Characterization of vine canopy through two dimensional imaging

Abstract

Abstract. In the last decade, reconstruction of plant dimensions and volume has gained a noticeable importance, in particular for the possibility of collecting data correlated to biomass, leaves area, etc. This is specifically of interest in the case of vineyards, where knowledge of variability can be not only a useful mean to evaluate the health condition of the vines and of the grapes, but also an important input to allow variable management practices. Different sensing technologies are available for shape reconstruction, as for instance ultrasonic sensors, laser scanners or depth cameras, however, their practical application is still limited by low quality information, high costs, or high speed data processing demand. For the present work two-dimensional imaging is proposed as a viable solution for shape reconstruction of canopies. The method takes advantage of low cost 2D commercial cameras, which can be installed on board the tractor allowing on the go collection of images of plants from the bottom. The paper describes the instrumentation set up and integration with a specific thresholding algorithm allowing segmentation of canopy profile. Tests were carried out on 20 different dates on a glera vineyard in an experimental farm in the North of Italy. The results were correlated with the number of leaves, leaf area index and canopy volume. High correlation was identified in the case of volume with coefficients of determination R² > 0.7 in most of the cases.Keywords: canopy, profile, imaging, segmentation. IntroductionThe opportunity of analyzing the variable profile of vineyard canopies is of the highest relevance for the possibility of precisely quantifying local biomass. Such interest arises from the correlation, which is typically found between the canopy size and plant needs in terms of water and nutrients as well as with the yield and quality [1]. As a consequence, knowledge of plant dimensions allows not only evaluation of the plant health conditions, but also exploitation of variable rate and management practices, including crop protection and canopy management [1][2][3].Many researches and approaches have been proposed in the last decade, which allow quantification of the plant dimensions [1; 3; 4], also taking advantage of the availability of miniaturized low cost sensors and of an increased speed in data processing [5; 6]. Available technologies can be referred mainly to ultrasonics, laser sensors and other optical methods [7].Ultrasonic technology is a low cost solution, which relies on the sound waves echoing off of a given target [8; 9] to provide its external contour. Its performance is influenced by ultrasound propagation, which often exhibits parasitic signals or losses especially in the cases of thin targets or high reflection angles, as is in the case of small branches or in the case of badly oriented leaves. Additionally, ultrasonic sensors are very much sensitive to humidity, dust and high temperatures, however, their practical implementation has provided interesting results especially cou...

Structure from Linear Motion (SfLM): An On-the-Go Canopy Profiling System Based on Off-the-Shelf RGB Cameras for Effective Sprayers Control

“…A target sprayer for orchards that identifies the crown by extracting the green color in the HSV color space followed by a morphological erosion operation, aiming the camera upwards at an angle of 30°to capture the entire crown while having the sky as a background, is described in [14]. With a similar approach, the authors of [15] placed the camera facing the sky near the ground and analyzed the correlation between the leaf area of the vineyard measured by the segmented image with its volume or using a professional instrument for measuring LAI [16], whose operating principle is the measurement of UV-blue (320-490 nm band) instead of green. The segmentation based on color information always suffers from the context, typically of the same color (grass of the ground or adjacent rows).…”

“…The depth data that the camera produces is treated in a similar way to what is proposed with LIDAR, or the depth image is binarized by setting a lower and upper threshold for the distances of interest. With this device, there are two advantages: the most important is the possibility of having a continuous mapping of distances even at speeds of a few meters per second or, in the case of low speed, there is a large number of points measured several times in successive frames, whose data can be discarded or averaged; another advantage is the availability of the corresponding aligned color image, which can also be segmented (as reported in the works in the Introduction based on colors [13][14][15], shapes [12] or using classifiers [25]) to increase the information available to more precisely recognize only the leaves.…”