**5. Conclusions**

We built a polarized multispectral for a low-illumination-level imaging system and a polarized multispectral low-light imaging system (PMSIS), incorporating a SCOMS camera, rotating filter wheel and linear polarizer, for monitoring the health status of plants outdoors at night. With a compact size, the developed system is portable, consisting of a minimum number of moving parts, which makes it ideal for field measurements. We have done some research on the potential of polarization technology to provide contrast enhancement information for vegetation assessments. Polarization imaging and multispectral imaging can provide complementary discrimination information for target monitoring and other applications. However, few people have proposed to fuse the information of polarized images and multispectral images to obtain better target monitoring results. Therefore, a fusion algorithm based on the spectral and polarization characteristics of the diffuse and specular reflection of vegetation is proposed to detect vegetation at night. The NDVI, DoLP

and AOP were all computed within a fusion framework to better monitor the health status of plants in a nighttime environment.

In order to evaluate the remote sensing power of the fusion algorithm in monitoring the plant health status at night, the leaf regions of spotted laurel in different health states were measured. The changes of the SPAD, NC, NDVI and NPSDI in different health levels of plants were determined, and the health degrees of spotted laurel were classified. The value of the NPSDI decreased with the decrease of the plant health status, and it was in a positive linear correlation with the physiological parameters such as the SPAD and NC. The correlation of the NPSDI with NC was 0.916, while, with the SPAD, it was 0.882. In applying the fusion algorithm in classifying healthy leaves from level-1 stress leaves, the sensitivity obtained was 89% and the specificity 92%, which suggests the effectiveness of the method for the early detection of the health status in field-grown plants.

In this paper, the fusion of the NDVI, DoLP and AOP was proposed for the first time. The proposed fusion algorithm enhanced the contrast effect, and the resultant image carried more abundant information of the object; thus, monitoring the health status of plants at night becomes more effective. The fusion algorithm makes up for the unreliability of the NDVI in a low-illumination environment. Based on the polarimetric imaging system that we developed, polarization information improved the monitoring accuracy at night with traditional NDVI information. In the study of monitoring the plant health status at night, we found that the NPSDI was excellent in tracking stress-induced plant changes. The NPSDI scatter plots were used to study for classifying different grades of the health state in spotted laurel, and the sensitivity and specificity were good, which indicated that this technique was suitable for the classification of the outdoor plant health status and various stresses. This study provided the basis for the promotion and use of PMSIS from various remote sensing platforms (such as cars and drones) to assess the crop health status by polarization spectral imaging at night. In the future, the system can be improved (a three-camera real-time imaging system), so that it can be installed on vehicle and airborne devices, enabling its application in a large area for monitoring the health of vegetation at night. The system has the following applications: (1) the detection of the vegetation status in low-illumination environments, (2) the detection of the vegetation status during the day, (3) the detection of the polarization of vegetation leaves and (4) the detection of artificial targets in vegetation environments.

We supplement a table at the end of paper. Acronyms are shown in Table 2.


**Table 2.** List of acronyms with their meanings.

**Author Contributions:** J.J. and C.W. conceived and designed the experiments; S.L. performed the experiments and analyzed and interpreted the results. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the National Natural Science Foundation of China (No. 61773249) and Shanghai Science and Technology Innovation Action Plan (No. 20142200100) for project funding.

**Acknowledgments:** Thanks to the colleagues and students who participated in the outdoor data collection.

**Conflicts of Interest:** The authors declare no conflict of interest.
