Flower load is one of the earlier indicators of potential yield in fruit orchards. Usually, a higher flower clusters number are present on a tree than needed for an optimal production. The most used techniques to manage the flower load are manual, mechanical, and chemical thinning. The main issue is to calibrate these techniques on the base of the desired yield. Drone imagery, being able to collect highly detailed information, could offer a solution to automate flower counting since manual flower counting would be too laborious. The main goals of this study were to test an easy to use and quick be analyzed data acquisition for sudden field interventions, short computing time and reliability. This was achieved by applying and comparing two methodologies that allow to map apple's flower clusters density at full bloom stage by processing Unmanned aerial Vehicle's (UAV) imagery with binary classification and K-Nearest Neighbor algorithm, respectively. A comparison between the flower cluster estimation and the actual cluster load analysis highlighted that mapping flower clusters by binary classification is more suitable than machine learning in terms of image processing because it allowed to have a quicker (8 minutes) , easier and less noise - affected image analysis with R2 values ranging from 0,60 to 0,71. Furthermore, the proposed methodology seemed to be able to manage with a high spatial variability since it produced a map that clearly corresponded with zonal field conditions.
We present preliminary results of a joint radio-optical test carried out by the BLM bistatic radar system and a LLTV camera located approximately below the radar hotspot. The observations were performed during the peak of the Perseids activity in 2004. Tens of coincidences have been detected during the shower display. Some advantages of simultaneous radar-optical observations are discussed here.
Fruit weight is one of the factors taken into account when performing yield estimations together with the trees density and orchard's area. Thus, having the possibility to collect data about the weight of a large number of fruits in the orchard gives the possibility to increase the reliability of the yield estimation. Over recent years, mathematical models able to convert the fruit size into fruit weight were evaluated as effective. Since then, manual data collection with calipers and automated/continuous fruit gauges were tested to collect fruit size data to perform yield predictions. Their main drawbacks are respectively the need for human-labour, repetitiveness, being time-requiring and the limited sample varying from 20 to 200 fruits per hectare. This research is trying to discover and deepen the use of AI in agriculture for doing a step further: sizing fruits after their detection with a YOLOv5 Neural network algorithm. To reach this goal, a system which takes as a input RGB-D depth-camera's color images and 16 bit depth maps was developed. After applying YOLOv5 detection, two different methodologies (by mean of squared bounding boxes and circular shapes) to extract from the depth map the distance data needed to size the target object were tested. Results from a preliminary data-set showed that the system could be a potential solution to increase the sample dimension and perform yield prediction. The main drawbacks of the developed vision-system are related to the errors in sizing the objects, which are ranging from an underestimation of about 9 mm to an overestimation of 24 mm. From the initial results was possible to identify the squared-bbox-mediated sizing process as a better pathway rather than the one performed with circular-bboxes, since the RMSE is always smaller with values of 7–9 mm
Food quality standards are evolving to meet consumer demands and sustainability goals. Quality controls are essential throughout the supply chain, but manual assessment is subjective, time-consuming, and costly. Computer vision systems (CVS) offer a potential solution by integrating cameras and computers to automate the evaluation and sorting process. The study contributes to explore the use of computer vision on a new market segment, the one of red-flesh kiwifruit, with the purpose of ensuring consistent quality attributes for consumers while supporting the supply chain. For red kiwifruit, accurately assessing flesh redness poses a challenge, which was addressed in this study through the implementation of a robust CVS that i) exploits an artificial neural network to detect human perceived and commercially determined red shades in kiwifruit images, ii) computes image descriptors, iii) and grades the fruit with the use of unsupervised learning algorithm. Cohen's K-score analysis showed that machines have higher and more consistent agreement (k = 0.60) with the reference evaluation made by experienced fruit quality graders. Human inspectors demonstrated to produce evaluations affected by perception subjectivity and high variance throughout the day (k between 0.38-0.54). The analysis of Pearson correlation highlighted that the CVS shows a Pearson correlation in range 0.87-0.91 when compared to human evaluations.
A recent trend in scientific computing is the increasingly important role of co-processors, originally built to accelerate graphics rendering, and now used for general high-performance computing. The INFN Computing On Knights and Kepler Architectures (COKA) project focuses on assessing the suitability of co-processor boards for scientific computing in a wide range of physics applications, and on studying the best programming methodologies for these systems. Here we present in a comparative way our results in porting a Lattice Boltzmann code on two state-of-the-art accelerators: the NVIDIA K20X, and the Intel Xeon-Phi. We describe our implementations, analyze results and compare with a baseline architecture adopting Intel Sandy Bridge CPUs.
Although superficial scald (SS) is well characterized on apples, there are only a few insights concerning the influence that agronomic and management variability may have on the occurrence of this physiological disorder on pears. In this study, we aimed to improve our understanding of the effect of different preharvest factors on SS development using a multivariate statistical approach. Pears (Pyrus communis L.) cv “Abate Fetel” were picked during two consecutive seasons (2018-2019 and 2019-2020) from twenty-three commercial orchards from three growing areas (Modena, Ferrara, and Ravenna provinces) in the Emilia-Romagna region of Italy. Bioclimatic indices such as weather and soil, agronomic management such fertilization and irrigation, orchard features such as rootstock and training systems, and SS incidence were carried out at harvest and periodically postharvest in all producers. Two different storage scenarios (regular atmosphere and use of 1-MCP) were also evaluated. Our data in both seasons showed high heterogeneity between farms for SS symptoms after cold storage either in the regular atmosphere or with 1-MCP treatment. Nevertheless, in 2018, all the producers showed SS at the end of the storage season, but in 2019 some of them did not exhibit SS for up to 5 months. In fact, some preharvest factors changed considerably between the two seasons such as yield and weather conditions. Indeed, some factors seem to affect SS in both growing seasons. Some can increase its occurrences such as physiological and agronomical factors: high yields, late date of blooming, heavy downpours, improper irrigation management (low watering frequency and high volumes), nitrogen (included that deriving from organic matter), soil texture (presence of clay), orchard age, and canopy volume in relation to training system and rootstock. Others can decrease SS such as climatic and management factors: late harvest dates, rain, gibberellins, calcium, manure, absence of antihail nets or use of photoselective nets, and site (probably related to better soils toward the Adriatic coast). Initial preharvest variability is an important factor that modulates physiological plant stress and, subsequently, the SS after cold storage in “Abate Fetel” pears. Multivariate techniques could represent useful tools to identify reliable multiyear preharvest variables for SS control in pear fruit different batches.
