Design And Implementation Of Green Perrer Recognition System

The population aging is serious in China,especially the shortage of labor force in rural areas,which hinders the development of agriculture.In this paper,through the identification of pepper fruit image processing and BP neural network technology,aims to develop a product capable of picking robot in the practical application of the field of agricultural production of fruits and vegetables,to change the current agricultural production in China for picking fruits and vegetables mainly rely on manual completion of the situation,in order to deal with the labor force,the rural population aging in our country is very serious the lack of conditions.At the same time,in the face of the automation degree of international agricultural production continues to increase,the international agricultural products prices continue to decrease,the research and development of a high performance of the picking system,our country is helpful for reducing the cost of agricultural products,improve the international competitiveness of China's agricultural products,allowing farmers to grow fruits and vegetables more enthusiasm,has the practical significance to the healthy development of the agriculture of our country.The green pepper recognition system is part of the robot.It can be applied both individually and in a robot system.It provides a technical guarantee for automatic picking of robot.The main contents of this paper are as follows:1.Analysis of color characteristics of pepper fruit and background from the library material,use of color knowledge were extracted from the range of fruit color and the background color and characteristics,find the ideal parameters of the original gray image processing,combined with the gray level distribution of the fruit and the background,calculate the threshold of the fruit and the background,the final image segmentation two value image.2.The use of image processing technology,pre processing of the image,through denoising,smoothing,edge detection,corrosion expansion,complete separation of pepper fruit and background image.In the experiment,four kinds of filtering methods and three kinds of edge detection methods were compared in detail,and the median filter algorithm and Canny operator edge detection algorithm were selected as the final design plan.3.Through analyzing the features of pepper fruit,extract feature five normalized values as the input of BP neural network.In the experiment of extraction of five with similar features of pepper fruit value,to speed up the BP neural network recognition rate.4.In this paper,the BP neural network is used to identify the fruit of green pepper.The default parameters are used to train and identify the five normalized eigenvalues of green pepper,and the parameters of the neural network are constantly changed.A group of parameter groups with fast convergence and high recognition rate are determined,and a stable BP neural network is established.5.Finally,the results were statistically analyzed.By studying the design of the color characteristic of the image,and then use the preliminary separation of the fruit and the background of the image histogram to two threshold estimation,then use denoising,smoothing,edge detection,image corrosion expansion and separation target recognition technology,finally extracted five representative features of the target shape.As the input data of BP neural network for intelligent recognition.The system designed in the experiment has realized the recognition of green pepper fruit,the recognition rate is above 90%,the recognition speed is less than 1 second,and it can meet the actual production demand.