International Journal of Pattern Recognition and Artificial Intelligence

Grape is an important cash crop that is susceptible to diseases when growing, resulting in lower yield and quality. In recent years, transformers have achieved excellent performance in a variety of natural language processing and image recognition tasks through the self-attention mechanism. Therefore, this paper proposes a grape leaf disease recognition model named Dense Convolutional Transformer (DensCT). The compact convolutional transformer (CCT) is used as the backbone in this model, which improves the convolutional module of the original model by introducing densely connected modules, enhancing the transfer and reuse of features between networks. This also modifies the single-scale feature extraction method of the original model to multi-scale, which improves the feature extraction performance. Finally, the model was trained on two small-scale datasets from scratch, and the recognition accuracy of the final model on the test sets reached 89.19% and 93.92%. Compared with CCT, DenseNet121, ResNet50, MobileNetV3 and ViT, the recognition accuracy improved by 4.73%, 3.38%, 10.81%, 0.68% and 18.24% on the first dataset and 6.08%, 5.41%, 1.35%, 3.38% and 12.84% on the second dataset. The experimental results show that the proposed model can effectively identify grape leaf diseases, which can provide a reference for building disease leaf recognition models on small-scale datasets.

“Crypto ransomware is defined as malware that blocks the user file’s access by encrypting them and demands them with a ransom for obtaining the decryption key”. This causes some major threats in many of the companies. Therefore, the detection of ransomware is needful for reducing the workloads of the analysts and also for finding the variations in unknown samples. The adopted scheme encompasses 3 phases: (i) feature extraction, (ii) feature selection and (iii) classification. Initially, the sequential frequent patterns are extracted using the apriori framework. However, the major challenge in this extracted sequential pattern is the curse of dimensionality. To overcome this, the selection of optimal features is very important, which is done as the second stage. In this, the optimization concept is evolved for the optimal selection of these extracted sequential patterns. Furthermore, the optimal patterns are given for classification, where DBN is deployed. Particularly, for the selection of the optimal sequential pattern, this work proposes a new crow search with adaptive awareness probability (CS-AAP) model. In the end, analysis is performed to authorize the supremacy of the developed scheme.

Ferrography analysis is one of main means to identify wear state of mechanical equipment, and its key is the intelligent recognition of wear debris ferrographic images. Ferrographic image acquisition is a complex and time-consuming work, so the direct deep learning cannot been carried out for the small tested samples. A virtual ferrographic image dataset is prepared firstly and then two-level transfer learning scheme is proposed to improve the identification rate of the tested samples based on the deep learning model trained by the virtual samples. A combined network of YOLOv3 and DarkNet53 is constructed, and the application effect of model is improved by two-level transfer learning of virtual dataset to open dataset and then open dataset to tested dataset, and the model errors before and after twice transfer learning are analyzed. The average identification accuracy of the model in the validation dataset is 86.1%, which is 44.5% higher than that without two-level transfer learning, and the average recall reaches 95.8%. The experimental results prove the proposed method have a high identification rate for the tested ferrographic images of an actual gearbox.

Type II Diabetes Mellitus (Type II DM) is a chronic condition that has detrimental effect on vital organs if left untreated, necessitating early diagnosis and treatment. Iridology, a subset of Complementary and Alternative Medicine (CAM), has the potential to serve as a tool for noninvasive early diagnosis of Type II DM. Iridology involves analyzing the characteristics of iris such as color and pattern for detection of organ and system defects. Deep learning algorithm is one of the promising methods in diagnosing various health-related issues. In this study, we have demonstrated the efficiency of iridology in diagnosing Type II DM using deep learning algorithms. Near Infra-Red images of iris were captured using iris scanner from 178 voluntary subjects belonging to two categories namely, Type II DM (95 subjects) and nondiabetic or healthy category (83 subjects). We have developed an algorithm using Fully Convolutional Neural network for effective iris segmentation. Normalized iris images were used to crop out our region of interest, pancreas, based on the iridology chart. Classification networks such as AlexNet, VGG-16, and ResNet-50 were used to classify Type II DM versus healthy category. Our proposed model for iris segmentation achieved an accuracy and sensitivity of 0.99, specificity and F-Score of 0.98, and a precision of 0.97. Results obtained using AlexNet classifier exhibits better classification accuracy of 95.85% for Zero-padding based resized image. The classifier yielded a sensitivity, specificity, and precision of 95.80%, 95.85%, and 96.11%, respectively. Our study results establish the efficacy and emphasize the importance of the proposed algorithm for diagnosing Type II DM.

The traditional embroidery identification technology cannot present the target image of the art network in a more comprehensive and three-dimensional manner. A research on the segmentation and synthesis of embroidery art images based on deep learning convolutional neural network is proposed. Based on the semantic image segmentation technology of deep learning, this paper analyzes the embroidery semantic image segmentation technology, obtains the information of image technology, analyzes the embroidery rendering technology of convolutional neural networks, and puts forward the embroidery rendering algorithm. In order to verify the effectiveness of the algorithm, a simulation test experiment was carried out on the target content image and the embroidery art network image. The test results show that compared with the traditional method, this method has more specific and flexible image generation, stronger three-dimensional sense, closer to the real art embroidery network, and the direction of its needlework is also more hierarchical.

Driver misbehavior is one of the major traffic safety hazards as car ownership increases year by year. So, it is important to have driver fatigue detection and behavior recognition. Initially, given the fatigue detection problem that the images captured by the visible light camera cannot capture the eyes of a driver wearing sunglasses or eye glasses. As a solution, this paper introduces a DCT-HSV preprocessing algorithm for infrared images, which is believed to enhance the target characteristics of infrared images. The paper also introduces a more efficient lightweight SSD detection model, which achieves a better balance in terms of model size and detection performance. It also shows a better performance in self-built datasets and basic vehicle operation datasets. Secondly, aiming at the problems of high complexity and poor accuracy of the existing driving behavior detection model, this paper designs a driver abnormal behavior discrimination model based on the comparison twin, which has good performance on the Kaggle public dataset. The relevant experimental detection results show that the method constructed in this paper has high detection accuracy, low warning delay, and good practical use value.

Digital recognitions are playing a vital function in the current era of technological advancements. Hence, they offer more possible ways of performing handwritten character recognition (HCR). Generally, recognizing the Tamil handwritten texts is highly complicated, in comparison to the Western scripts. Nevertheless, many researchers have presented several real-time approaches to achieve Tamil character recognition (TCR) in offline mode. This paper introduces a new handwritten TCR (HTCR) approach with two phases: (1) pre-processing and (2) classification. Primarily, the scanned document in the Tamil language is pre-processed via the steps like RGB to grayscale conversion, binarization with thresholding, image complementation, application of morphological operations and linearization. The pre-processed images are then classified using an optimized convolutional neural network (CNN) model. Further, the fully connected layer (FCL) and the weights are tuned optimally via a new sea lion with self-adaptiveness (SL-SA) algorithm. Lastly, the adopted model is evaluated using various measures to prove its supremacy over the existing schemes.

Adversarial attack is a serious threat to neural network-based natural language processing applications. Adversarial attack uses tiny well-crafted perturbations to mislead neural networks. While existing adversarial text attacks can achieve good attack effects, they still do not guarantee efficiency and robustness. The adversarial text attacks are more efficient if they use less perturbation to achieve a higher attack success rate. The attacks are more robust if they can achieve a higher success rate when defense strategies are applied. To improve the efficiency and robustness of the adversarial attack, we propose SMAL: Saliency Map Attack with Levenshtein-similarity. The proposed attack consists of two parts: (1) The saliency map measures the perturbation priority of each word. It considers not only the influence of each word on the classification result but also how to maintain the misled classification result to improve the robustness of the attack. (2) Levenshtein-similarity network embeds words into edit distance space. When perturbing sentences, some words are replaced by substitutions with less edit distance. This can reduce the amount of modification, which improves the efficiency of the attack. Since the words are embedded in edit distance space rather than semantic space, the semantic-based defense is not effective for this attack, which improves the robustness. The experiments show that SMAL achieves a higher attack success rate with fewer perturbations. Also, the proposed attack is better when attacking a classifier defended by adversarial training.

Automatic Facial Expression Recognition (FER) has become essential today as it has many applications in real time such as animation, driver mood detection, lie detection, and clinical psychology. The effectiveness of FER systems mainly depends on the extracted features. For extracting distinctive features with low dimensions, a new local texture-based image descriptor named Dimensionality Reduced Chess Pattern (DRCP) is proposed for recognizing facial expressions in a person independent scenario. DRCP, an improvement over Chess Pattern (CP), is mainly proposed for effectively reducing the feature vector length of CP. For feature extraction, DRCP also considers the movements of chessmen in a $5\times 5$ neighborhood, as like CP. As a part of feature extraction through DRCP, apart from the center pixel, the remaining 24 pixels are arranged into four groups in such a manner that each group contains the pixels corresponding to three chessmen. From each group, one feature is extracted and thus corresponding to four groups, four features are extracted in a $5\times 5$ neighborhood. The extracted features are fed into multi-class Support Vector Machine (SVM) for expression recognition. The experiments are performed on five “in the lab” datasets (MUG, TFEID, JAFFE, CK$+$ and KDEF) and on two “in the wild” datasets (RAF and SFEW) in person independent setup to simulate a real world scenario.

Maize is one of the main crops in Shangluo. The maize yield and quality are adversely affected by leaf spot and rust. In order to realize the early prevention of maize leaf spot and rust and avoid problems of environmental pollution induced by the conventional chemical reagents, computer vision technology was proposed for disease detection in this research. The algorithm of $K$-means was used to process the image samples obtained from the test field. The healthy area, disease area and background area were separated. Based on the healthy area and disease area, ten parameters were extracted, including four parameters of color characteristic, four parameters of texture feature and two parameters of shape feature, which were taken as the classification criteria in the classification training by KNN algorithm. In total, 200 test image samples (100 samples of leaf spot and 100 samples of maize rust) were sent into the training model for disease identification. The results showed that the algorithm proposed can efficiently and nondestructively identify maize leaf spot and rust based on image segmentation and multi-feature fusion. The method was convenient and environmentally friendly. It can provide technical support for plant protection and can supply research ideas of precise control of crop diseases.

Pests and diseases of tomato fruits are the foremost factors leading to the reduction in produce quantity. Effective and prompt differentiation of these lesion symptoms is essential for agricultural production. Due to the lack of a global receptive field, commonly used convolution neural network models restrict the capability on tomato images under complicated backgrounds. In this paper, we present an efficient and robust method for field diagnosis of tomato fruits based on the context data fusion and capsule network. First, in order to make better the point of inadequate images, the Cycle-GAN algorithm is used to realize the augmentation of the diseases dataset. To improve the quality of the generated images, the feature reconstruction loss function is proposed for the Cycle-GAN method, named Fruit-GAN. The proposed Fruit-GAN learns features of healthy and diseased tomatoes, respectively, and forms lesions based on healthy tomato fruit images. Adding lesion tomato images produced on Fruit-GAN to training sets improves the diagnostic capability compared to classical data augmentation techniques. Subsequently, context fusion studies were used to construct intermediate feature maps from different layers into the fruit diseases recognition compact representation to efficiently leverage the feature reuse mechanism. Finally, the fusion model’s enlightened multi-scale capsule network is integrated to form deep features and determine the disease diagnosis. After analyzing the hyper-parameters, the integral technique provides 96.16% accuracy. This research of tomato diseases diagnosis demonstrates the context capsule network model is efficient and establishes a foundation for tomato fruits determination in natural growing environment.

Meshless Shepard interpolation method (MSIM) is used to simulate the destruction process within the arch reinforcement area of tunnel lining. In this method, the shape functions are formed by the partition of unity and the finite cover technology, which is not affected by discontinuous domains, with delta property at any desired node, and applies boundary conditions in an easy and correct way. The MSIM method combines the advantages of both the conventional meshless method and the numerical manifold method. The finite cover technology enables the shape functions to not be affected by the discontinuities in the solution domain, which overcomes the difficulty resulting from the conventional meshless method. The finite covers and the partition of unity functions are formed using the influence domains of a series of nodes, which removes the obstacle from conventional numerical manifold method and has simpler formation of finite covers than numerical manifold method. Virtual crack closure technique is used to calculate the intensity factor of crack-tip stress. The results of progressive destruction process simulation on the cracking patterns within the arch reinforcement area indicate that the method is suitable for tracking the crack propagation in complex stress conditions.

The blind spots brought by a car’s A-pillar are main reasons of most of accidents. In this work, a vision system that focuses on eliminating A-pillar blind spot of a car without any affection of driver’s operation is investigated. The driver’s facial features are captured by a binocular vision system that is mounted on A-pillar, head poses and gaze line directions are reconstructed. The generated blind spot by A-pillar is then simultaneously calculated according to the position of driver’s gaze. A field-of-view of the blind spots is displayed in a screen system mounted on the A-pillar. The screen conjointly with front and side windows thus provides a full view field for the driver, which can effectively reduce the occurrence of accidents.

EEG has been proved to be an effective tool for researchers’ cognition and mental workload by detecting the changes of brain activity potential. The mental workload of pilots in aviation flight is closely related to the characteristics of flight tasks. The previous methods have problems such as lack of objectivity, low EEG analysis ability and lack of real-time analysis ability. In order to solve these problems, this paper proposes a multi-dimensional data fusion brain workload calculation method based on flight effect evaluation, which integrates vision, operational behavior and visual gaze, and classifies and analyzes them in combination with EEG data. This method evaluates the mental workload of pilots from three aspects: visual gaze behavior, control behavior and flight effect in the simulated flight experimental environment, and realizes a more objective mental workload analysis. Then, the synchronously collected EEG data are segmented and sampled to form a dataset, and an LSTM neural network model integrating attention mechanism is established, in which the attention mechanism is used to improve the feature processing ability of the network model for the classification of complex EEG data. After machine learning training, the final model can achieve 94% detection accuracy for 2-s EEG data, and has the ability of real-time analysis in the application environment. Compared with the previous similar LSTM model, the accuracy is improved by 6%, which also shows the effectiveness of the model.