proximity between people and equipment; Camera module: HD cameras installed inside and outside the car wash machine to monitor the status of vehicles and personnel in real time; infrared cameras installed inside the car wash machine to monitor the status of vehicles and personnel in dark environments; Controller module: Microprocessor: responsible for receiving data from sensors and cameras, and processing and analyzing them; Memory: used to store monitoring data and system parameters; Communication module: used to communicate with the remote monitoring center to achieve remote monitoring and management; Alarm module: an audible and visual alarm installed inside and outside the car washer, used to sound and light alarms when abnormal situations occur; a text message alarm used to send text message alarms to car owners and managers when abnormal situations occur; The monitoring scheme of the above-mentioned human and vehicle safety monitoring system is divided into personnel safety monitoring, vehicle safety monitoring and remote monitoring, among which: Personnel safety monitoring includes: Personnel approach detection: Human infrared sensors detect whether someone is approaching at the entrance and exit of the car washing machine. Once a person is detected approaching, the system automatically issues an audible and visual alarm to remind the person to stay away from the car washing machine; when a person enters the car washing machine, the proximity sensor detects the degree of proximity between the person and a specific part of the equipment in real time. If a person approaches a part of the equipment that may cause danger, the system immediately issues an alarm and prevents the person from getting closer; Personnel behavior monitoring: High-definition cameras and infrared cameras monitor the behavior of people entering the car wash machine in real time. The system uses image recognition technology to analyze the behavioral characteristics of people. If abnormal behavior is detected, the system will immediately issue an audible and visual alarm and stop the operation of the car wash machine. At the same time, the alarm information is sent to the remote monitoring center through the communication module to notify the management personnel to handle it in time. Vehicle safety monitoring includes: Vehicle entry detection: The vehicle position sensor detects the position and movement status of the vehicle when it enters the car wash machine; if the vehicle position is incorrect or the movement status is abnormal, the system immediately issues an audible and visual alarm and stops the operation of the car wash machine; Vehicle door and window detection: Use high-definition cameras to detect the vehicle's doors and windows; if the doors and windows are found not closed tightly, the system will prompt the owner to confirm that the vehicle's doors and windows are closed tightly through voice, and at the same time suspend the car washing process and wait for the owner to handle it; Vehicle type recognition and water pressure adjustment: The high-definition camera uses image recognition technology to identify vehicles of different brands and models; the system automatically adjusts the water pressure of the water gun based on the pre-stored vehicle parameters of different brands and models; Vehicle damage detection: During the car wash process, the high-definition camera continuously scans the vehicle surface; once scratches, dents and other damages are found on the vehicle, the system immediately records them and notifies the owner; Remote monitoring includes: Data transmission: The data collected by sensors and cameras are transmitted to the remote monitoring center in real time through the communication module; Remote management: Managers can view the operating status and monitoring screen of the car wash machine in real time through the remote monitoring center; car owners can receive the operating status and alarm information of the car wash machine through mobile phone APP or SMS.

[0008] Preferably, the person is close to the detection: Assume that the infrared intensity detected by the human infrared sensor is I , set a threshold T 1 ,when I>T 1 When someone approaches, it determines that someone is approaching; when it detects that someone is approaching, the sound and light alarm device is triggered, emitting a sound of a specific frequency and loudness and a light of a specific color and flashing frequency; Assume the distance detected by the proximity sensor is d , set a safe distance threshold T 2 ,when d<T 2 When it is detected that a person is too close to a dangerous device, the alarm device is triggered, emitting a sound of a specific frequency and loudness and a light of a specific color and flashing frequency, while an electromagnetic lock is used to prevent the person from getting closer.

[0009] Preferably, the personnel behavior monitoring includes a subsystem that uses image recognition technology to analyze personnel behavior characteristics and predict personnel behavior intentions. The system includes a data acquisition module, a data annotation module, a feature extraction module, a behavior sequence modeling module, a behavior intention prediction model training module, a real-time behavior intention prediction and response module, and a model update and optimization module.

[0010] Preferably, the specific implementation of each component module of the personnel behavior intention prediction subsystem is as follows: The data acquisition module is used to continuously collect image data of people in the car washing machine using the high-definition camera and infrared camera installed in the unmanned car washing machine, as well as to collect sensor data; wherein, the image data is collected using the high-definition camera and infrared camera at a fixed frame rate, denoted as ,in Indicates time; proximity sensor uses time intervals The distance between the detection personnel and the equipment is recorded as ; The data annotation module manually annotates the collected image data, classifies human behaviors into different categories, and annotates the corresponding behavioral intentions; The feature extraction module uses a convolutional neural network (CNN) to extract features from images and extracts additional features in combination with sensor data. Perform feature extraction and record the extracted features as , using the pre-trained network architecture ResNet or VGG, the expression is ; Combined with proximity sensor data , extract additional features , the expression is ,in is a function that defines features based on distance; The behavior sequence modeling module uses a recurrent neural network (RNN), namely a long short-term memory network (LSTM) or a gated recurrent unit (GRU) to model the behavior sequence; the time window size is set to , the continuous The frame image features and the corresponding additional features form a sequence, recorded as ; Use LSTM or GRU to sequence For modeling, the output of LSTM is , the expression is ; The behavior intention prediction model training module builds a behavior intention prediction model and inputs the extracted features into the model for training; a multi-layer perceptron (MLP) is built as a classifier and the output of the LSTM is input into the model for training; Input into MLP to get the behavior intention prediction result , the expression is ; Define the cross entropy loss function ,in is the real behavioral intention label, is the predicted behavioral intention; Adam optimizer is selected to train the model. During the training process, the model parameters are continuously adjusted to minimize the loss function; specifically, for the model parameters , in each iteration, according to the current loss function gradient and learning rate Update the parameters, i.e. ; The real-time behavior intention prediction and response module performs real-time feature extraction and behavior intention prediction during the operation of the unmanned car wash machine, and takes corresponding measures based on the prediction results; during the operation of the unmanned car wash machine, it collects image data and sensor data in real time, performs feature extraction and behavior sequence modeling, and obtains the current behavior intention prediction results If the prediction result indicates that the person has dangerous behavior intentions, an audible and visual alarm will be issued immediately, the car washing machine will be stopped, and the remote monitoring center will be notified; The model update and optimization module continuously collects new data to update and optimize the model; continuously collects new image data and sensor data and annotates them; regularly uses new data to update and optimize the model, using the incremental learning method. Specifically, the real label and prediction result of the new data are and , the true labels and prediction results of the old data are and , the learning rate parameter is and ; During the update process, for the model parameters , first according to the loss function gradient and learning rate of the new data To update, , and then combine the loss function gradient and learning rate of the old data Further updates, namely ; At the same time, the hyperparameters of the model can be adjusted to improve the performance of the model.

[0011] Preferably, in the behavior intention prediction model training module, the multi-layer perceptron (MLP) consists of an input layer, a hidden layer and an output layer; the input layer receives the output features from LSTM or GRU, the hidden layer further processes the features through a combination of multiple neurons, and the output layer outputs the prediction results of the behavior intention; the number of nodes in the input layer is equal to the output feature dimension of LSTM or GRU; the activation function selects the ReLU function as the activation function of the hidden layer, and the output layer uses the softmax function for multi-classification; the calculation method of the cross entropy loss function is to compare the probability of each behavior intention category predicted by the model with the true label and calculate the cross entropy between them; assuming that the probability distribution of the behavior intention predicted by the model is , the true label is ,in The number of categories representing behavioral intentions, the calculation formula of the cross entropy loss function is .

[0012] Preferably, the model update and optimization module adopts an incremental learning method, specifically, when using mini-batch gradient descent for incremental learning, a fixed mini-batch size is set, and for newly collected data, it is randomly shuffled and divided into several mini-batches; for each mini-batch of data, it is first input into the current model to obtain the model's prediction result; then, the loss function between the prediction result and the true label is calculated, and the cross entropy loss function can continue to be used here; next, the gradient of the loss function to the model parameters is calculated by the back propagation algorithm; according to the gradient and the learning rate, the model parameters are updated; for the true label and the prediction result of the new data, and , the cross entropy loss function of the new data is recorded as ,in is the predicted probability distribution of new data, is the true label of the new data; when considering the influence of old data, the forgetting factor is used to weight the loss function of the old data. Assume that the loss function of the old data is , the loss function for the new data is , forgetting factor is (0< < 1), the total loss function is ; Then, the gradient is calculated based on this total loss function and the model parameters are updated.

[0013] Preferably, the specific implementation method of the vehicle safety monitoring is: Vehicle entry detection: The vehicle position sensor uses a laser rangefinder sensor or an ultrasonic sensor to determine the distance between the vehicle and the sensor by emitting a specific signal and measuring the reflection time, and determines the vehicle position and motion state based on the measurement results of multiple sensors; laser rangefinder sensors are installed on both sides of the car wash entrance to measure the width and position of the vehicle, and ultrasonic sensors are installed on the internal track to measure the forward distance and speed; when it is detected that the vehicle position is incorrect or the motion state is abnormal, the system emits a red flashing light and a high-frequency alarm to sound and light alarms, and stops the operation of the car wash machine motor, water pump and other equipment by controlling the power supply; Vehicle door and window detection: Install a high-definition camera at the top center or on both sides of the car wash machine to collect vehicle images and transmit them to the image processing system; use a deep learning target detection algorithm to process images, pre-process to improve image quality, and then use a pre-trained model to detect the state of doors and windows. Set a threshold based on the angle between the doors and windows and the vehicle body to determine whether they are closed tightly. If they are not closed tightly, mark the position and calculate the degree of looseness. If it is detected that they are not closed tightly, the owner will be prompted to confirm through voice, and the car wash program will be suspended at the same time. The program will be restarted after the owner confirms that they are closed tightly. Vehicle model recognition and water pressure adjustment: Use high-definition cameras on the top and side of the car wash machine to collect vehicle images, and use deep learning image classification algorithms to identify vehicle models. When a vehicle enters the car wash machine, the camera collects images and inputs the trained model to output brand and model information; establish a vehicle parameter database to store the shell material parameters and suitable water pressure ranges of different brands and models. After the system recognizes the model, it queries the database to obtain the suitable water pressure range, adjusts the water pressure by controlling the speed of the water pump motor or the valve opening, and uses the PID control algorithm to calculate the control signal based on the difference between the current and target water pressures to adjust the water pressure to make it close to the target water pressure; Vehicle damage detection: During the car wash process, high-definition cameras are installed on the top, sides, and bottom of the car wash machine to collect images of the vehicle surface and transmit them to the image processing system. The images are processed using image recognition technology, scratches are detected using the Canny edge detection algorithm, and dents are detected using an image segmentation algorithm combined with deep learning U-Net. When damage is detected, the damage location, type, and severity information are recorded and stored in a database, and the owner is notified via voice prompts, SMS notifications, or mobile APP push notifications.

[0014] Preferably, the specific implementation method of the remote monitoring is: Data transmission: Use 4G / 5G, Wi-Fi or wired network modules, configure the correct network parameters to establish a connection with the remote monitoring center, and compress the sensor and camera data and send it to the remote monitoring center server via TCP / IP or MQTT protocol; Remote management: Establish a remote monitoring center including servers, databases and monitoring software. The server receives data and stores it in the database. The monitoring software can display the operating status and monitoring screen in real time. It adopts C / S or B / S architecture. Managers log in and view the center through client software or browser. They can view parameters such as motor speed in real time and display them in charts. In case of abnormality, remote control is available, and alarm information has sound and image prompts. Car owners can receive operating status and alarm information through mobile phone APP or SMS. APP can display cleaning progress, etc., and push notifications in case of abnormality. SMS is a backup method, and the content includes car wash machine number and fault type.

[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. Accurate prediction of personnel behavior intention: Through the fusion collection of high-definition cameras, infrared cameras and sensor data, a rich data foundation is provided for accurately predicting personnel behavior intention. The use of convolutional neural networks for image feature extraction can effectively capture personnel posture, movement, expression and other features. Combined with long-term and short-term memory networks to model behavior sequences, it can accurately grasp the subtle changes and time dynamic characteristics of personnel behavior, thereby greatly improving the accuracy of behavior intention prediction.

[0016] 2. Real-time response and safety assurance: During the operation of the unmanned car wash machine, feature extraction and behavior intention prediction can be performed in real time. Once a dangerous behavior intention is detected, an audible and visual alarm will be immediately issued, the car wash machine will be stopped, and the remote monitoring center will be notified to provide strong protection for personnel safety. At the same time, there are also comprehensive monitoring measures for vehicle safety, including vehicle entry detection, door and window detection, vehicle model identification and water pressure regulation, vehicle damage detection, etc., to ensure that the vehicle is not damaged during the washing process, and to achieve dual safety protection for personnel and vehicles.

[0017] 3. Intelligent model update and optimization: Continuously collect new data to update and optimize the model, and use incremental learning methods to enable the model to adapt to changes in human behavior and environment. Use reasonable hyperparameter adjustment strategies to continuously optimize the performance of the model, improve the accuracy and stability of predictions, and ensure that the system always maintains a good operating state. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a schematic diagram of the system flow of the present invention; Figure 2 This is a flow chart of the personnel behavior intention prediction subsystem of the present invention. DETAILED DESCRIPTION

[0019] In order to make the technical means, creative features, objectives and effects achieved by the present invention easy to understand, the present invention is further explained below in conjunction with specific implementation methods.

[0020] like Figure 1-2 As shown in the figure, the unmanned car washing machine equipped with the human-vehicle safety monitoring system is implemented in actual application, and the specific implementation process is as follows: 1. Implementation of personnel safety monitoring Personnel proximity detection When a pedestrian passes by the car wash, the human infrared sensors installed at the entrance and exit of the car wash machine detect changes in infrared intensity. Assuming that the detected infrared intensity is I=150, and the threshold T1 is set to 120, it is judged that someone is approaching. The sound and light alarm device is triggered, emitting a sound with a frequency of 1000Hz and a loudness of 80 decibels and a red light with a flashing frequency of 5 times / second to remind people to stay away from the car wash machine.

[0021] A staff member enters the car washing machine for routine inspection. The proximity sensor detects the distance between the person and a specific part of the equipment in real time. Assuming that the distance to the key part of the equipment is d=0.5 meters, and the safety distance threshold T2 is set to 0.8 meters, it is judged that the person is too close to the dangerous equipment part. The alarm device is triggered, emitting a sound with a frequency of 1200Hz and a loudness of 70 decibels and a yellow light with a flashing frequency of 3 times / second. At the same time, the electromagnetic lock prevents the person from getting closer.

[0022] Personnel behavior monitoring The HD camera and infrared camera collect image data of people in the car washing machine at a fixed frame rate of 30 frames per second, and the proximity sensor detects the distance between people and equipment at a time interval of 1 second. Suppose at this time a person in the car washing machine shows abnormal behavior, such as quickly approaching the running equipment.

[0023] The system uses image recognition technology to analyze the behavioral characteristics of personnel. The image is extracted through a convolutional neural network (CNN). Using the pre-trained ResNet network architecture, the extracted features are recorded as . Combine proximity sensor data to extract additional features , where the distance function is defined as the risk level characteristics divided into different levels according to the distance.

[0024] The behavior sequence is modeled using a long short-term memory network (LSTM), and the time window size is set to 5 seconds. The continuous frame image features and the corresponding additional features are combined into a sequence, which is recorded as The output of LSTM is recorded as , the expression is =LSTM(S).

[0025] A multi-layer perceptron (MLP) is constructed as a classifier, and the output of LSTM is input into the MLP to obtain the behavior intention prediction result. Assuming that the prediction result indicates that the person has a dangerous behavior intention, an audible and visual alarm is immediately issued, the car washing machine is stopped, and the remote monitoring center is notified.

[0026] 2. Implementation of Vehicle Safety Monitoring Vehicle entry detection A car drives into the car wash and the vehicle position sensors start working. The laser range sensors on both sides of the entrance measure the width and position of the vehicle, and the ultrasonic sensors on the inner track measure the vehicle's forward distance and speed. Assume that the vehicle position has a certain deviation from the track, exceeding the set normal range threshold.

[0027] The system immediately emits red flashing lights and high-frequency alarm sounds as an audible and visual alarm, and stops the power supply of the car washer motor, water pump and other equipment to prevent further damage to the vehicle.

[0028] Vehicle door and window detection The high-definition camera is installed at the top center of the car wash machine to collect vehicle images and transmit them to the image processing system. Using the deep learning target detection algorithm, the image is first pre-processed by denoising and contrast enhancement to improve image quality.

[0029] The pre-trained object detection model is then used to detect the vehicle doors and windows in the image. If the vehicle doors and windows are detected to be not closed, the system will mark the location of the doors and windows and calculate the degree to which they are not closed.

[0030] The system prompts the owner through voice to confirm that the vehicle doors and windows are closed, and pauses the car wash program, waiting for the owner to process.

[0031] Vehicle type identification and water pressure regulation When a vehicle enters the car wash, high-definition cameras on the top and sides of the car wash collect images of the vehicle. Using a deep learning image classification algorithm, the image is input into a trained model to identify the brand and model information of the vehicle. Assuming that the vehicle is identified as an SUV of a certain brand, the system queries the pre-established vehicle parameter database to obtain the appropriate water pressure range for the model. The water pressure is adjusted by controlling the speed of the water pump motor to keep it within the appropriate range. Assuming that there is a certain gap between the current water pressure and the target water pressure, the PID control algorithm is used to calculate the control signal of the water pump motor based on the difference between the current water pressure and the target water pressure, so that the water pressure gradually approaches the target water pressure.

[0032] Vehicle damage detection During the car washing process, high-definition cameras on the top, side and bottom of the car washing machine continuously collect images of the vehicle surface and transmit them to the image processing system.

[0033] Using image recognition technology, we first use the Canny edge detection algorithm to detect scratches. If we detect continuous edge lines on the vehicle surface and their features match those of scratches, such as length, width, and direction, we determine it to be a scratch. Then we use the image segmentation algorithm combined with deep learning U-Net to detect dents. If we detect that the depth of a certain area is significantly lower than that of the surrounding area and its features match those of a dent, such as area and depth, we determine it to be a dent.

[0034] The system immediately records the location, type, severity and other information of the damage and stores it in the database, and notifies the owner via a mobile phone APP push notification.

[0035] 3. Remote Monitoring Implementation Data Transfer The unmanned car wash machine uses a 4G communication module, configures the correct network parameters such as IP address and port number, and establishes a connection with the remote monitoring center. The data collected by the sensor and camera are compressed by JPEG and packaged into a JSON format data packet. The data packet is sent to the server of the remote monitoring center through the TCP / IP protocol.

[0036] Remote Management The server of the remote monitoring center receives data from the car wash machine and stores it in the database. The monitoring software adopts B / S architecture, and managers can log in to the remote monitoring center through a browser to view the operating status and monitoring screen of the car wash machine in real time.

[0037] Managers can view parameters such as motor speed, water pressure, temperature, etc. in real time and display them in the form of graphs for easy analysis and judgment. If abnormal water pressure is found, managers can remotely send control instructions to the controller of the car washing machine to adjust the water pressure.

[0038] When an abnormal situation occurs in the car washing machine, the monitoring software will sound an alarm and display an alarm message on the screen.

[0039] Owner Notification Car owners can receive the operating status and alarm information of the car wash machine through the mobile phone APP. If a car is scratched during the car wash, the APP will push a notification to the car owner, saying "Dear car owner, we found scratches on your car during the car wash. Please check and deal with it in time." At the same time, the APP can also display information such as the vehicle's cleaning progress and estimated completion time.

[0040] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A human and vehicle safety monitoring system for an unmanned car washing machine, characterized in that: The human and vehicle safety monitoring system consists of: Sensor module: Human infrared sensors installed at the entrance and exit of the car washing machine are used to detect whether someone is close to the car washing machine; vehicle position sensors installed inside the car washing machine are used to detect the position and movement status of the vehicle; water pressure sensors installed on the water gun of the car washing machine are used to detect the water pressure of the water gun; temperature sensors installed inside the car washing machine are used to detect the temperature inside the car washing machine; proximity sensors installed in various places inside the car washing machine are used to detect the proximity between people and equipment; Camera module: HD cameras installed inside and outside the car wash machine to monitor the status of vehicles and personnel in real time; infrared cameras installed inside the car wash machine to monitor the status of vehicles and personnel in dark environments; Controller module: Microprocessor: responsible for receiving data from sensors and cameras, and processing and analyzing them; Memory: used to store monitoring data and system parameters; Communication module: used to communicate with the remote monitoring center to achieve remote monitoring and management; Alarm module: an audible and visual alarm installed inside and outside the car washer, used to sound and light alarms when abnormal situations occur; a text message alarm used to send text message alarms to car owners and managers when abnormal situations occur; The monitoring scheme of the above-mentioned human and vehicle safety monitoring system is divided into personnel safety monitoring, vehicle safety monitoring and remote monitoring, among which: Personnel safety monitoring includes: Personnel approach detection: Human infrared sensors detect whether someone is approaching at the entrance and exit of the car washing machine. Once a person is detected approaching, the system automatically issues an audible and visual alarm to remind the person to stay away from the car washing machine; when a person enters the car washing machine, the proximity sensor detects the degree of proximity between the person and a specific part of the equipment in real time. If a person approaches a part of the equipment that may cause danger, the system immediately issues an alarm and prevents the person from getting closer; Personnel behavior monitoring: High-definition cameras and infrared cameras monitor the behavior of people entering the car wash machine in real time. The system uses image recognition technology to analyze the behavioral characteristics of people. If abnormal behavior is detected, the system will immediately issue an audible and visual alarm and stop the operation of the car wash machine. At the same time, the alarm information is sent to the remote monitoring center through the communication module to notify the management personnel to handle it in time. Vehicle safety monitoring includes: Vehicle entry detection: The vehicle position sensor detects the position and movement status of the vehicle when it enters the car wash machine; if the vehicle position is incorrect or the movement status is abnormal, the system immediately issues an audible and visual alarm and stops the operation of the car wash machine; Vehicle door and window detection: Use high-definition cameras to detect the vehicle's doors and windows; if the doors and windows are found not closed tightly, the system will prompt the owner to confirm that the vehicle's doors and windows are closed tightly through voice, and at the same time suspend the car washing process and wait for the owner to handle it; Vehicle type recognition and water pressure adjustment: The high-definition camera uses image recognition technology to identify vehicles of different brands and models; the system automatically adjusts the water pressure of the water gun based on the pre-stored vehicle parameters of different brands and models; Vehicle damage detection: During the car wash process, the high-definition camera continuously scans the vehicle surface; once scratches, dents and other damages are found on the vehicle, the system immediately records them and notifies the owner; Remote monitoring includes: Data transmission: The data collected by sensors and cameras are transmitted to the remote monitoring center in real time through the communication module; Remote management: Managers can view the operating status and monitoring screen of the car wash machine in real time through the remote monitoring center; car owners can receive the operating status and alarm information of the car wash machine through mobile phone APP or SMS.

2. The vehicle and person safety monitoring system of an unmanned car washing machine according to claim 1, characterized in that: The person is approaching the detection: Assume that the infrared intensity detected by the human infrared sensor is I , set a threshold T 1 ,when I>T 1 When someone approaches, it determines that someone is approaching; when it detects that someone is approaching, the sound and light alarm device is triggered, emitting a sound of a specific frequency and loudness and a light of a specific color and flashing frequency; Assume the distance detected by the proximity sensor is d , set a safe distance threshold T 2 ,when d<T 2 When it is detected that a person is too close to a dangerous device, the alarm device is triggered, emitting a sound of a specific frequency and loudness and a light of a specific color and flashing frequency, while an electromagnetic lock is used to prevent the person from getting closer.

3. The vehicle and person safety monitoring system of an unmanned car washing machine according to claim 1 is characterized in that: The personnel behavior monitoring includes a subsystem that uses image recognition technology to analyze personnel behavior characteristics and predict personnel behavior intentions. The system includes a data acquisition module, a data annotation module, a feature extraction module, a behavior sequence modeling module, a behavior intention prediction model training module, a real-time behavior intention prediction and response module, and a model update and optimization module.

4. The vehicle and person safety monitoring system of an unmanned car washing machine according to claim 3 is characterized in that: The specific implementation methods of each component module of the personnel behavior intention prediction subsystem are as follows: The data acquisition module is used to continuously collect image data of people in the car washing machine using the high-definition camera and infrared camera installed in the unmanned car washing machine, as well as to collect sensor data; wherein, the image data is collected using the high-definition camera and infrared camera at a fixed frame rate, denoted as ,in Indicates time; proximity sensor uses time intervals The distance between the detection personnel and the equipment is recorded as ; The data annotation module manually annotates the collected image data, classifies human behaviors into different categories, and annotates the corresponding behavioral intentions; The feature extraction module uses a convolutional neural network (CNN) to extract features from images and extracts additional features in combination with sensor data. Perform feature extraction and record the extracted features as , using the pre-trained network architecture ResNet or VGG, the expression is ; Combined with proximity sensor data , extract additional features , the expression is ,in is a function that defines features based on distance; The behavior sequence modeling module uses a recurrent neural network (RNN), namely a long short-term memory network (LSTM) or a gated recurrent unit (GRU) to model the behavior sequence; the time window size is set to , the continuous The frame image features and the corresponding additional features form a sequence, recorded as ; Use LSTM or GRU to sequence For modeling, the output of LSTM is , the expression is ; The behavior intention prediction model training module builds a behavior intention prediction model and inputs the extracted features into the model for training; a multi-layer perceptron (MLP) is built as a classifier and the output of the LSTM is input into the model for training; Input into MLP to get the behavior intention prediction result , the expression is ; Define the cross entropy loss function ,in is the real behavioral intention label, is the predicted behavioral intention; Adam optimizer is selected to train the model. During the training process, the model parameters are continuously adjusted to minimize the loss function; specifically, for the model parameters , in each iteration, according to the current loss function gradient and learning rate Update the parameters, i.e. ; The real-time behavior intention prediction and response module performs real-time feature extraction and behavior intention prediction during the operation of the unmanned car wash machine, and takes corresponding measures based on the prediction results; during the operation of the unmanned car wash machine, it collects image data and sensor data in real time, performs feature extraction and behavior sequence modeling, and obtains the current behavior intention prediction results If the prediction result indicates that the person has dangerous behavior intentions, an audible and visual alarm will be issued immediately, the car washing machine will be stopped, and the remote monitoring center will be notified; The model update and optimization module continuously collects new data to update and optimize the model; continuously collects new image data and sensor data and annotates them; regularly uses new data to update and optimize the model, using the incremental learning method. Specifically, the real label and prediction result of the new data are and , the true labels and prediction results of the old data are and , the learning rate parameter is and ; During the update process, for the model parameters , first according to the loss function gradient and learning rate of the new data To update, , and then combine the loss function gradient and learning rate of the old data Further updates, namely ; At the same time, the hyperparameters of the model can be adjusted to improve the performance of the model.

5. The vehicle and person safety monitoring system of an unmanned car washing machine according to claim 4 is characterized in that: In the behavior intention prediction model training module, a multi-layer perceptron (MLP) is composed of an input layer, a hidden layer, and an output layer; the input layer receives output features from LSTM or GRU, the hidden layer further processes the features through a combination of multiple neurons, and the output layer outputs the prediction results of the behavior intention; the number of nodes in the input layer is equal to the output feature dimension of LSTM or GRU; The activation function selects the ReLU function as the activation function of the hidden layer, and the output layer uses the softmax function for multi-classification. The calculation method of the cross entropy loss function is to compare the probability of each behavior intention category predicted by the model with the true label and calculate the cross entropy between them. Assume that the probability distribution of the behavior intention predicted by the model is , the true label is ,in The number of categories representing behavioral intentions, the calculation formula of the cross entropy loss function is 。 6. The vehicle and person safety monitoring system of an unmanned car washing machine according to claim 4, characterized in that: The model update and optimization module adopts an incremental learning method. Specifically, when using mini-batch gradient descent for incremental learning, a fixed mini-batch size is set. For newly collected data, it is randomly shuffled and divided into several mini-batches. For each mini-batch of data, it is first input into the current model to obtain the model's prediction result. Then, the loss function between the prediction result and the true label is calculated. The cross entropy loss function can continue to be used here. Next, the gradient of the loss function to the model parameters is calculated through the back propagation algorithm. According to the gradient and the learning rate, the model parameters are updated. For the real label and prediction result of the new data, and , the cross entropy loss function of the new data is recorded as ,in is the predicted probability distribution of new data, is the true label of the new data; when considering the influence of old data, the forgetting factor is used to weight the loss function of the old data. Assume that the loss function of the old data is , the loss function for the new data is , forgetting factor is (0 < < 1), the total loss function is ; Then, the gradient is calculated based on this total loss function and the model parameters are updated.

7. The vehicle and person safety monitoring system of an unmanned car washing machine according to claim 1, characterized in that: The specific implementation method of the vehicle safety monitoring is: Vehicle entry detection: The vehicle position sensor uses a laser rangefinder sensor or an ultrasonic sensor to determine the distance between the vehicle and the sensor by emitting a specific signal and measuring the reflection time, and determines the vehicle position and motion state based on the measurement results of multiple sensors; laser rangefinder sensors are installed on both sides of the car wash entrance to measure the width and position of the vehicle, and ultrasonic sensors are installed on the internal track to measure the forward distance and speed; when it is detected that the vehicle position is incorrect or the motion state is abnormal, the system emits a red flashing light and a high-frequency alarm to sound and light alarms, and stops the operation of the car wash machine motor, water pump and other equipment by controlling the power supply; Vehicle door and window detection: Install a high-definition camera at the top center or on both sides of the car wash machine to collect vehicle images and transmit them to the image processing system; use a deep learning target detection algorithm to process images, pre-process to improve image quality, and then use a pre-trained model to detect the state of doors and windows. Set a threshold based on the angle between the doors and windows and the vehicle body to determine whether they are closed tightly. If they are not closed tightly, mark the position and calculate the degree of looseness. If it is detected that they are not closed tightly, the owner will be prompted to confirm through voice, and the car wash program will be suspended at the same time. The program will be restarted after the owner confirms that they are closed tightly. Vehicle model recognition and water pressure adjustment: Use high-definition cameras on the top and side of the car wash machine to collect vehicle images, and use deep learning image classification algorithms to identify vehicle models. When a vehicle enters the car wash machine, the camera collects images and inputs the trained model to output brand and model information; establish a vehicle parameter database to store the shell material parameters and suitable water pressure ranges of different brands and models. After the system recognizes the model, it queries the database to obtain the suitable water pressure range, adjusts the water pressure by controlling the speed of the water pump motor or the valve opening, and uses the PID control algorithm to calculate the control signal based on the difference between the current and target water pressures to adjust the water pressure to make it close to the target water pressure; Vehicle damage detection: During the car wash process, high-definition cameras are installed on the top, side, and bottom of the car wash machine to collect images of the vehicle surface and transmit them to the image processing system; The image is processed using image recognition technology, scratches are detected using the Canny edge detection algorithm, and dents are detected using an image segmentation algorithm combined with deep learning U-Net. When damage is detected, the damage location, type, and severity information are recorded and stored in a database, and the owner is notified via voice prompts, SMS notifications, or mobile phone APP push notifications.

8. The vehicle and person safety monitoring system of an unmanned car washing machine according to claim 1, characterized in that: The specific implementation method of the remote monitoring is: Data transmission: Use 4G / 5G, Wi-Fi or wired network modules, configure the correct network parameters to establish a connection with the remote monitoring center, and compress the sensor and camera data and send it to the remote monitoring center server via TCP / IP or MQTT protocol; Remote management: Establish a remote monitoring center including servers, databases and monitoring software. The server receives data and stores it in the database. The monitoring software can display the operating status and monitoring screen in real time. It adopts C / S or B / S architecture. Managers log in and view the center through client software or browser. They can view parameters such as motor speed in real time and display them in charts. In case of abnormality, remote control is available, and alarm information has sound and image prompts. Car owners can receive operating status and alarm information through mobile phone APP or SMS. APP can display cleaning progress, etc., and push notifications in case of abnormality. SMS is a backup method, and the content includes car wash machine number and fault type.

Images