Abstract:The morphological traits are important to investigate the state of plant. Measuring the morphological traits periodically during plant growing and fitting the growth model can be helpful to monitor the state and get dynamic growth rule of the plant. And growth model’s visualization can be more directly to show the dynamic changes and predict the plant growth tendency. To speed up and promote the normalization of the measurement of morphological phenotypes, using Arabidopsis thaliana for example, a lowcost machine vision system was designed which can be used to measure the morphological phenotypes of Arabidopsis thaliana during its growth process. With the growth data getting from the system, the plant growth equations and visualization model can be built. A platform was set which consisted of two main parts, fixed part for loading plant and moving part for carrying visible camera, to make sure that the plant would not shake so that can get clearer image sequences. Structure from motion (SfM) was used to get the 3D point cloud from the image sequence. Because of the weakness of SfM, which made the coordinate system generated each time different, a preprocessing algorithm to point cloud based on color panels board was designed to standardize every plant 3D point cloud model’s coordinate system as one. Under the stage for loading plant of the platform’s fixed part, a color panels board was set, which was a black board on which two red panels consisted of two linestyle and a rectanglestyle and one blue panel, and would be transformed to a part of the 3D point cloud. After filtering procedures, the areaofinterest of Arabidopsis thaliana was extracted from the original point cloud. To test the reliability of the color panels board, a 3mm×3mm blue square was fixed on the platform for a repeat trails. Firstly, three kinds of board were used, on which red panels were only linestyle, only rectanglestyle and both of them respectively, for three testing groups. Each testing group had 30 3D point cloud models from the same 10 plants and each plant was collected from three different camera perspectives. Secondly, the method to standardize every 3D point cloud model’s coordinate system was used. Then the centroid coordinate of 3mm×3mm blue square’s point clouds on each model was got, and the Euclidean distance between the centroids in each testing group was calculated. Throughout the value of contrast test, the mean absolute percentage error of leaf width, leaf length, main stem’s length, leaf area and angle between leaves were 9.83%, 10.10%, 1.07%, 4.09% and 4.37%, respectively. A timeseries morphological phenotyping data of three Arabidopsis thaliana samples were collected and used to fit a mathematical model. After that, the model was visualized on Lstudio with L-system.