Browsing by Subject "Machine vision"

Now showing 1 - 1 of 1

Laser backscattering imaging in agriculture
(2023) Wu, Zhangkai; Müller, Joachim
Non-destructive optical sensor technology (NDOST) is an essential part of agriculture. The unique capabilities of laser notably enhanced NDOST. Laser backscattering imaging (LBI) is a technology that captures light patterns scattered by a material to analyze its properties. It is particularly suitable for agriculture due to its affordability and the optical scattering nature of agricultural products. The images generated by LBI are related to the optical parameters of the examined objects. Crucial tasks in LBI include the selection of an appropriate laser, the extraction of image features, and the utilization of a prediction model for analysis. LBI has been employed in numerous scenarios, such as maturity detection and drying monitoring. The main challenges for LBI involve establishing a precise theoretical framework and uncovering new applications within agriculture. This study aims to enhance the foundational knowledge about LBI and explore additional application scenarios. The first study focused on basic research about LBI. Currently, researchers rarely document the cell size of their samples and treat the optical coefficient as a constant within agricultural products, which is questionable. This studys purpose was to use glass filter matrices as controlled models and to evaluate the effects of pore size, different solutes, concentrations, and wavelength. The used porous glass discs had pore diameters ranging from 1 to 160 µm. We applied aqueous solutions of NaCl (0, 1, 2, 3, and 4 mol/L) and NaH₂PO₄ (0, 0.8, 1.6, 2.4, and 3.2 mol/L) to fill the pores. The LBI system incorporated laser modules at three different wavelengths (405, 635, 780 nm). The results illustrated that three of the four examined experimental factors (the pore diameter, the solutes, and their concentrations) have a substantial impact on LBI. However, no clear differences in LBI patterns were observed among the three utilized wavelengths. Consequently, when deploying LBI on fruits, for instance, a thorough consideration of cell sizes at various depths from the fruit surface is required. The second study focused on a possible application scenario of LBI: the sedimentation process of crude sesame oil. Oil sedimentation is a process where gravity is used to remove solid impurities, resulting in a clearer oil. This study examined the sedimentation process in crude sesame oil using LBI. In situ and laboratory experiments were conducted over 30 days, involving an LBI system directly attached to a transparent sedimentation tank with 120L of crude oil. Both the oil properties and sedimentation curve were analyzed along with the LBI images. There was a dramatic drop in oil particle-related properties (at least 87%), a 90% decrease in water content, and minor changes in oxidation-related properties. The sedimentation speed was about −7 mm/h, then became less than −2 mm/h, revealing two stages: diluted and hindered sedimentation. The crude oils surface displayed a distinctive scattering spot and a Tyndall effect within the oil, showing an increasing path length as sedimentation proceeded. The findings offer practical insights for enhancing sedimentation tank and LBI system design. The third study focused on another possible application scenario of LBI: leaf wetness measurement. Leaf wetness plays a pivotal role in managing plant fungi diseases. Existing optical techniques categorize leaf wetness as a binary problem – either wet or dry. In contrast, this research developed a platform capable of semi-automatically measuring droplet deposition on grape leaves using an LBI system. The leaf area, mean intensity per pixel in the red channel, and droplet count using information from the green channel were extracted from the scattering images. The study employed a generalized additive model (GAM) to predict leaf wetness with the extracted features. The prediction of the test dataset achieved an R-squared value of 0.78. The extraction of image features was found to be influenced by factors such as image resolution and leaf orientation. The method introduced in this study offers the potential for precise quantification of leaf wetness with an LBI system. In conclusion, our study highlights the importance of considering cell size in agricultural applications of LBI. Besides, LBI was found useful in monitoring plant oil sedimentation and quantifying leaf wetness. This suggests its potential for scenarios involving state changes in suspensions or colloids and differentiating materials with distinct optical properties. However, using a porous matrix as a model introduces an inherent error. Additionally, advancements are necessary to transition the application studies into practical use. Future LBI development could be facilitated by building a comprehensive database on light interactions with diverse cells and tissues.