Data Processing for Laser Screen Velocity Measurement System Based on LSTM and Spatio-Temporal Constraints

The Laser Screen Velocity Measurement System is a typical velocity measurement device for high-speed and small-size moving objects. However, the system faces serious challenges in real-time signal processing when measuring dense supersonic spherical fragment clusters. Traditional methods cannot solve the problems of accurate labeling of fragment signals and signal matching between the two channels. To address these challenges, a hybrid intelligent data processing framework is proposed in the paper. This framework is based on a Long Short-Term Memory (LSTM) network and Spatio-temporal constraints matching. The proposed framework integrates the Fourier Synchrosqueezed Transform (FSST) to enhance the time-frequency characteristics, and the Particle Swarm Optimization (PSO) algorithm to optimize the parameters of the LSTM network, thereby facilitating the accurate annotation of fragment signals. Furthermore, the spatio-temporal constraints are constructed through the fragment kinematic model, and the dynamic planning matching mechanism is established to solve the cross-screen matching problem of high-density fragment clusters. Data acquisition and method validation were executed by means of three models of fragment generators. Experimental results demonstrate: the signal annotation module achieves 94.1% recall, while the cross-screen matching module attains 95.8% accuracy. With their product yielding a 90.1% overall detection rate, this method establishes an effective framework for adaptive processing of parallel laser screen data.