Denoising Atmospheric Temperature Measurements Taken by the Mars Science Laboratory on the Martian Surface

TL;DR

This paper compares denoising methods for Mars temperature data, showing that Discrete Wavelet Transform offers better performance and feature detection than the traditional Moving Average filter, especially considering computational efficiency.

Contribution

It introduces and evaluates DWT and HHT methods for denoising Martian temperature measurements, demonstrating DWT's suitability for large-scale data processing.

Findings

DWT outperforms MA in filtering accuracy.

DWT enables better feature and artifact detection.

DWT is more computationally efficient for large datasets.

Abstract

In the present article we analyze data from two temperature sensors of the Mars Science Laboratory, which has been active in Mars since August 2012. Temperature measurements received from the rover are noisy and must be processed and validated before being delivered to the scientific community. Currently, a simple Moving Average (MA) filter is used to perform signal denoising. The application of this basic method relies on the assumption that the noise is stationary and statistically independent from the underlying structure of the signal, an arguable assumption in this kind of harsh environment. In this paper, we analyze the application of two alternative methods to process the temperature sensor measurements: the Discrete Wavelet Transform (DWT) and the Hilbert-Huang Transform (HHT). We consider two different datasets, one belonging to the current Martian measurement campaigns, and the other to the Thermal Vacuum Tests. The processing of these datasets allows to separate the random noise from the interference created by other systems. The experiments show that the MA filter may provide useful results under given circumstances. However, the proposed methods allow a better fitting for all the realistic scenarios, while providing the possibility to identify and analyze other interesting signal features and artifacts that could be later studied and classified. The large amount of data to be processed makes computational efficiency an important requirement in this mission. Considering the computational cost and the filtering performance, we propose the method based on DWT as more suitable for this application.