A High-similarity shellfish recognition method based on convolutional neural network

The high similarity of shellfish images and unbalanced samples are key factors affecting the accuracy of shellfish recognition.This study proposes a new shellfish recognition method FL_Net based on a Convolutional Neural Network(CNN).We first establish the shellfish image(SI)dataset with 68 species and 93574 images,and then propose a filter pruning and repairing model driven by an output entropy and orthogonality measurement for the recognition of shellfish with high similarity features to improve the feature expression ability of valid information.For the shellfish recognition with unbalanced samples,a hybrid loss function,including regularization term and focus loss term,is employed to reduce the weight of easily classified samples by controlling the shared weight of each sample species to the total loss.The experimental results show that the accuracy of shell-fish recognition of the proposed method is 93.95%,13.68%higher than the benchmark network(VGG16),and the accuracy of shellfish recognition is improved by 0.46%,17.41%,17.36%,4.46%,1.67%,and 1.03%respectively compared with AlexNet,GoogLeNet,ResNet50,SN_Net,MutualNet,and ResNeSt,which are used to verify the efficiency of the proposed method.

High performance vegetable classification from images based on AlexNet deep learning model

Deep learning techniques can automatically learn features from a large number of image data set.Automatic vegetable image classification is the base of many applications.This paper proposed a high performance method for vegetable images classification based on deep learning framework.The AlexNet network model in Caffe was used to train the vegetable image data set.The vegetable image data set was obtained from ImageNet and divided into training data set and test data set.The output function of the AlexNet network adopted the Rectified Linear Units(ReLU)instead of the traditional sigmoid function and the tanh function,which can speed up the training of the deep learning network.The dropout technology was used to improve the generalization of the model.The image data extension method was used to reduce overfitting in the learning process.With AlexNet network model used for training different number of vegetable image data set,the experimental results showed that the classification accuracy decreases as the number of data set decreases.The experimental verification indicated that the accuracy rate of the deep learning method in the test data set reached as high as 92.1%,which was greatly improved compared with BP neural network(78%)and SVM classifier(80.5%)methods.

A hybrid model for dissolved oxygen prediction in aquaculture based on multi-scale features

To increase prediction accuracy of dissolved oxygen(DO)in aquaculture,a hybrid model based on multi-scale features using ensemble empirical mode decomposition(EEMD)is proposed.Firstly,original DO datasets are decomposed by EEMD and we get several components.Secondly,these components are used to reconstruct four terms including high frequency term,intermediate frequency term,low frequency term and trend term.Thirdly,according to the characteristics of high and intermediate frequency terms,which fluctuate violently,the least squares support vector machine(LSSVR)is used to predict the two terms.The fluctuation of low frequency term is gentle and periodic,so it can be modeled by BP neural network with an optimal mind evolutionary computation(MEC-BP).Then,the trend term is predicted using grey model(GM)because it is nearly linear.Finally,the prediction values of DO datasets are calculated by the sum of the forecasting values of all terms.The experimental results demonstrate that our hybrid model outperforms EEMD-ELM(extreme learning machine based on EEMD),EEMD-BP and MEC-BP models based on the mean absolute error(MAE),mean absolute percentage error(MAPE),mean square error(MSE)and root mean square error(RMSE).Our hybrid model is proven to be an effective approach to predict aquaculture DO.

Fast processing of foreign fiber images by image blocking

In the textile industry,it is always the case that cotton products are constitutive of many types of foreign fibers which affect the overall quality of cotton products.As the foundation of the foreign fiber automated inspection,image process exerts a critical impact on the process of foreign fiber identification.This paper presents a new approach for the fast processing of foreign fiber images.This approach includes five main steps,image block,image predecision,image background extraction,image enhancement and segmentation,and image connection.At first,the captured color images were transformed into gray-scale images;followed by the inversion of gray-scale of the transformed images;then the whole image was divided into several blocks.Thereafter,the subsequent step is to judge which image block contains the target foreign fiber image through image pre-decision.Then we segment the image block via OSTU which possibly contains target images after background eradication and image strengthening.Finally,we connect those relevant segmented image blocks to get an intact and clear foreign fiber target image.The experimental result shows that this method of segmentation has the advantage of accuracy and speed over the other segmentation methods.On the other hand,this method also connects the target image that produce fractures therefore getting an intact and clear foreign fiber target image.

Multi-kernel dictionary learning for classifying maize varieties

The automatic classification and identification of maize varieties is one of the important research contents in agriculture.A multi-kernel maize varieties classification approach was proposed in this paper in order to improve the recognition rate of maize varieties.In this approach,four kinds of maize varieties were selected,in each variety 200 grains were selected randomly as the samples,and in each sample 160 grains were taken as the training samples randomly;the characteristics of maize grain were extracted as the typical characteristics to distinguish maize varieties,by which the dictionary required by K-SVD was constructed;for the test samples,the feature-matrixes were extracted by dimension reduction method which were mapped to the high-dimension space by muti-kernel function mapping.The high-dimension characteristic matrixes were trained by K-SVD method and the corresponding feature dictionary was obtained respectively.Finally,the test samples representing were trained and classified by l2,1 minimization sparse coefficient.The experiment results showed that recognition rate was improved obviously through this approach,and the poor-effect to maize variety identification from partial occlusion can be eliminated effectively.

Recognition of abnormal body surface characteristics of oplegnathus punctatus

To identify the abnormal characteristics of the oplegnathus punctatus is great importance to the detection of iridovirus disease in the breeding environment.In this paper,an advanced neural network model to identify the characteristics of the oplegnathus puncta-tus and predict its different periods of suffering from iridovirus disease is proposed based on the establishment of a data set.First of all,a standard format data set of oplegnathus punctatus and an abnormal format date set are established in order to verify the effective-ness of the method in this paper.And then,the feature extraction fusion method is used for preprocessing in terms of the abnormal format data set,which combines the edge fea-tures extracted by the improved multi-template Sobel operator and the color features extracted by the HSV model.Finally,an improved VGG-GoogleNet network recognition model comes into being through the fusion and improvement of the VGG and GoogleNet neural network structure.The experiments results show that the prediction accuracy rate for oplegnathus punctatus suffering from iridovirus disease in the the abnormal format data set and the standard format data set are improved,which reach 98.55%and 69.18%.

A multi-scale features-based method to detect Oplegnathus

It is of great significance to use underwater video and image processing technology to detect and analyze fish behaviors.In this paper,an Oplegnathus image dataset for fish behaviors study by deep learning algorithm is constructed,and the data is captured from two cameras(one above water and another below water);and then an improved Neural Network model based on multi-scale features is proposed for fish behaviors learning auto-matically.To overcome the occlusion and blur problems of the images,the lightweight neu-ral network MobileNet-SSD is improved by adding a dilate convolution,and SE blocks are added to the feature maps at different scales to establish a self-attention mechanism;the Focal Loss function is used to calculate the classification loss and to balance the propor-tion of background and target samples.The results of the experiments show that the aver-age behaviors detection accuracy of our method reach 90.94%and 88.36%in both overwater and underwater datasets.