Large swivel bridge construction equipment real-time action detection method based on 3D CNN

The invention provides a large swivel bridge construction equipment real-time motion detection method based on 3DCNN. The motion detection method comprises the following steps: real-time motion detection of building equipment in a video is carried out; based on the single-flow features of the 3DCNN, a three-dimensional attention module and a feature pyramid network developed in the invention are utilized, and network extraction and network detection learning are designed to improve the performance; according to the method, a single-stage model based on 3DCNN can be realized, and the method is used for detecting the action of construction equipment in real time in videos of different, variable and large swivel bridge complex construction sites. The provided method paves a way for improving the safety, productivity and environment management of a construction project; according to the method, the security of the nodes is verified in a grouping manner, and inert nodes and illegal nodes can be quickly detected. 本