Telemetering spectrum noise suppression method suitable for infrared small target recognition

Abstract

The invention discloses a telemetering spectrum noise suppression method suitable for infrared small target recognition. The method includes the steps that normalization processing is carried out on infrared spectrum data; the first-order derivative of spectral intensity obtained after normalization processing is calculated; the noise level of a spectrum is estimated according to the first-order derivative of the spectrum, and thus two important parameter values of a bilateral filter are determined; a spectrum database is trained, and an optimal parameter combination is solved and used for guiding parameter setting of the bilateral filter; the bilateral filter with the optimal parameters is used for processing the noise-containing spectrum to obtain a high-resolution infrared spectrum. According to the method, the noise suppression advantage of the bilateral filter is effectively utilized, the spectral resolution is increased by about 35% when the signal-to-noise ratio is higher than 30 dB, application of the infrared spectrum in the fields of target detection and recognition, space technology, meteorology, petrochemical engineering and the like is promoted, and a corresponding technical support is provided.

Description

technical field [0001] The invention relates to infrared spectrum signal preprocessing technology, in particular to a remote measurement spectrum noise suppression method suitable for infrared small target recognition. Background technique [0002] Infrared telemetry spectrum is often used in long-distance target detection, air pollution gas detection, etc., and has outstanding advantages such as long distance, large range, and simple operation. However, the measured spectral signal often contains a lot of high-frequency noise and is interfered by the low-frequency baseline. Random noise is affected by factors such as experimental conditions and the error of the spectrometer itself. It affects the subsequent processing of the spectrum and hinders the effective use of data. To solve this problem, many noise suppression methods have been proposed to improve spectral resolution. [0003] The traditional noise suppression method generally designs the filter template and weigh...

AI Technical Summary

Problems solved by technology

Among them, the differential method has a good effect on eliminating baseline drift, strengthening band characteristics, and overcoming band overlap. However, the spectral noise of the original spectrum will increase after differentiation, and further spectral denoising is required.

The smoothing method can effectively smooth the high-frequency noise in the signal, but it will also smooth the target spectrum at the same time, resulting in signal distortion

The Fourier transform method separates the time domain signal from the frequency domain signal, and uses the frequency characteristics of the signal for denoising, but it only has a good denoising effect on stationary signals

Images