Machine Vision Based Measure Of Phenotypic Characteristics Of Grape Organ

The grape fruit is an important fruit tree, it’s volume is only after the production ofcitrus fruit, ranking second in all kinds of fruit. In recent years, China’s grape productionand cultivation area and production continues to increase, which has become the world’slargest grape wine producer. But compared with the developed countries, such as Italy andFrance, China’s production as well as the quality of grape needs to be improved, whereirrigation on grape yield and quality has a significant impact. At present, China’s wineindustry is divided into ground furrow irrigation type, sprinkler irrigation, drip irrigationand so on. But with the increasing scarcity of water resources as well as the rapiddevelopment of modern precision agriculture, plant water status under the precise controlof the implementation of irrigation has become an important way to improve water useefficiency and fruit quality. The roots are vital organs of plants to obtain nutrients from thesoil and water, thus the growth conditions of the roots has a critical influence on the plant.Meanwhile, the plant roots act as an indispensable organ, its phenotypic characteristics ofthe dynamic changes indicate the plant moisture, growth modeling of real-time monitoringand an important basis for plant growth. At present, research on phenotypic characteristicsof plant roots in the dynamic observation method is still relatively backward. UnitedStates, Italy, France and other developed countries have made major developments inagricultural root phenotype research areas. For roots hidden below ground, it’s difficult tomake direct observations of this feature. Thus scholars and experts have proposed microroot window technique, MRI magnetic resonance imaging techniques and CT computedtomography technology and other non-contact non-destructive measurement method, tofacilitate people with study on characteristics of root growth, which has brought greatconvenience. Compared to destructive contact measurement method, non-contact methodhas the following advantages: no destruction of plant organs and growing environment,with test repeated several times, and no data loss and distortion. Therefore, on the basis ofnon-contact and non-destructive measurement, measurement and analysis on plant roots morphology parameters and phenotypic characteristics was carried out, which castimportant guiding significance on the research.In this paper, we base on the root-domain-limit greenhouse grape cultivation methodfor the study, combined with robot vision and image processing measurement and otherkey technologies, through continuous monitoring grape root growth and by observinggrape phenotypic variation, we provide data support and strategic guidance for grapegrowing modeling and precision irrigation. In this paper, we target on the theory ofmachine vision measurement and processing algorithms, regarding root restriction undergrape fruit and roots phenotypic characteristics extracted as research content, we conductmeasurements research on the characteristics of grape critical organs systematictheoretically and experimentallyThe main research contents are as follows:1. According to characteristics of root domain restriction on greenhouse environmentgrape growing, we developed a set of visual monitoring system for greenhouse grapeswith high throughput and noise immunity, which includes image acquisition vision sensor,MCU-based controller, and soft lighting and position-check continuous measurementsystem for non-overlapping/overlapping phenotypes grapes micro changes. The system isrobust with high precision, high throughput and good noise immunity. It has overcome thecomplex background and uneven illumination of grape images, and thus can better meetrequirements for continuous non-destructive monitoring of grape fruit.2. Regarding the issue of access to phenotype of grape fruit under complexbackground, a FCM-based fuzzy clustering algorithm as well as local information basedactive boundary model ACM algorithm for image segmentation method were proposed,combined with the morphology algorithms of digital image processing to remove falseboundaries. In this way, contour eventually converges to the borders of target fruit, andeventually accurate information of grape outline was obtained.3. For the phenotypic characteristics of grape root and particular issues for theirgrowth environment, a set of visual monitoring system was developed suitable for plantroots micro-variable observation, where micro root window methods was applied,combined with scanner and macro shot video camera, so that images of grape roots localinformation was acquired. Then employ FCM fuzzy clustering algorithm on the image oflocal information, and thus precise contour of grape root was acquired, which providesdata support for subsequent analysis.4. Based on grape root image segmentation results, and combined with soil waterpotential and temperature parameters, the experimental results were corrected, andregression analysis and summary were conducted. Then the phenotypic characteristics of grape root enlargement growth rhythm were derived, which in turn lead to the grape rootgrowth model.Dissertation focuses on the content of the above, carrying out research onnon-destructive measurement of grape root based on computer vision; designed a set ofmachine vision inspection system for grape organs under conditions of greenhousecultivation, as well as studied the grape root image segmentation algorithm and testifiedthe feasibility and effectiveness of the machine vision solutions through experimentalmethods.