Based On Improved Wolf Pack Algorithm To Optimize Bayesian Network Structure Learning In The Field Of Pest Identification

The most basic development of the national economy is agriculture.The development of agriculture is closely related to our lives.Agricultural pests and diseases are one of the main reasons for the development of agriculture.The main methods for controlling pests are pesticides,natural enemies and organisms.The most common ones are pesticide control methods.However,the blind use of pesticides,sometimes,can not only effectively control pests,but will destroy the soil,burn seedlings,pollute water sources,leading to a decline in crop yields,and even pose a great threat to human health.However,the types of pests are very complicated.The traditional method of pest identification relies only on the human eye.It is often objective and one-sided to classify pests according to the characteristics of the pests they see,and it is very labor-intensive and time-consuming.Therefore,the classification and identification of pests has become an urgent issue.Only by accurately classifying can crop diseases be controlled.We propose a pest image recognition method based on improved Wolf Pack Algorithm to optimize Bayesian network.First,we use the GrabCut algorithm to automatically segment the pest target region,then perform grayscale processing on it,save it to the data file,and form the data set.For training and testing of models.Secondly,the pre-trained convolutional neural network is used to extract the image features on the training set and the test set and input into the Bayesian network.Then the traditional Wolf Pack Algorithm is improved as a search algorithm,Bayesian Information Criterion(BIC)as a scoring function to learn the structure of Bayesian network.Then use the Maximum Likelihood(ML)algorithm to learn the parameters of the Bayesian network to form a Bayesian classifier.The main research work of this paper is as follows:1.Proposed an improved binary wolf colony algorithm(I-BWPA)The Wolf Pack Algorithm is improved.The mutation operator is added to the wolf's wandering behavior.The approximation operator is added to the summoning behavior,and the interaction operator is added to the siege behavior.And in the update step of the wolves,it is proposed to use the chaotic map to generate a new artificial wolf to replace the artificial wolves.2.Bayesian Network Construction algorithm based on improved binary wolf group algorithm.(Bayesian Network Construction algorithm using I-BWPA,BNC-I-BWPA)The Bayesian network structure can be represented by an adjacency matrix whose corresponding structure matrix is coded as{x11,x12,...,x1 n,x21,x22,...,x2 n,...,xn1,xn2,...,xnn}.Then,using the improved binary wolf group algorithm as the search algorithm,BIC is used as the scoring function to find the optimal Bayesian network structure.3.Combining convolutional neural networks and Bayesian networks for identification and processing of pest images.The pre-trained convolutional neural network is used to extract the features of the training set and the test set,input the feature attributes and classification extracted on the training set,and use the BNC-I-BWPA to learn the Bayesian network structure,and then use ML.The Bayesian network parameters are learned to form a Bayesian network that best matches the input data set as a Bayesian classifier.The Bayesian classifier is tested by inputting the extracted feature attributes and classifications into the Bayesian classifier.