توصيفگر ها :
ريشه ميانگين مربعات خطا , ميانگين مطلق درصد خطا , ميانگين اندازه نقاط كليدي , گرهنقطه , خطاي بازافكنش , ساختار ناشي از حركت
چكيده انگليسي :
Measuring the vegetation structure of the plant with the help of 3D images is a non-destructive method that can be used to monitor the plant status during the growing season. Among the passive optical distance measurement methods, structure from motion is considered as one of the most common approach in which the two-dimensional images are used to reconstruct a three-dimensional model. The aims of this research is: 1) to design and develop a 3D image capturing system using the structure from motion technique, 2) to investigate the optimal number of images in order to minimize the noise of 3D reconstructed models, and 3) to determine the effect of drought stress on the 3D geometry of tomato plant and comparing them with physical measurements. Therefore, a system with the ability to rotate camera 360 degrees was developed which includes of: a step motor, a chassis, a power transmission mechanism (including toothed belt and pulley), a pot holder plate, a camera support rod, a carrier wheel, a digital camera, a controller, an imaging chamber and two lighting lamps. The control part of the automatic imaging system was included of: an Arduino board, a stepper motor driver with microstep capability, power supply and a LCD screen. To investigate the effect of irrigation regime on tomato plant geometry, 60 seedlings of CH Falat cultivar were divided into three groups of 20 and subjected to three different irrigation regimes. Plants images were taken at every 10° rotation interval of camera and at 3 different camera installation angles. Three drought stress treatments were tested in this study including: irrigation up to field capacity (C), the interruption of irrigation until plants with wilt severity reaching 80% and then rewatering (T1), and the interruption of irrigation until the plant leaves wilted thoroughly (T2). To minimize the reconstruction of unwanted items, a black background was used, and to increase the number of recognizable features in the images, the surrounding of the plant pot was covered with a checkered pattern. A checkerd cube was placed near the plant pot to help align photos taken from the viewing angle of 180 degrees. Images were loaded into Agisoft PhotoScan software, an automatic feature extraction algorithm were conducted to align images. In the following a point cloud and a 3D solid model were reconstructed. After scaling the 3D models, surfaces of the geometric models of tomato plants were estimated by Cloudcompare software in which the total surfaces of triangular meshes are measured. Stem height and diameter were determined by a caliper and the leaf area was estimated by scanning and using ImageJ software. These dimensions were considered as the reference values. Results showed that the position of checkerboards has an effect on accuracy of the 3D models. 3D models quality decreased by increasing the interval angle of consecutive images by more than 10 degrees. Background elimination helped to reduce the 3D model reconstruction time by 42%. Results also showed that the minimum error in leaf area estimation will occur when images of the three camera installation angles (i.e. 75, 135 and 180 degrees) are used. Comparison of the estimated and the reference values revealed that the coefficients of determination in estimating stem height, diameter and leaf area were 0.98, 0.96 and 0.99, respectively. For the control plants the increase of 59.68%, 3.75%, and 17.97% in leaf area, stem diameter and height were observed, respectively. For T1 treatment, before rewatering, the leaf area, stem diameter and height respectively decreased by 43.86%, 12.11% and 3.2%. However, after rewatering, the leaf area, stem diameter and height increased by 34.4%, 0.61% and 32 2%, respectively. For T2 treatment, seven days after the interruption of irrigation and when plants wilted permanently, 51.99% reduction in leaf area, 17.38% reduction in stem diameter and 18.67% reduction in stem height were observed.
