Original papersDetecting fruit surface wetness using a custom-built low-resolution thermal-RGB imager
Introduction
The surface wetness and temperature of fruit are important parameters in pre- and post-harvest crop loss management. Some varieties of fresh-market fruit, such as sweet cherry, tomatoes and grapes are sensitive to the presence of moisture on fruit surface (i.e. fruit surface wetness), which may lead to fruit cracking/splitting and loss of fresh-market values. For example, sweet cherry faces the challenge of crop loss up to 90% in some varieties due to fruit splitting/cracking from seasonal rains prior to harvest (Zhou et al., 2016). Sweet cherry fruit cracking caused by rainwater is a major source of financial loss to growers around the globe and especially in the pacific region of the United States. Timely removal of rainwater from fruit surface requires accurate and real-time monitoring of fruit wetness in the field as part of a crop loss management system. Surface wetness and duration are also used for monitoring, prediction and management of disease in high value crops (Sutton et al., 1984, Rowlandson et al., 2015). Previous evidence has shown that bacteria and fungi are closely related to the humidity and wetness of fruit and canopy, and can be used to determine critical times for chemical spray applications (Llorente et al., 2000, Peres and Timmer, 2006, Duttweiler et al., 2008). The surface humidity and temperature of diseased fruit may be different from that of healthy fruits. Quantification of fruit surface wetness and temperature could thus provide a promising way for monitoring and prediction of disease/pests in tree fruit crops.
Considering the difficulty involved in direct measurements of fruit surface wetness, leaf-shaped flat plate type sensors are currently being used as the main indirect methods of quantifying surface wetness (Sentelhas et al., 2007). This type of wetness sensor, however, has shown to be very unreliable. The principle behind wetness sensing is relating either the dielectric constant or resistance of a grid of copper wires printed on the substrate surface to the presence or absence of water. The wetness sensor sampling size is limited to the area of the sensor and in some cases, the sensor may not even detect any moisture due to its small size. Moreover, thermal properties and shape of current wetness sensors are very different from that of an actual fruit making them incompetent of mimicking fruits (Hatfield, 1982).
In recent years, thermal sensing has found its way into many applications in precision agriculture (Khanal et al., 2017). Field scale remote sensing based on infrared thermography is known to be capable of characterizing leaf wetness duration (Sankaran et al., 2010, Mahlein et al., 2012). Ramalingam et al. (2004) showed the feasibility of using a sensor array comprised of a multispectral imager and an infrared thermometer in detecting and quantifying leaf surface wetness in a greenhouse. They developed a spraying system based on information from the sensor array, which was able to detect and spray a dry spot within a canopy. Leaf surface wetness in the form of dew has proven to be detectable using ground-based radiometry (Pinter, 1986, Hornbuckle et al., 2006) and satellite imagery (Cosh et al., 2009). Heusinkveld et al. (2008) developed a remote optical wetness sensor for field applications, which scans the surface and analyzes resulting spectral reflectance at two wavebands. They reported that the sensor was able to detect both leaf surface wetness and internal water.
In a previous study, we harnessed the power of recent technological advancements in the area of thermal sensing and single-board computers to develop an inexpensive thermal-RGB imaging system (Osroosh et al., 2018). This system was later used for monitoring apple sunburn in two apple varieties (Chandel et al., 2018). In the present study, our main goal was to evaluate the same system for monitoring the level and duration of cherry fruit surface wetness. To the best of our knowledge, combined thermal and RGB imagery has not been used for monitoring surface wetness and duration especially for small fruit surface wetness monitoring. The specific objectives were to (i) use a fully automated thermal-RGB imagery-based data acquisition system for monitoring sweet cherry fruit surface under field conditions, (ii) develop image-processing and computer vision algorithms to extract cherry fruit surface temperature from acquired images, and (iii) relate fruit surface temperature and microclimate measurements to fruit surface wetness, and evaluate the performance of the system in detecting fruit surface wetness.
Section snippets
Experimental field
A field experiment was carried out in plots of Skeena (Y-trellised) and Selah (vertical canopy architecture) cherry varieties at Washington State University Roza Farm near Prosser, WA (latitude: 46.38°N, longitude: 120.46°W, and elevation: 239 m above sea level). The experiment was conducted in the 2017 season at the fruit maturing stage of selected cultivars, two weeks prior to the commercial harvesting window.
Microclimate monitoring system
The microclimate information including wind velocity, air temperature and relative
Lab calibration of thermal-RBG imaging modules
The thermal-RGB imager calibration results are listed in Table 1. It can be seen that the coefficient of determination was high (), and the RMSE, and MAE after calibration were small (; ). The results indicated that the measurements were free of outliers (). The average error of thermal modules without calibration was . The highest error reached was which was about half the value reported by the manufacturer ().
Microclimate during the field trials
Fig. 4 illustrates the fluctuations
Conclusion
In this effort, the possibility of using a low-resolution thermal-RGB imagery-based system for monitoring the surface temperature of sweet cherry fruits and quantifying the wetness level and duration was assessed. The system relied on inexpensive revolutionary thermal and RGB camera modules. An algorithm was developed to extract the surface temperature of cherries from thermal images. Such systems were deployed in a cherry orchard and a rain simulator was used to apply water and wet the cherry
Acknowledgements
The authors would like to thank Dr. Matthew Whiting, Mr. Rajeev R. Sinha, Mr. Haitham Y. Bahlol, and Mr. Chongyuan Zhang, Washington State University for their help in the completion of this study. The authors would also like to thank Dr. Lav Khot, Washington State University for his help with editing the manuscript draft and facilitating the experiments. The authors also acknowledge the assistance, funding and support from the Center for Precision and Automated Agricultural Systems (CPAAS) at
References (20)
- et al.
Thermal-RGB imager derived in-field apple surface temperature estimates for sunburn management
Agric. For. Meteorol.
(2018) - et al.
Observations of dew amount using in situ and satellite measurements in an agricultural landscape
Agric. For. Meteorol.
(2009) - et al.
A new remote optical wetness sensor and its applications
Agric. For. Meteorol.
(2008) - et al.
The effect of free water in a maize canopy on a microwave emission at 1.4 GHz
Agric. For. Meteorol.
(2006) - et al.
An overview of current and potential applications of thermal remote sensing in precision agriculture
Comput. Electron. Agric.
(2017) - et al.
Economical thermal-RGB imaging system for monitoring agricultural crops
Comput. Electron. Agric.
(2018) - et al.
Evaluation of the Alter-Rater model for spray timing for control of Alternaria brown spot of Murcott tangor in Brazil
Crop Prot.
(2006) Effect of dew on canopy reflectance and temperature
Remote Sens. Environ.
(1986)- et al.
A review of advanced techniques for detecting plant diseases
Comput. Electron. Agric.
(2010) - et al.
In-field sensing for crop protection: efficacy of air-blast sprayer generated crosswind in rainwater removal from cherry canopies
Crop Prot.
(2017)
Cited by (10)
Highly sensitive graphene oxide leaf wetness sensor for disease supervision on medicinal plants
2022, Computers and Electronics in AgricultureCitation Excerpt :During the field measurements, the LWS sensors are kept at 45° angle as suggested in (Sentelhas et al., 2004). The rationale for selecting the Phytos 31 LWS as reference system is that, many researchers have used it as reference sensor for the LWD measurements (Osroosh et al., 2019, Hornero et al., 2017, Gao et al., 2020, Jia et al. 2019) Fig. 8 (a) shows the measured frequency data from the fabricated GO LWS from three different nodes recorded for about 9 days. As depicted in Fig. 8 (a) it comprises of 6 events, where event is referred to the time taken by the sensor to change its value from the base line (t1) and again attaining the baseline value (t2).
Intelligent thermal image-based sensor for affordable measurement of crop canopy temperature
2021, Computers and Electronics in AgricultureCitation Excerpt :It would also be possible to implement an image classification model to determine the specific phenotype (Azlah et al., 2019) for which the measurement is to be made, thus allowing to use dedicated models to enhance the performance on image segmentation. Morphological shape and colourimetrical variations of the fruit (Li et al., 2016; Lin et al., 2019b, 2019a; Osroosh and Peters, 2019) are also possible indicators derived from visible image processing that can be used for optimal harvesting predictions or phenological stage estimation of the crop (Vicente-Guijalba et al., 2014). Currently, several low-cost devices for the determination of the crop water status based on the measurement of the canopy temperature have been developed (García-Tejero et al., 2018; Noguera et al., 2020).
The Vision-Based Target Recognition, Localization, and Control for Harvesting Robots: A Review
2024, International Journal of Precision Engineering and ManufacturingFactors influencing fruit cracking: an environmental and agronomic perspective
2024, Frontiers in Plant ScienceA comprehensive review on acquisition of phenotypic information of Prunoideae fruits: Image technology
2023, Frontiers in Plant Science