Study On Field Lea Image Segmentation And3D Reconstruction From A Single Image Machine Vision Algorithms

With the rapid development of computer technology and the Internet of things. to realize orchard production intelligent decision management by using the machine vision technology has become possible. To obtain the plant agronomic parameters visually is a important part of the whole process digital management of orchard production.The study on orchard leaf image segmentation and image reconstruction algorithm. is to solve the intelligent decision problems of production management.by the way of the real-time remote extraction of agronomic parameters in. This has important theoretical and practical significance.This paper. using orchard field leaves as the object of study. solve the problem of extracting parameters of agronomic using machine vision system by the remote access to agricultural production spot. Combining Integrated machine vision technology, image processing technology and pattern recognition technology, we research the segmentation method of online leaf image using adaptive threshold. Stduy the3D reconstruction method for recovering three-dimensional surface from a single image based on the region segmentation of leaf, and calculate leaf area. The usage of the real-time remote extracted agronomic parameters, can provide intelligent decision support for the production management. Focus on the following aspects of the work:QJln natural light conditions, some segmentation success ratio of field plant leaf with complex background captured by a real-time system is not high. In this paper we propose the gray mapping function and an adaptive threshold segmentation, the mapping method makes the segmentation algorithm used in this paper can obtain higher segmentation success rate. In the segmentation algorithm,we combine the edges segemented Canny and OTSU operator, remove the complex background and the strong interference targets.For the blade edge background noise, that cannot be removed by the operators based on edge detection and threshold segmentation, we design optimization method based on mathematical morphology and internal and external template, the target area segmentation results obtained after the optimization method continuous, remove adhesions and background, the leaf segmentation effect is very ideal.(2)In order to solve the problem of extracting the target area of plant leaf blades of different morphological features after segmentation, shape recognition algorithm and three kinds of connected region extraction method is designed, leaf area can be successfully identified from the connected region after segmentation, the method makes the segmentation obtaining complete leaf ratio is greatly improved. At the same time, the two-dimensional image is impossible in the problem of field vane position judgment. We design a template library of blade pose, extract the leaf different3D pose information from the projection of normalized shape characteristics of two-dimensional. We use K-nearest neighbor clustering to match the template, obtaining leaf pose information, and from the two only leaves the edge points related moment (Xian Changju chord angle moment). ③Set up the algorithm of3D reconstruction from single image surface according to brightness change. In Lambertian modle. using the fitting equation of the p.q value of the surface vector and the gray value, we establish the continuous integrable equation by blade micro Lambertian body composition. and use the constraint equation. obtain a set of solutions of the surface system of solutions. We use the point of each pixel on the surface of micro Lambertian body to construction triangulation to calculate leaf area.To meet the effect adjustment needs of brightness jump of the blade region reconstruction. we research an effective triangular interpolation method.④We integrate the remote network control machine vision system based on WEB and Internet.The system can set camera shooting movement rules. and remote real-time schedule the equipment movement, according to the control algorithm and the feedback signal. The system also can obtain the equipment parameters and real-time image data. Machine vision system designed in this paper can provide sample images for image segmentation method and leaf area calculation. and image recognition results feedback to the decision system, parameters of the plant traits can give support for the follow-up work of the greenhouse production management intelligent decision system.