Research On The Geometric Parameters Measurement Method Of Plant Leaves With High Progress Based On Machine Vision

High precision measurement of geometric parameters of plant leaves has been widely used in intelligent agriculture.The measurement results of leaf geometric parameters can be used as important technical indexes in plant growth potential analysis,plant species identification and crop yield prediction based on LEAF area index.The accuracy and efficiency of measurement will directly affect the reliability of subsequent processing results.At present,the measurement method of geometric parameters of plant leaves mainly relies on the combination of manual assisted calibration and hardware equipment.Although this method can achieve the measurement of geometric parameters of plant leaves,there are still some defects.For example,manual assisted calibration is prone to calibration errors,resulting in precision loss.In the process of blade measurement,only one blade parameter can be obtained at a time.The measuring equipment has a large volume and is not convenient for long distance handling.This paper mainly studies the above three kinds of problems in depth.A geometric distortion correction algorithm based on the marker matrix combination and improved Hough transform is proposed.Multi-blade clustering algorithm based on improved connected region;At the same time,an automatic drawing algorithm of external moment based on spindle rotation Angle is proposed.In addition,with the help of the idea of image pooling operation in deep learning,the algorithm is optimized to improve its efficiency.The specific work of this paper is as follows:(1)In view of the large volume of the existing measuring equipment,which is not easy to carry,this paper designs a simple blade sample image acquisition equipment with CMOS HD industrial camera as the core,with an external ring light source and a high-brightness back light plate that can rotate 360 degrees.The equipment has the advantages of easy to carry and strong environmental adaptability.(2)In terms of algorithm design,due to the portability of hardware devices,related components of peripheral circuits are reduced to the greatest extent,and only the core sample image acquisition module is retained.Therefore,the geometric parameter measurement of blade samples mainly depends on the algorithm.In terms of algorithm implementation,this paper adopts the traditional Hough Transform(Hough Transform,HT)and Connected Domain Clustering(Connected Domain Clustering,CDC)combined with the way of the location of the distortion of image restoration,using the Connected area Clustering,position of blade samples for testing,and to adjust its grey value,the leaves gray value is set to agree with background grey value,eliminate the influence of the blade for straight line detection,then using Hough Transform,corner detection target area,complete the geometric distortion correction;In order to realize multi blade multi-parameter measurement at the same time,this paper puts forward a multi blade clustering algorithm based on connected area,for the traditional clustering algorithm,mostly USES the grey value as the clustering based on,in different regions of a single goal for clustering,the method has good treatment effect,but for multi-objective clustering,is slightly less than,by traditional clustering algorithm,it is easy to be different color of target area as part of the get together for a class,in order to solve the problem,this paper USES the connected domain clustering algorithm based on the location attribute,avoid color attribute of cross-regional clustering problem;Finally,for the drawing of the minimum external moment of the blade,this paper proposes a drawing method based on the principal axis rotation method.By drawing the external moment,the rotation Angle is determined,the convex envelope is adjusted,and the boundary point of the external moment is located by using the contour attribute of the convex envelope.(3)Software interface design.In order to better display the measurement results of multiple leaf geometric parameters and make it easy to use,this paper USES Matlab GUI to design a plant leaf geometric parameter measurement system software.With UI design as the core,the software is divided into user login interface,functional menu interface and blade parameter measurement interface.The user login interface can realize login by adding user name and password or login by using authentication code.When the user information is entered incorrectly,a prompt for re-entry will be made at the bottom of the login interface.In the functional menu interface,in addition to the built-in blade geometric parameter measurement system,commonly used modules such as image morphology processing,contrast enhancement,noise reduction and so on are added to ensure that the system can process the diversity of images.To sum up,this paper has carried out in-depth research on the high-precision measurement method of multi-blade geometric parameters,and designed a portable field image acquisition equipment and corresponding upper computer software module.For the core algorithm implementation module,the geometric distortion correction algorithm based on marker matrix combined with Hof transformation,the multi-blade clustering algorithm based on improved connected region and the minimum external moment rendering method based on the principal axis rotation method are proposed.The experimental results show that compared with the traditional geometric parameter measurement method,this method has higher measurement accuracy and faster measurement speed.