Abstract
Resolving adjacent submeter-scale hot spots in Raman distributed temperature sensing (DTS) is critically limited by the finite pump-pulse duration. This pulse-width-limited spatial resolution inherently merges thermal responses from nearby events into an overlapped profile. Chaotic Raman DTS can overcome the pulse-width limitation in principle. However, the correlation peaks still broaden and overlap when multiple hot spots are adjacent, preventing reliable identification of individual events. Here, we propose and demonstrate an optimal chaotic segment correlation (OCSC) method that exploits the nonuniform local correlation property of a broadband chaotic pulse. A 1-ns sliding window is applied to a 50-ns chaotic reference pulse to select the segment with the minimum autocorrelation full width at half maximum (FWHM), and this optimal segment is then used as the correlation kernel against a self-differentially reconstructed Raman anti-Stokes signal. This local-reference strategy effectively suppresses redundant correlation components arising from the full pulse and enhances the separability of adjacent temperature-induced responses. In the experiment, two adjacent 20-cm hot spots with 60-cm separation are clearly resolved at a distance of 2.52 km, with consistent demodulation performance obtained at 60, 70, and 80°C. The proposed OCSC method provides a general signal-processing route for multi-event discrimination in correlation-based distributed fiber sensors that use chaotic laser or random light sources.