**1. Introduction**

Vegetation plays an important role in the global ecosystem [1], but natural disasters such as drought, pests and diseases affect their growth of greatly, even leading to their death. In the past 30 years, monitoring the health status of vegetation via remote sensing has been applied in many areas successfully, such as the analysis of crop diseases and pests, monitoring forest coverage and vegetation growth status, etc. However, such remote sensing-based activities are mainly carried out during the daytime, with limitations in sensing the vegetation status at night. Crop diseases and pests are important factors for analyzing crop yields. For example, *Spodoptera frugiperda*, as one kind of nocturnal animal, spreads extremely fast and seriously threatens human food security [2]. Thus, it is necessary to monitor them efficiently with state-of-the-art techniques. Therefore, it is essential and important to develop an effective and accurate method to monitor the health status of vegetation in the night environment.

With the development of remote sensing technologies based on multiple platforms (space, air and ground), various remote sensing-based methods are used to monitor plant health [3–11]. For the optical remote sensing methods, the visible and near-infrared wave bands are mainly used for analysis of plant diseases and insect pests [12–21]. As the

**Citation:** Li, S.; Jiao, J.; Wang, C. Research on Polarized Multi-Spectral System and Fusion Algorithm for Remote Sensing of Vegetation Status at Night. *Remote Sens.* **2021**, *13*, 3510. https://doi.org/10.3390/rs13173510

Academic Editor: Xanthoula Eirini Pantazi

Received: 2 August 2021 Accepted: 1 September 2021 Published: 4 September 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

vegetation of different healthy status have different absorption and reflection characteristics at different wavelengths, various vegetation indices (VIS) with visible and near-infrared wave bands have been developed to monitor vegetation growth and health [22], such as the normalized vegetation index (NDVI, Rouse et al. 1973) [23], soil-adjusted vegetation index (SAVI, Huete 1988) [24], modified soil-adjusted vegetation index (MSAVI, Qi et al. 1994) [25] and normalized difference water index (NDWI, Gao 1996) [26]. More and more indices are being developed to study and analyze the crop damage caused by pests with remote sensing.

For the ground-based remote sensing technologies, typically, handheld instruments are widely used monitoring crop diseases [27], and tower-based platforms [28–30] and other huge near-ground platforms are used to obtain crop spectral information under different disease states at the leaf and canopy scales. For example, Graeff et al. (2006) [31] used a digital imager (Leica S1 Pro, Leica, Wetzlar, Germany) to analyze the spectrum of wheat leaves infected with powdery mildew and take-all disease, finding this disease resulted in strong spectral response at 490 nm, 510 nm, 516 nm, 540 nm, 780 nm and 1300 nm. Yang et al. (2009) [32] used a hand-held Cropscan radiometer and found that it was feasible to use the ratio vegetation index (400/450 nm and 950 nm/450 nm) to identify and distinguish plants infested by green aphids and Russian wheat aphids. Liu et al. (2010) [33] used a portable spectrophotometer (Field Spec Full Range, ASD Inc., Boulder, CO, USA) to analyze the spectral characteristics of Rice Panicles and found that there was a correlation between the reflectance change in the 450–850-nm band and rice glume blight disease. In addition, Prabhakar et al. (2011) [34] found that the spectral reflectance of healthy plants was significantly different from the plants infested by leafhoppers in the visible and near-infrared bands with a Hi-Res spectroradiometer (ASD Inc., Boulder, CO, USA; spectral range: 350–2500 nm), and the new leaf hopper index (LHI) was applied to monitor the severity of leafhopper. Prabhakar et al. (2013) [35] used a Hi-Res spectroradiometer (ASD Inc., Boulder, CO, USA; spectral range: 350–2500 nm) for the experiment and found that the most sensitive bands to cotton mealybug infestation were concentrated at 492 nm, 550 nm, 674 nm, 768 nm and 1454 nm, and a new cotton mealybug Stress Index (MSI) was developed.

Besides ground-based remote sensing technologies, various air and space-based remote sensing technologies have been developed and widely applied to the monitoring of plant diseases and insect pests, with the advantages of good temporal, spatial and spectral resolutions. In general, the airborne platform used for monitoring crop diseases can be integrated with different sensors, such as an imaging camera, multispectral/hyperspectral spectrometer, infrared camera, lidar and other detection systems [36,37]. For example, Yang et al. (2010) [38] used high-resolution multispectral and hyperspectral aerial image data to extract the occurrence range of cotton root rot and indicated that multispectral data was promising for large-scale disease monitoring. Calder ó n et al. (2013) [39] used UAV with the multispectral camera and thermal infrared camera to diagnose the verticillium wilt of olive trees. Sanches et al. (2014) [40] calculated a new Plant Stress Detection Index (PSDI) from the chlorophyll characteristic center (680 nm) and green edge (560 nm and 575 nm) with an airborne imaging spectrometer (ProSpecTIR-VS) and the continuum division method, showing this index could be used for an analysis of the canopy status successfully. Lehmann et al. (2015) [41] analyzed the UAV multi-spectral images with object-based image processing methods to monitor the pests of oak trees. For the cases of space-based remote sensing applications, Yuan et al. (2016) [42] proposed a monitoring method with the help of SPOT-6 images to analyze the occurrence of Wheat Powdery Mildew in the Guanzhong area of Shaanxi Province, with an accuracy of 78%. Chen et al. (2018) [43] applied high-resolution multi-spectral satellite images to monitor wheat rust, with a classification accuracy of 90%. Zheng et al. (2018) [44] used the Sentinel-2 Multispectral Instrument (MSI) to distinguish the severity of yellow rust infection in the winter and proposed the Red Edge Disease Stress Index (REDSI) to detect yellow rust infections of different severities. Meiforth et al. (2020) [45] used the WorldView-2 (WV2) satellite and

LiDAR data to detect the stress of kauri canopies in New Zealand. The results showed that this method can be used to monitor kauri canopies economically and efficiently in a large area. Li et al. (2021) [46] developed a machine learning model to analyze the vegetation growth by retrieving NDVI from the satellite sensor.

It is clear that vegetation indices based on different remote sensing platforms play an important role in monitoring the status of crops. When the polarization information is supplemented, more complex and accurate indices and models can be developed, and the polarization shows directional characteristics when light interacts with the vegetation [47,48]. Polarization has attracted many researchers to study remote sensing together with polarization information. For example, Vanderbilt et al. pointed out that polarization was affected by vegetation canopy morphology and leaf surface characteristics [49,50]. Since vegetation canopy morphology and leaf surface characteristics are affected by stress, activity and the growth stage, polarization-based parameters, such as the degree of linear polarization (DoLP) of vegetation, can reflect such information well [51]. Compared to traditional remote sensing methods without polarization information, polarization-based remote sensing can provide vegetation canopy information, such as the structure of leaf layers [52] and the emergence of panicles above canopies [53]. Theoretical and practical research indicates the possibility of detecting the geometry of vegetation canopies with polarization [54,55]. Besides, many researchers have measured various leaves in an extensive way [56–60]. Hu et al. studied polarization imaging in low-light environments [61]. For the above-mentioned research, it is worth noting that the researchers are still in the phase of ground experiments in studying remote sensing technologies.

It is shown that the health of vegetation can be monitored successfully with different remote sensing technologies, and various methods have been carried out during the daytime. However, limited efforts have been done in monitoring the health of vegetation at night, which requires high-sensitivity imaging equipment but is quite important and necessary. In this paper, a ground-based remote sensing system is developed and used for monitoring the health of vegetation at night. Since polarization information can supplement traditional spectral imaging to develop more complex and accurate models for vegetation monitoring, in this paper, we developed a polarized multispectral low-illumination-level imaging system, which is used to record the spectral images of vegetation at 680 nm and 760 nm during the night, as well as the polarization images at 0◦, 60◦ and 120◦. Moreover, a fusion algorithm is proposed to improve the contrast of vegetation detection by fusing polarization images and spectral images of vegetation. The vegetation index (NDVI), degree of linear polarization (DoLP) and angle of polarization (AOP) are all calculated in the fusion algorithm for better detection of the health status of plants at night. In addition, a new index, the night plant status detection index (NPSDI), was calculated based on the fusion images of NDVI, DoLP and AOP. The index was compared with physiological parameters of plants, such as nitrogen content (NC) and SPAD, to assess the applicability of the fusion algorithm for the remote sensing of vegetation. Moreover, based on our novel ground-based remote sensing research, it is promising to transform the platform into air and space-based ones for large-scale remote sensing applications in the future.
