Assessment for crop water stress with infrared thermal imagery in precision agriculture: A review and future prospects for deep learning applications

Highlights

• Thermal imaging cameras are widely used in crop water stress assessment.

• Thermal imaging cameras with ground-based and unmanned aerial vehicles are reviewed.

• Image processing strategies for thermal images has been discussed.

• Presents critical review of crop water stress assessment through thermal images.

• Discussing prospects and challenges of deep learning applications of thermal images.

Abstract

With the increasing global water scarcity, efficient assessment methods for crop water stress have become a prerequisite to perform precision irrigation scheduling. The 1accessibility of infrared thermal sensor provides a powerful tool to detect and quantify crop water stress. This paper reviews the current practices of infrared thermal imagery utilized to assess crop water stress. Overall, three technological aspects of infrared thermal sensing applications for crop water stress assessment are reviewed along with the challenges and recommendations: (i) introduction of uncooled thermal camera and platforms, including ground-based platform and unmanned aerial vehicles (UAVs) platforms, for thermal imaging acquisition, (ii) strategies of canopy segmentation in thermal imaging used to obtain average canopy temperature for CWSI calculation, (iii) correlation between three forms of crop water stress index (CWSI) i.e. theoretical CWSI (CWSIt), empirical CWSI (CWSIe), and statistic CWSI (CWSIs) and physiological indicators. The emphasis is on imaging process techniques for canopy segmentation in thermal imaging. As a future perspective, the potential use of deep learning approaches to assess crop water stress has been elaborated highlighting the future trends.

Introduction

Nearly two-thirds of the freshwater consumption is accounted for crop irrigation purpose (Navarro-Hellín et al., 2015). With the increasing world population, it is expected that food production and agricultural water consumption will be significantly increased in near future which will cause water scarcity. Irrigated area provides 45% of the global food supply covering only 18% of the cultivated land (Jorge Gago et al., 2014). It means that the main increase in food production is more likely coming from irrigated area. Considering the limited availability of water resources and the increased demand of water by other sectors and climate change, it is becoming of utmost importance to efficiently utilize the available water resources for crop productivity in irrigated area. Under this scenario, sophisticated irrigation scheduling strategies are especially required to enhance efficient water utilization and maintain crop yield and quality. Irrigation management is crucial to optimize crop yield and quality while conserving limited water resources and protecting environment sustainability. Plant physiological parameters including leaf water potential, stem water potential, and stomatal conductance have been used as indicators to evaluate crop water stress. Traditionally, irrigation scheduling mainly relies on the direct measurement of these physiological indicators, such as evapotranspiration-based estimation, plant water potential measurement, and soil moisture measurement (Acevedo-Opazo et al., 2008, Acevedo-Opazo et al., 2010, Fernández-Novales et al., 2018). However, traditional measures are highly labor intensive, also require expert and trained personnel to operate the instruments (Salvador Gutiérrez et al., 2018). Non-destructive and rapid crop water stress monitoring is enormously needed to support precision irrigation management. The accessibility to infrared thermal imagery offers a fast and reliable alternative to detect and quantify crop water stress.

Stomatal closure induced by water deficits (even for very short time) reduces the transpiration rate, resulting in reduction of evaporative cooling and increase of leaf temperature (Buckley, 2019, Kögler and Söffker, 2019). Crop water stress index (CWSI), a thermal-derived indices based on canopy temperature measurement, has been performed to assess water deficit for diverse field crops, such as grapevine (Gonzalez-Dugo et al., 2015, Pôc et al., 2017, Gutiérrez et al., 2018), olive (Ortega-Farías et al., 2016, Egea et al., 2017;), wheat (Elsayed et al., 2017), potato (Gerhards et al., 2016), cotton (Bian et al., 2019), and almond tree (García-Tejero et al., 2018a, García-Tejero et al., 2018b). Various studies has shown the significant spatial variability of crop water stress under different irrigation treatments (Bellvert et al., 2016; S. Gutiérrez et al., 2017, Romero et al., 2018). Mapping the variability of crop water stress has become crucially important for irrigation scheduling. In early 1960s, hand-held thermometer was primitively used to measure vegetative surface temperature for evaluating plant-water and plant-health status (Tanner, 1963, Fuchs and Tanner, 1966). During 1980s, CWSI was carried out for crop water stress detection by Idso et al., 1981, Jackson et al., 1981. However, thermal remote sensing techniques had not been widely adopted for irrigation planning until measurement instruments became affordable and efficient for precise measurement of vegetation surface temperature. In previous literatures, infrared thermal sensing for crop water stress detection was discussed as a part of remote sensing application in precision agriculture with aerial imaging system (Gago et al., 2015, Gerhards et al., 2019; Wouter H. Maes and Steppe, 2019, Karthikeyan et al., 2020). Maes and Steppe (2012) reviewed on the applications of ground-based thermal remote sensing in agriculture, specifically addressed different approaches (i.e. analytical approach, empirical and adaptive empirical approach) to calculate CWSI for evaluating crop water stress. In the past ten years, with the cost reductions of ground manned vehicles and unmanned aerial vehicles (UAVs), and advancement in technology of thermal cameras (with higher accuracy and reduction in size and weight of thermal cameras) has increased flexibility for thermal sensing in crop water stress detection under both ground and aerial scopes (i.e. canopy level and plot level).

Thermal imagery has been used to obtain average canopy temperature to calculate CWSI for diverse crops under different scales, fostering various image processing approaches for the segmentation of green vegetation in thermal imaging (Hoffmann et al., 2016, Lima et al., 2016, Matese and Di Gennaro, 2018, Petrie et al., 2019, Shivers et al., 2019). Technical developments in image processing and deep learning have offered new capabilities to extract canopy temperature and predict crop water stress. From the aspects that have not covered by recent review literatures, this paper focuses on the latest image processing techniques to obtain canopy temperature in infrared thermal imaging acquired on ground-based and UAV-based platforms, and addresses the correlation between different forms of CWSI and plant biological indicators. Additionally, the trends of deep learning for assessment of crop water stress with infrared thermal imagery is presented. Due to the recent rapid advancement in thermal imaging techniques and their applications for precision agriculture, synthesis and analysis of literature in this area is significantly important in providing state-of-the-art guidelines and potential future directions of thermal imaging in crop water stress monitoring, which lacks in the recent literature. Therefore, this article is focused on most up-to-date studies carried out in the area of thermal imaging applications in crop water stress monitoring, as well discussing the challenges and potential opportunities in this area. The paper is organized as follows: image acquisition platforms are introduced in Section 2, image processing methods for canopy segmentation are presented in Section 3, different forms of CWSI and the correlation with crop physiological indicators are summarized in Section 4, and finally, the potential use of deep learning approaches to assess crop water stress and the future trends are discussed in Section 5.