Research On The Key Techniques Of Plants Growthform Point Cloud Date Processing And Precision Irrigation Control

In the face of increasingly scarce water resources,agricultural water shortage and other outstanding problems,the application of precision irrigation technology has become the inevitable development trend of modern agriculture.Plant environmental parameter collection,growth pattern monitoring,high-efficiency irrigation model,intelligent irrigation equipment and precision control strategy are the core contents of precision irrigation.According to the basic national conditions of precision irrigation technology in China,this paper focuses on plant growth morphology monitoring technology and precise control technology of irrigation flow valve to solve the problem of difficulty in obtaining plant growth information and poor control precision of intelligent irrigation equipment,so as to lay a foundation for the precision irrigation of plants.In order to obtain plant growth state without damage,high efficiency,real time,and no pollution,this paper builds a three-dimensional system of plant morphology based on low cost TOF depth sensor camera,and obtains,de-noises,and coordinates plant three-dimensional cloud data.The construction of plant growth state is realized to provide technical support for real-time monitoring of plant growth status.In order to realize precise irrigation,the motor control theory of flow valve for permanent magnet synchronous motor based on non-smooth composite control is proposed and experimental research is carried out.The main research contents and conclusions include:1.The accurate 3D point cloud data for the reconstruction of plants and realize the digitization of plants can be implemented by high-precision 3D laser scanning technology.However,it need to use the large amount of equipment,which will cause the higher cost.And also,the corresponding processing algorithm is complex and needs large amount of computation to deal with the complicated figure data,which limit its use in the more demanding fields of real-time.From the above,it is great theoretical and practical significance to design a good practicability and low-cost method for obtaining plant 3D point cloud data.TOF(time of flight)depth sensor is a novel scheme to acquire the depth information based on the principle of time of flight.The TOF method has some advantages,such as acquiring the depth information of the image quickly,the calculation accuracy did not change with distance and the depth calculated was not affected by surface gray level and characteristics.In spite of those superiorities,according to author knowledges,there are few research results on the application of TOF depth sensor in plant digitalization and visualization.For the purpose of improving the accuracy of 3D reconstruction for plants and realize the digitization of plants,in this thesis,an obtaining and denoising method of plants 3D point cloud data by using TOF depth sensor algorithm is proposed.The proposed method has the advantages of low cost and good practicability.The detailed steps are described as follows.Firstly,the TOF depth sensor is used to obtain the plant point cloud data,and then a direct pass filter is applied to reduce the background noise in advance.Secondly,the noise in the point cloud data has been detected and filtered by utilizing the improved outlier denoising algorithm,which combining the 2 characteristic parameters of neighboring average distance and neighborhood point number together.Finally,small-size noise in the point cloud was detected and denoised by bilateral filtering algorithm.In experiment,the tomato plants were selected to verified the validity of the proposed method.According to the experimental results,compared with the traditional bilateral filtering algorithm,the maximum error is reduced by 11.2%,and the average error is reduced by 23.2%.And in contrast to the Laplace filtering algorithm,the experimental results showed that the maximum error is reduced by 20.6%,and the average error is reduced by 39.2%.Thus,the proposed method could not only keep the characteristics of point cloud,but also simply and effectively reduce the noise in different scales of 3D point cloud data.2.Three-dimensional reconstruction technology is used to build a precise three-dimensional shape model of crops.Analysis and study of the plant morphological structure and physiological function is of great significance.At present,there are three main methods for three-dimensional reconstruction of agricultural and forestry crops,including rule-based methods,scanning methods and three-dimensional digitizer based methods.At present,although three-dimensional digitizer based method is a accurate method to describe plants,and the trueness of model is good,but measurement process efficiency is lower,more time consuming.In order to improve the accuracy and real-time performance of plant point cloud registration in three-dimensional reconstruction,a three-dimensional point cloud registration method was proposed based on ISS-ICP algorithm.Firstly,two plant point cloud images from different angles were acquired by the TOF(Time of Flight)depth sensor.The key points were detected by ISS feature on the two preprocessing point clouds,and the FPFH(Fast Point Feature Histograms)descriptor was developed to obtain the characteristic vector.Secondly,the initial registration was completed by SAC-IA(Sample Consensus Initial Alignment)algorithm to obtain the initial conversion parameters.Finally,the ICP(Iterative Closest Point)algorithm was used to achieve accurate registration for the two point clouds with better initial positions.The experimental results showed that the average mean square error of two point cloud registrations with the interval angle of no more than 30° was 0.55cm,and the average registration time was 54.2s.Compared with the traditional ICP algorithm,the registration error was reduced by 95.6%.Compared with the initial registration based on SIFT and NARF characteristics,the registration errors was reduced by 86.5%and 69.9%respectively,and the registration time was reduced by 57.3%and 53.9%respectively.In this paper,the point cloud registration algorithm had good accuracy and real-time performance.Three-dimensional point cloud registration is the key problem of the three-dimensional reconstruction,and it has important research value and research difficulty.Due to large amount of calculation of point cloud registration and the complexity of algorithm is higher,so it is higher requirements on the speed and accuracy of computing.The methods of cloud computing and artificial intelligence can be used to improve the processing speed and matching accuracy of registration method.3.For the purpose of improving the performance of disturbances rejection,this paper proposed a composite control scheme based on extended state observer(ESO)and continuous finite time control(FTC)to improve the disturbance rejection property of system.First,an extended state observer is introduced.to estimate the disturbances of system.The estimated value is used in the feed-forward compensation design.Second,a continuous feedback-based finite-time control technique is applied in the feedback design.Then the stability of the controller was analyzed.The experimental results show that,when PMSM operates at 1000rpm,the rate of convergence of proposed method is 0.02ms,compared with conventional PI control and the P+ESO control,the performance improvements are 44.4%and 33.3%,respectively.Nevertheless,it has the less overshoot(1.3%).When the PMSM is running at steady-state of 1000rpm,the load torque TL=2N.m is added suddenly,compared with conventional PI control(0.14s)and the P+ESO control(0.08s),the proposed controller gives a shorter settling time(0.05s),and the performance improvements are 64.3%and 44.7%,respectively.Nevertheless,the proposed method has a lower speed drop(80rpm)than the PI controller(200rpm)and P+ESO(150rpm),the performances of disturbance rejection are impro ved 60%and 46.7,respectively.