Abstract:Plant protection machine is widely used in agricultural production, but its relatively long boom with bolted joints may have a large response under various external loads, resulting in structural damage. To serve the control of the response under complicated circumstances and protect the machine, the global dynamics of the boom is studied from both modal and dynamic analysis. The finite element model of the boom structure established by ABAQUS are reduced by the Component Modal Synthesis method to significantly improve the computational efficiency, by which the motion equation of the system is obtained. The modal analysis is then carried out. The first four order modes are solved and the amplitude-dependent frequency and damping ratio of the second order mode are analyzed by a method of Quasi-Static Modal Analysis. Finally, the responses of the whole structure are calculated under static load and dynamic load, indicating that the dynamic load has a larger potential hazard. The parametric characteristic of the response is analyzed by changing a variety of parameters in turn.
Abstract:Object detection and tracking of dairy cows based on machine vision is an important foundation for behavioral analysis. The accuracy of detection is directly affected by the accuracy of the detector. Compared with laboratory conditions, object detection in natural breeding sites is often affected by changes in illumination and occlusions between objects. Because appearance similarity, it is greater difficult for cow detection and tracking with the changes in natural light. Therefore, this paper introduced an improved YOLO v4 algorithm combined with the channel pruning and a tracking algorithm to achieve object tracking which combined the location and appearance information between different frames. Firstly, a video collection platform for dairy cows was established. About 2320 images were from the surveillance video, included cow objects of three different postures, under different light condition and with different degrees of occlusion. To achieve detection of multi-cow, YOLO v4 model based on CSPDarkNet53 was built and pruned. On the test set of 420 images including single and multiple cows during the whole day, the average precision reached up to 98.01%. The detection speed reached up to 41.25fps after pruning. To address the temporal association across video frames, a tracking algorithm based on Deep Sort was proposed. The result showed that the object still maintained the original mark number after being occluded. The algorithm in this paper exhibited high precision for multi-object tracking in complex scenes and effectively reduced ID switches. The proposed method can provide a theoretical reference for multi-object detection and tracking for dairy cow in large-scale precision farming.
Abstract:Multi-Channel, Semantic Segmentation, Thyroid Nodules\nAbstract: Semantic segmentation of medical images is a fundamental task and a basic step in medical image processing. However, the segmentation of medical images by artificial intelligence algorithms has not yet reached a professional level. Thesegmentation results are often not impressive. Regarding the issue above, the Multi-Channel Atrous Space Convolution Pooling Pyramid module is proposed in this paper. This module extracts feature information from channels and combines the convolution results of different convolution kernels for each pixel point on the image. This module establishes links between different channels for the model, introducing information between different channels to provide higher dimensional information and improves the ability of the model to extract minute features efficiently. 3967 images of thyroid nodules are used as the dataset. The optimal structure is chosen by extensive experiments under the dataset of this paper. The final result of segmentation is Mean Intersection over Union of 84.8%. The average error rate is reduced by 13.6% compared to the state-of-the-art model. The feature extraction capability of the model and the accuracy of the model for image segmentation are enhanced. These modules can be easily applied to various computer vision tasks and improve the performance of the model. Our code [https://github.com/dia123blo/MCASPP_Deeplabv3-]
Abstract:The demand for intelligent management of animal husbandry is becoming more and more obvious. It has become a new research focus on identifying individual animals by means of technology. In this paper, to address the low recognition accuracy caused by the difficulty in collecting cattle iris image, an improved feature extraction network SqueezeNet-MS was proposed. The Mish function was used to replace the ReLU function in the original Fire module of the SqueezeNet network. In this way, the robustness of the network was enhanced and the speed of feature extraction was accelerated. The Stochastic Pooling was utilized to improve the generalization ability of the network. In this paper, 80 cattle were randomly selected with 150 photographs of each of the left and right eyes. The proposed method was compared with the existing network. The experimental results showed that the proposed method achieves a better recognition effect with a small number of parameters.
Abstract:Background: It is observed that the crude suicide rate has increased approximately 2.5 times in the years between 1975 and 2019 in Turkey, with suicides of an unknown cause accounting for a significant part among completed suicides. However, very limited information is available about completed suicide of an unknown cause. Objective: This study was conducted to compare completed suicides of an unknown cause by geographical regions in Turkey. Methods: In this study, suicide statistics of the Turkish Statistical Institute (TURKSTAT) for the years between 1975 and 2019 were analyzed retrospectively. The distribution of completed suicides of an unknown cause was examined by geographical regions for the specified period of 45 years. Data were evaluated by numbers, percentages, and mean and standard deviation. RMANOVA, Duncanï¿½s multiple comparison test, and Bonferroniï¿½s multiple comparison test were used.Results: The review of the total number of suicide cases per 100,000 population in seven geographical regions of Turkey in a period of 45 years reveals that suicides were most common in the Aegean region followed by the Eastern Anatolia, Mediterranean, Central Anatolia, Black Sea, Marmara, and Southeastern Anatolia regions, respectively. However, the examination of suicides of an unknown cause reveals that the rates were the highest in the Eastern Anatolia region, which was followed by the Aegean, Black Sea, Mediterranean, Southeastern Anatolia, and Marmara regions, respectively in the decreasing order of frequency.Conclusion: This study has shown that suicides of an unknown cause constituted the most significant category among completed suicides by the cause in Turkey in the years between 1975 and 2019. It has been found out that suicides of an unknown cause in Turkey were most common in the Eastern Anatolia region and least common in the Marmara region.
Abstract:The coronavirus disease of 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) recently has become fatal pandemic in the past hundred years. The ACE2 is a major entry receptor for causative coronavirus SARS-CoV-2 and expressed in various extrapulmonary tissues, hence, direct viral tissue damage has been proposed a possible mechanism of injury. Since pandemic started, various already existed drugs were approved as first-line treatment for COVID-19 patients. Although, several already existing drugs, including tocilizumab, remdesivir, hydroxychloroquine, chloroquine, baricitinib, and arbidol have shown great potential for COVID-19 management by inhibiting SARS-CoV-2, however, various studies have discouraged their use in the treatment of COVID-19 patients. Here, we review the pathophysiology and transmission of COVID-19, multi-organ dysfunction caused by SARS-CoV-2, efficacy and safety of major repurposed drugs that have been initially proposed or already being investigated in clinical trials to treat COVID-19 patients.
Abstract:One of the most important qualities of pedagogue-musician is his ability to organize interaction with children, so to communicate with them, to study their abilities, to manage their activities. The actuality of problem of teacher\'s understanding of pupil\'s personality is growing due to fact that teaching of music is carried out in condition of pedagogical communication, which is associated with formation of very subtle and complex relationships between teacher and pupils and it requires high compatibility, correlation and mutual understanding, success of development of musical, artistic-creative abilities of pupils depends on them. The ideas of outstanding of pedagogues-musicians on question of study and understanding of pupils’ personality, ability to communicate with them in process of musical teaching have been highlighted in the article. The peculiarities of study and understanding of originality of musical abilities of certain pupils have been aroused interest by teachers. On the base of conducted research, specially developed program of observation of pedagogical influences and methods of estimation of effectiveness of activity of teachers during individual lessons has been offered, and also ways of perfect of social-perceptual skills of teachers for improvement of study and understanding of musical abilities of pupils have been outlined
Abstract:Introduction: Urinary tract infections (UTIs) are inflammatory urinary tract infections caused by an unusual pathogen. The aim of this study is to investigate the utility of catalase test in the early diagnosis of UTIs. Methodology: The study involved 201 patients suspected to be infected with UTI and admitted to Near East Hospital between March and September 2020. The urine samples of the patients were microscopically analyzed after testing for catalase, nitrite and leukocyte esterase. Moreover, urine samples were inoculated onto blood agar and eosin methylene blue agar. Conventional microbiological techniques and vitek2 automated identification system were used for bacterial identification. Results: Total participants represented 87 (43.3%) male and 114 (56.7%) female. The bacterial growth was observed in 37 (18.4 %) specimens. Among them, 29 (78%) was tested positive for catalase activity, leucocyte esterase and nitrite tests. Erythrocyte strip positivity and catalase test positivity were 65 (78.0%), while leukocyte positivity in patients with cystitis were 26 (24.1 %). The sensitivity, specificity, negative predictive value and positive predictive value of the catalase test were 90.6 %, 33.7 %, 95%, and 20.5 %, respectively, compared to the catalase test results of the culture test. Conclusion: Catalase test is a simple, fast, and low-cost test that detects all positive results. It would be an effective alternative screening test for detecting bacteriuria or pyuria. However, catalase test is very sensitive but not very specific for early detection of UTI.
Abstract:In the field of drug discovery, accurately predicting the interaction between drug entities can reduce the risk of patients in clinical treatment. Meanwhile, accurate prediction of compound entity relations can also improve the efficiency of drug development. Aiming at the shortcomings in the extraction model of chemical entity reaction relations, we apply graph convolutional neural network to the task of chemical reaction relation extraction, and propose a multi-view algorithm to extract compound reaction relation features. The algorithm considers not only the internal structure feature information of compound entities, but also takes into account the interactive graph information between entities. Thus, it can fully mine the structure feature information in the graph. Experimental results show that the proposed method has higher accuracy than the state-of-the-art methods.
Abstract:Location-based social network (LBSN) is a new type of social network. Each user can not only establish friendships with other users, but also share their own events in different places by checking in. There is a large amount of auxiliary information in the location social networks, such as comment information, social information, geographic information, and so on. A large amount of auxiliary information has unique spatiotemporal characteristics, which can be used to alleviate the problems of cold start and data sparseness of the recommendation system. In this paper, we propose a joint convolution matrix factorization method that considers time relations, referred to as CMF-J. In a unified probabilistic matrix factorization recommendation framework, the method jointly considers the related comments of items, items relationships, users social influence and users comments. And integrates the processing results of the convolutional neural network to perform rating prediction. Finally, we conduct extensive experiments on the dataset Yelp. The experimental results demonstrate the superiority of CMF-J compared to several state-of-the-art methods in the two performance indicators of Root Mean Square Error and Mean Absolute Error.