Smartphone image-based framework for quick, non-invasive measurement of spray characteristics

• Autor(a) principal: Carreira, Vinicius dos Santos [UNESP]
• Data de Publicação: 2023
• Outros Autores: Nuyttens, David, Langenakens, Jan, Pereira, João Victor, da Silva, Rouverson Pereira [UNESP]
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Repositório Institucional da UNESP
• Texto Completo: http://dx.doi.org/10.1016/j.atech.2022.100120; http://hdl.handle.net/11449/249228
• topic: Agricultural spraying; Cloud computing; Digital system; Manufacturing; Spray nozzle

Agricultural spray nozzles are small yet key components for delivering agrochemicals to the target surface. Their complexity demands specific assessment to guarantee the spray quality. The traditional equipment measures spray characteristics accurately; however, it can be time-consuming and invasive. An emerging method for agricultural assessments is image processing. Therefore, our goal was to develop an innovative smartphone image-based framework to analyze spray characteristics (angle, single distribution and solid stream jets) and detect poor spray nozzles under indoor or outdoor environments. We used flat fan and hollow cone nozzles, with three sizes and three working pressures. A mobile phone with 12 MP was used to capture spray images. The images were automatically stored and processed using a cloud computing platform. Spray angle was estimated by finding edges in binary images and using Hough transform function. Single spray distribution shape was characterized by extracting the pixel grayscale value profile across a horizontal line in three regions of the image. Finally, solid stream jets were detected using a function to calculate grayscale peaks. The actual spray characteristics were measured using an automatic horizontal patternator with 120 collectors (25 mm grooves). As a result, image-based spray angle mean absolute error (MAE) and mean absolute percentual error (MAPE) were 4.45º and 4.40%, respectively, leading to an underestimation in most cases. Specific regions of the image accurately resembled the measured spray distribution shape, except for hollow cone nozzles when working at low pressures. Moreover, solid stream jets were correctly detected at three different flow rates. Our framework was validated for both indoor (laboratory) and outdoor (field) conditions and had similar results in detecting poor spray nozzles. The only restriction was the combination low/medium pressures with small nozzle sizes. Therefore, our insights absolutely contribute for quick, non-invasive measurements of spray nozzles characteristics through a digital approach and open pathways towards affordable and flexible methods for agricultural assessments.