Independent Component Analysis-Based Infrared Dim and Small Target Detection

This is the authors' abstract. We don't add key points for this paper.

Infrared (IR) dim and small target detection is crucial in various military and civilian fields but faces challenges in complex environments and low signal-to-noise ratio. Considering the limitations of the existing methods in capturing the characteristics of different components, this paper proposes an IR small target detection method based on frequency data sphering and independent component analysis (FICA-TD). Specifically, for the yet-to-be-detected IR frame, we construct a corresponding spatial-temporal tensor with its adjacent frames. Then, we suppress the low-frequency information with a designed frequency high-pass filtering operator and a frequency data sphering operator. Subsequently, we develop independent component analysis theory for determining the number of signal sources and the basis of independent component space ΩIC, followed by projecting the frequency data sphered tensorDspinto the constructed independent component space. Furthermore, as for the achieved independent component tensorI, a sparsity information filter is developed to filter out the residual low-response information of the middle frontal slice. Finally, the target detection result sequence can be obtained. Comprehensive experiments and analysis on five real IR sequences demonstrate that FICA-TD is superior to seven state-of-the-art competitive methods in terms of target detectability, background suppressibility, and overall performance.