AC-LSTM: Anomaly State Perception of Infrared Point Targets Based on CNN+LSTM

Anomaly perception of infrared point targets has high application value in many fields, such as maritime surveillance, airspace and early warning systems. This kind abnormality includes the explosion target, separation between stages, disintegration caused by abnormal strike, etc. By extracting radiation characteristics continuous frame targets, it is possible to analyze warn target state time. Most anomaly detection methods adopt traditional outlier detection, which problems poor accuracy a false alarm rate. Driven data, this paper proposes new network structure, called AC-LSTM, combines Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM), embeds Periodic Time Series Data Attention module (PTSA). The can better extract spatial temporal one-dimensional time series PTSA consider periodic process movement, focus on data. In addition, also data enhancement method, slices re-amplifies long method significantly improves detection. Through large number experiments, AC-LSTM achieved higher scores our collected datasets than other methods.