Agricultural Image Processing System Design And Algorithm Research Based On Cloud Computing

Image processing has been widely used in detection of agricultural products, diagnosis and identification on diseases, pests and weeds of crops, precision spraying and fruit harvesting, etc; it has achieved lots of research results in agricultural field and promotes the development of agricultural modernization. But these achievements haven’t been accepted by users because of the complexity and the high cost of image processing system, the inconvenience of use and maintenance, and then they don’t give full play in economic benefit and social benefit. So it is a problem to realize the real-time processing of image and make low cost that could be accepted by users.Agricultural image cloud is proposed by combining the cloud computing and agricultural image processing and then the architecture of agricultural image cloud is designed. The cloud platform of agricultural image cloud is divided into three levels: control layer is the core, transition layer makes the choice and algorithm layer executes the functions. For the purpose of designing the control layer, the parametric design method of agricultural image processing system is proposed, and then the system can set different parameters for different objects. The model of top-level controller of the system is built, and the model can schedule algorithm module automatically according to the parameters of different objects. At last, image segmentation, image de-noising and image sharpening are researched in the paper. The optimization and integration of these algorithms are researched for improving the quality of image processing.The main research contents are as follows:(1) Agricultural image cloud is proposed for improving the real-time of the system, the extensiveness of the system functions, the convenience of the operation, the adaptation of the working environment and the cheapness of the using cost, and then the architecture of agricultural image cloud is designed. The cloud platform, as the core of the agricultural image cloud, is divided into three levels:control layer, transition layer and algorithm layer. The control layer schedules the algorithm transition modules which the service needs in the transition layer according to the service that selected by users; the algorithm transition module in the transition layer should judge whether it is scheduled and select the idle algorithm modules in the algorithm group which corresponds with the algorithm transition module to executive the function when the execution conditions are favorable. There is also a service parameter library used to store the parameters of various objects.(2) For the purpose of designing the control layer in the cloud platform of agricultural image cloud, the parametric design method of agricultural image processing system is proposed. Users can select the algorithms which are needed and set the parameters into the system, and then the system will execute the algorithms according to the parameters to complete the functions that the users requested. The design method improves the flexibility of image processing system and makes the system more widely used; it is applicable to design the large-scale, distributed parallel computing system, such as cloud platform, and it is also can be used to design independent parametric agricultural image processing system.(3) In order to solve the modeling problem of agricultural image processing system which has a parallel structure, a Petri nets modeling method of algorithm image processing system is proposed and an algorithm call model based on parameters is constructed. The model of top-level controller of the agricultural image processing system based on Petri net is built, it is transferred to PNML by using SnoopylOPT software, and then the VHDL is got through the PNML2VHDL software; at last, the VHDL is simulated in QuartusII and the simulation result shows that the model is correct.(4) Color image cone segment algorithm based on G-component is proposed for getting better segment results of cotton leaves and other green crops. The paper also proposed color image frustum segment algorithm of any color, reduced the expression and verified it. For the purpose of limiting the range of pixel gray value of the target, the color image frustum segment algorithm based on G-component is proposed by fusing the color image cone segment algorithm based on G-component and the thresholding method. The new algorithm makes a cone, the apex of the cone is (0,0,0), the center axis is G-axis and the radius is r. The pixels which in the cone at the same time the pixel gray value is in the range are the target pixels. The unknown quantities are determined by histogram method. At last, the new algorithm is compared with color difference method and thresholding method. The result shows that the pixel error number is the least and the pixel error rate is the smallest by using the new algorithm.(5) Fusion algorithm research of image de-noising based on median filter and mask de-noising method. The de-noising effect is good but the image is fuzzy by using median filter5X5; and the de-noising effect is not good but the image is clear by using mask de-noising method. The fusion algorithm is got by fusing the two methods through assigning different weights to the results of the two methods. The result shows that the bigger the k is, the better the de-noising effect is and the fuzzier the image is; the de-noising effect is better and at the same time the image is clearer by using the fusion algorithm which the k is0.6.