Research On Method Of Crop Phenotypic Parameter Measurement And Growth Status Monitoring Based On Image Processing

With the increase of population and the decrease of cultivated area,it is more and more important to ensure crop yield and quality for China,a large agricultural country.Measuring the crop's phenotype,which refers to the characteristics of crops,is the basis for monitoring the growth status of the crop that ensures the crop's yield and quality.There are disadvantages of the traditional phenotypic parameter measurement,such as manual measurement and sensor measurement:strong subjectivity,low repeatability,and damage to crops.Using image processing technology to restore the three-dimensional information of the crop and measure its phenotypic parameters is not only nondestructive and repeatable,but also brings more accurate measurement results.For accurate measurement of crop phenotypes,this thesis focuses on the development of non-destructive and accurate measurement system for crop phenotypic parameters,which is based on image processing technology with the cost and practicality in consideration,to monitor the growth status of crops during their growth.This thesis takes pepper,a common crop,as the experimental objects,uses the binocular stereo vision model for crop images collecting,and combines with the camera calibration,stereo matching of image pair,optimization of disparity maps,and crop segmentation,to realize the phenotypic parameters of crop measurement by the restoration of its three-dimensional information.The main research contents are as follows:The image capture is built by binocular stereo vision system,with the cameras and lens selected under a comprehensive consideration of several conditions,such as target field size and measurement accuracy requirements.The conversion from world coordinate system to image coordinate system is deduced.After analyzing the advantages and disadvantages of various calibration methods,Zhang Zhengyou calibration method is selected for camera calibration and the calibration error is less than 0.1 pixels.Finally,the cause of the error is analyzed.To improve SAD's poor matching effect on simple textures and the Census'limited accuracy for overall matching,the combination of the two is used as the matching cost for stereo matching.Then,the disparity map is further optimized by left-right consistency detection,median filtering,and disparity filling.The matching results meet the needs of subsequent research.The whole crop segmentation is realized by the color factor for grayscale,calculated by the normalized color components,combined with fixed threshold segmentation and small area deletion of the connected domain.The experimental results show that this method can completely segment the crop from the complex background which is less affected by light.For single plant crop segmentation,threshold segmentation of the disparity map is proposed to achieve the segmentation of the crop to be tested from other crops.The experimental results show that when the crops have a certain distance,the front crop can be completely separated from multiple crops.This method provides a solution for measuring single crops in the practical application.The crop point cloud image is obtained by the three-dimensional coordinate calculation formula,which verifies the reliability of the method for the conversion from 2-D to 3-D.The extraction and measurement methods of crop phenotypic parameters are designed:the plant height and width are calculated by the distance between the key points of the crop.The stems of crops are extracted by morphological operations and then refined to obtain the skeleton.The stem angle was calculated according to the three-dimensional coordinate linear fitting of the skeleton pixel.Compared with the manual measurement results,the calculation errors of these phenotypic parameters are within 2%.Finally,the experiment on phenotypic parameter measurement was conducted on the same crops with different measurement distances,different camera angles and multiple crops with different growth environments.The experimental results show that the measurement results of the measurement system are accurate and can better implement the function of tracking and monitoring the growth status of crops during their growth.Then the system construction and measurement efficiency are analyzed.The system realizes the non-destructive automatic accurate measurement of the crop phenotypic parameters and the monitoring of the growth status.And the function of multiple crops segmentation provides a solution for the practical application of crop segmentation.At the same time,the three-dimensional measurement function in this system is also common to other types of crops,and the phenotypic parameter extraction method also provides reference value for the phenotypic measurement of crops with similar structures.