| China is largest river crab farming country in the world.In the present there is a major problem in aquatic plants cleaning in river crab farming.However,its work efficiency and quality are relatively low.With the further expansion of farming scale and the change of agricultural labor force structure,there is a serious shortage of agricultural labor.Therefore,it is an urgent problem to improve the automation and intelligence of river crab farming.To realize intelligent autonomous navigation of aquatic plants cleaning workboat,the navigation efficiency and positioning accuracy are the key to improve the efficiency and quality of river crab farming.In this study,the aquatic plants cleaning workboat is taken as the research object,and improving the work efficiency and position accuracy of the workboat is taken as research goal.We have carried out related research around water surface aquatic plants image information processing and navigation filter positioning technology.It mainly includes image acquisition and preprocessing,image segmentation and navigation line fitting,filter positioning technology,workboat navigation control system and filter positioning experiment,etc.In this paper,the aquatic plants image denoising and segmentation algorithm,optimizing unscented kalman filter(UKF)?particle filter(PF)technology are described for details.The main contents of this study are as follows:(1)Firstly,the image of aquatic plants on the surface of the crab pond is collected.According to the collected aquatic plants image environment,the physical method of loading the polarizing plate is considered to reduce the influence of the diffuse reflection light on the water surface of the crab pond.Then,the aquatic plants image is analyzed by the difference image method to obtain the image noise model.By combining the advantages of wavelet transform(WT),fuzzy theory and independent component analysis(ICA)in image denoising,a WT and ICA combined denoising algorithm(WT-ICA),and an improved wavelet threshold denoising algorithm(Fuzzy-WT)are proposed.Through simulation experiments,the mean values of PSRN values of aquatic plants image after WT-ICA and Fuzzy-WT denoising methods are increased by 12.27% and 12.17%,respectively.In terms of computational efficiency,both proposed algorithms have improved,and Fuzzy-WT method is slightly better than WT-ICA method.(2)Through image analysis,the gray level of crab pond aquatic plants image is not uniform.Combining with pulse coupled neural net-work(PCNN)image segmentation and particle swarm optimization(PSO)algorithm for spatial optimization,an aquatic plants image segmentation algorithm based on PSO optimized PCNN(PSO-PCNN)is proposed.The new algorithm mainly optimizes the connection coefficient,amplification coefficient of connection weights and iterative decay time coefficient through PSO,and formulates the PSO-PCNN parameter optimization strategy.According to the experiment result,compared with the other three methods,it can be obtained that the error segmentation rate ?,false negative rate FNR and false positive rate FPR are all decreased to different degrees,and the segmentation accuracy is improved.In this study,the segmented images of aquatic plants are divided into cluster distribution and linear distribution for navigation line fitting,and visual navigation assistant lines are obtained.(3)In view of the low navigation accuracy of the workboat,the navigation model of the workboat is first established.The UKF method is used to filter and locate the workboat,and the PSO global optimization intelligent algorithm is introduced.The immune algorithm(IA)is used to improve the diversity of the PSO to avoid the algorithm falling into the local extremum.The immune particle swarm optimization UKF(IPSO-UKF)navigation filter positioning method is proposed.Through the simulation experiment of the workboat navigation model,the root mean square error(RMSE)of the east and north directions of the proposed method can decrease by 46.09% and 71.51% respectively;compared with GPS navigation method,and decrease by 23.92% and 58.26% respectively compared with integrated navigation method.In addition,combining with the positioning advantage of PF technology in the nonlinear non-Gaussian system,and the general regression neural network(GRNN)is introduced to adjust the weight of particles.The fruit fly optimization algorithm(FOA)is used to adjust the smoothing factor of mode layer transfer function for improving the performance of GRNN.A new FOA-GRNN-PF workboat navigation filter positioning method is proposed.Through the simulation experiment,FOA-GRNN-PF has advantages in both calculation accuracy and running time compared with PF and GRNN-PF,and can achieve better positioning accuracy and higher computational efficiency under fewer particle numbers.(4)In this study,the workboat experimental platform and control system of the workboat are designed.The workboat structure and hardware equipment are included in the experimental platform,and the structure of the aquatic plants cleaning workboat is described in the structural design part.The workboat hardware equipment mainly includes the control cabinet,the shipborne mobile station equipment,the image acquisition equipment,and the ground base station equipment.In the aspect of the workboat control system,the design of the overall control method of the workboat is described.The visual aid navigation system and the straight-forward,turning and route switching control strategies of the work ship are analyzed.The heading and speed fuzzy PID control methods of the workboat are given.(5)Combined with the relevant methods proposed in this study,the filter positioning experiments of aquatic plants cleaning and automatic feeding are carried out respectively.From the experimental results,it can be found that when the workboat cleans the clusters or strips aquatic plants by the visual aided navigation model,IPSO-UKF and FOA-GRNN-PF method can improve the navigation efficiency and positioning accuracy.When the workboat is in GPS navigation model,the deviation error based on the real-time interpolation IPSO-UKF and FOA-GRNN-PF method can be greatly reduced.Through the above study,a series of image processing algorithms and filter positioning techniques are proposed.The navigation efficiency and positioning accuracy can be greatly improved.The automatic operation level can be significantly improved.The labor intensity has been reduced,the scale development of China's river crab breeding industry can be promoted,which can also provide a certain theoretical basis and technical support for the transformation and upgrading of China's agricultural machinery equipment. |
PDF Full Text Request