There are usually hundreds of sensors in the multi-sensor system. Each sensor will
produce a large amount of parameter data over time. Once the state distribution of
some parameters changes, it is likely to lead to abnormal conditions of other
parameters in the system, and even the whole system can not operate normally. In
this project, the parameters of all sensors in the Autoencoder depth structure
real-time monitoring system based on Bi-LSTM will give an alarm in time and notify
relevant personnel to repair in time in case of abnormal parameters. This method can
not only effectively maintain the efficient operation state of the system, but also
avoid the loss caused by system abnormalities.
The main contents of this project include the vertical acceleration VRTG ≥ 2g during
landing, great safety risk, possible casualties, real-time analysis of sensor
parameters and prediction of future VRTG. Time series prediction is realized based
on time-delay neural network.
The main contents of the project include real-time statistics of passenger flow at
entrances and exits, warning of over threshold passenger flow and prediction of
passenger flow in peak hours. Key technologies such as target object segmentation
technology in image / video, moving target tracking technology in video and
regression analysis based on statistical passenger flow are adopted, which can be
used to monitor passenger flow in special scenes such as subway station doorway and
escalator entrance in real time, The situation exceeding the passenger flow
threshold shall be reported to the customer service management platform for timely
personnel counseling to prevent personnel stampede and other scenes.
