Research On Image Recognition Of Pepper Pests And Diseases Based On Image Recognitio

Crop diseases and insect pests are a major problem in agricultural production.Due to their diversity and complexity,they are easy to break out under many specific conditions,resulting in crop reduction or even extinction.With the continuous development and progress of society,the total population of the world continues to increase,the agricultural land area continues to shrink,how to ensure the productivity of crops,and meet the market demand of the problem is increasingly prominent,and pest management is a key aspect of this research.Agricultural production is seriously restricted by crop diseases and pests,because crop diseases and pests have a wide variety,wide distribution area and high concentration,which is very easy to cause a large amount of crop yield reduction.At the same time,due to the slow speed and low accuracy of traditional artificial identification of pests and diseases,it is easy to lead to the abuse of pesticides,great damage to the natural environment,and threaten the harmonious symbiosis between man and nature.At present,with the emergence and development of the concept of precision agriculture and intelligent agriculture,the use of information technology to support agricultural production,achieve intelligent identification of crop diseases and pests,and then reduce unnecessary pesticide spraying,which is of great significance to protect the balance of ecosystem,ensure the safe production of crops,improve crop quality.In this paper,a convolutional neural network based image classification of pepper pests is proposed to identify pepper pests and diseases.Pepper is indispensable in People’s Daily diet and is a very popular food.Due to the continuous improvement of human living conditions,as well as the rapid growth of the global market demand for condiment business and vegetables,the demand for pepper and mass production and processing of pepper is growing.But in the process of production and processing,the quality of pepper is controlled mainly by manual labor.Artificial control of pepper quality has congenital defects such as low efficiency,large labor input,serious lack of effectiveness,and ultimately leads to the continuous low level of pepper quality.In order to meet the needs of practical application,the relevant image recognition technologies in earlier years used manual processing or shallow neural network to extract the features of insect and disease images,which resulted in low accuracy of image extraction,time consuming and labor cost.Therefore,a more effective recognition and classification algorithm is needed to detect pepper pests and diseases.With the advent of high-performance,high-capacity picture processors(Gpus)and open-source database technology,the concept of deep learning is moving from theory to practice.Convolutional neural network(CNN)is the most famous deep learning algorithm for image classification.In the case of large data sets,convolutional neural networks are superior to shallow networks and the human eye in recognizing and classifying results.In addition,due to its unique characteristics,the image can be automatically processed in the early stage of model operation,which also reduces the pretreatment work of the image before putting into the model.Therefore,this paper conducted an in-depth study on the chili leaf disease and insect pests picture data set(a total of 11 categories,3650 pictures)provided by iffy Global AI Challenge open Platform of USTCM.Convolutional neural network was adopted in pepper defect recognition and classification,and the accuracy of pepper disease and insect pest defect recognition was improved by improving the model.The main work content of this paper is as follows:1.In addition to reviewing the basic overview of pepper pest identification and the current applications of convolutional neural networks in image recognition and classification,a brief introduction was also given to the history and research significance of this topic.2.This article introduces the theoretical foundation of convolutional neural networks and the deep learning framework Py Torch.3.For the task of crop pest classification,improved classification models based on SSD,VGG-16,Mobile Net,etc.were used to classify pests into 11 categories.Experiments have shown that this model can effectively identify and classify crop pests and diseases,with an overall recognition accuracy of 74.77%,and has good application prospects.4.Use an improved convolutional neural network to recognize and classify images of pepper pests and diseases.We established a chili pepper color RGB image dataset and a chili pepper gray image dataset for classification.Finally,we compared the recognition results of two types of convolutional neural networks based on grayscale images and RGB color image training sets for chili pepper images.The results indicate that the convolutional neural network based on grayscale image datasets has a high accuracy in image recognition and classification of pepper diseases and pests.Finally,compared to traditional methods,the theoretical method for identifying crop pests and diseases in this article has significant changes and improvements in key data performance,and more effectively solves the problems of pests and diseases that may occur in actual production.