Research On The Technology Of Environment Monitoring And Early Warning Based On Multi-sensor

With the rapid development of the transportation industry today,the problems of safety and pollution have become increasingly prominent.The monitoring of the traffic environment is of great significance to traffic safety,traffic efficiency and environmental pollution control.This subject has designed a multi-sensor-based traffic environment monitoring and early warning system to collect real-time environmental data and make prediction and early warning of visibility-related factors.This subject has designed a 485 gateway with 485 bus communication function,network port communication function,and 4G communication function,then,completed the design of its hardware,software and the corresponding cloud preparation.Its processor is the STM32F103RET6 chip.Based on it,this subject completed the transplantation of the u COSII operating system and transplantation of the Lw IP protocol stack,then,modificated it to supports dual network card communication.According to the functional division of the hardware,this subject designed many tasks including 485 bus task,network inspection task,data transmission task and cloud interactive task.It can connect the 485 bus transceiver through the serial port to complete the collection of485 data and the control of the 485 circuit breaker.The 485 data is provided by multiple sensors including temperature,humidity,atmospheric pressure,wind speed,wind direction sensors,four gas and two dust sensors which is collected once in about 3seconds.It can connect the Ethernet module through the SPI bus to complete the data communication with the load camera equipment and the data processing platform,and complete the network status monitoring and sensor data transmission every 3 seconds.It can communicate with the cloud through the 4G module and receive cloud power off or start commands,IP configuration commands,restart commands,and inspection requests,all these responses can be completed within 1 second,and it can upload local data and device status once in 3 seconds.When a power failure occurs,the alarm information can be sent to the cloud within 1 second,and the current data and historical data at any time can be observed in the cloud.In addition,based on the data processing platform,this subject finished the environment construction and data transmission program design,and then selected the items to be used to predict and preprocess them,after that,relying on the LSTM neural network to build a model based on the codec and attention mechanism,and draw the loss function of training and verifying this model,the model with better performance is selected to input the test set data,after drawing the prediction curve,performed error analysis and evaluation.According to the input of 60 hours,the pollutant concentration of the next 10 hours can be predicted,and the RMSE of the 250th iteration is 0.493.