| Leaf Area Index(LAI)is an important parameter to measure plant growth.It is also the basic parameter of many mathematical models in agricultural science,ecological science,and remote sensing science.The Leaf Area Index controls many biological and physical processes in the water,nutrient,and carbon cycle.LAI is defined as the area of a green leaf per unit of surface area on one side.For most of the current LAI indirect measurement methods,there are defects such as large errors,low accuracy,and low stability.Based on the photography method,this paper applies computer vision technology to LAI measurement,and conducts research on LAI value extraction algorithms from two aspects:single&top-view method and multi-view precise construction of plant three-dimensional models.This article mainly studies the leaf area index extraction algorithm based on photography from the following aspects:(1)Research on leaf area index extraction algorithm based on single&top-view method.Exploring and researching four aspects of the relationship between camera light radiation,photographic material’s photosensitive characteristics,digital camera ISO,and object relative light radiation measurement.The measurement principle of LAI based on the top-view method is proposed:L=Inη.The error of the measurement result from the commonly used measuring instrument LAI-2000 is within 0.5,and the result is highly accurate.(2)Study the basic framework of plant 3D reconstruction based on multi-views,and reconstruct the 3D plant model with higher accuracy based on Open MVG and Open MVS framework.A multi-view 3D reconstruction system was built.Based on the image sequence,the three-dimensional structure of the plant in the laboratory environment and the field environment can be accurately reconstructed.(3)Perform preprocessing operations on the three-dimensional structure of the plant,including:outlier removal based on the K-nearest neighbor algorithm and point cloud segmentation based on the random sampling consensus algorithm.The skeletonization algorithm and the random intercept node algorithm are used to automatically detect the stems and completely filter the stems of the plants,which is convenient for the subsequent experimental research to analyze the influence of the stems on the measurement.(4)Extract the leaf area index based on the three-dimensional point cloud of plants after pretreatment.First,count the leaves,and then use the SOM neural network model to calculate the area of the leaves.The experiment showed that the R~2 between the measured value of the leaf area of the plant and the real value measured based on LI-3000reached 0.97.The experimental results before and after the smoothing treatment of peppergrass increased from 0.9657 to 0.9741,and the R~2 of simulated corn increased from0.8319.To 0.9683.(5)Analyze and compare the deficiencies of a variety of leaf area index extraction methods based on the three-dimensional structure of plants.A method for calculating LAI based on the ratio of the plant leaf point cloud area to the plant’s land area based on the three-dimensional reconstruction structure is proposed.The calculation method of leaf area index used in this paper is compared with the real value of LAI,and the accuracy rate is as high as 98.63%.The LAI measurement values before and after the smoothing treatment and the stem filtering treatment are compared.The experimental results show that the smoothing treatment effect on accuracy of the LAI is 6%to 18%,and the effect of the stem is 10%～22%. |
