Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition最新文献

{"title":"Brain control jigsaw puzzle system based on hybrid brain computer interface of motor imagery and steady-state visual evoked potential","authors":"Rongxin Jie, Banghua Yang, Zhaokun Wang, Jun Ma, Xinxing Xia, Shouwei Gao","doi":"10.1145/3581807.3581872","DOIUrl":"https://doi.org/10.1145/3581807.3581872","url":null,"abstract":"With the development of information decoding technology, the field of Brain-computer interface (BCI) has developed rapidly in recent years. Among them, Motor Imagery Brain-computer Interface (MI-BCI) and Steady state visual evoked potential Brain-computer Interface (SSVEP-BCI) have been effectively applied in some brain-controlled rehabilitation training systems to assist stroke patients in their normal life. In this paper, a brain-controlled jigsaw puzzle system based on Motor Imagery and Steady state visual evoked potential (MI-SSVEP) hybrid brain-machine is constructed. In this system, the left-right moving jigsaw puzzle uses the MI-BCI paradigm and the up-down moving jigsaw puzzle uses the SSVEP-BCI paradigm. To reduce the difficulty for patients, the system will set the moving route of the puzzle in advance. When the puzzle piece needs to move left or right, the system will remind the patient through voice and words that the patient needs to Imagine clenching his fist with his left or right hand at this time. When the puzzle piece needs to move up and down, the system will remind the patient to gaze at the upward or downward flashing arrow. If the patient makes an incorrect recognition, the system will re-open the recognition at the current position until it is correct. Compared with the ordinary rehabilitation training system, this system adds the elements of the jigsaw puzzle, so that patients can complete the training in the process of enjoying the game. The success of the jigsaw puzzle will also increase the sense of achievement for patients, and play the effect of rehabilitation training while maintaining the healthy state of mind of patients. The average recognition time of MI is 2.5s, and the accuracy is 65%. The average recognition time of SSVEP is 1.5s, and the accuracy is 95%. The system operates stably, each subject was able to complete the puzzle task quickly. The experimental results demonstrate the feasibility and potential of this hybrid brain-machine system and provide a new idea for the rehabilitation training of stroke patients.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"53 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115858521","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Infrared Sea Surface Ship Target Detection Algorithm Based on Improved YOLOV5","authors":"Yang Shi, Qiang Wu, Xin Zheng, Bin Yue","doi":"10.1145/3581807.3581823","DOIUrl":"https://doi.org/10.1145/3581807.3581823","url":null,"abstract":"Ship target detection on the sea surface is one of the most common scenes in the field of target detection. The accuracy of target detection plays an important role in the field of sea rescue and defense. However, complex infrared sea surface scenes, such as island shore, fish scale wave, sea surface bright band and other disturbances, bring great challenges to target detection. In this paper, we propose an improved YOLOV5 model by analyzing the characteristics of infrared image imaging and ship target on the sea surface. Aiming at the problem of information loss of small targets in the deep layer of the network, we redesigned the backbone network, which was composed of four Multi-scale residual blocks, and each block was connected by CBAM (Convolutional Block Attention Module) attention mechanism to improve information transmission and fusion between different feature layers. Aiming at the problem of complex sea surface scene interference, we use FPN and PAN (Feature Pyramid Network and Personal Area Network) mechanism to construct feature fusion network in neck, and add shallow feature fusion in FPN. Our model achieves good performance both in terms of latency and accuracy. In a self-collected sea surface scene dataset, multiple SOTAs for detection tasks are compared and the results demonstrate the superiority of the proposed method.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130421014","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Plant Diseases Recognition on Digital Images using Swin Transformer","authors":"Hongjun Song, Yuanyuan Gao","doi":"10.1145/3581807.3581839","DOIUrl":"https://doi.org/10.1145/3581807.3581839","url":null,"abstract":"Plant diseases seriously affect the safety of food production and must be quickly recognized and detected. In recent years, the traditional convolutional neural network has been widely used to diagnose plant diseases. Swin Transformer was used to train and evaluate the PlantVillage dataset, which includes 54303 healthy and unhealthy leaf images that divided into 38 categories by species and disease. The trained model based on Swin transformer learning achieves an accuracy of 98.1% on training data set, 98.7% on testing data set, which proves the feasibility of this method.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131104631","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Recommendation Algorithm Incorporating Self-Attention Mechanism and Knowledge Graph","authors":"Jingjing Hou, Yuchen Jin, Yi-Wei Liu, Zhenhua Zhang, Qinghua Zhao","doi":"10.1145/3581807.3581858","DOIUrl":"https://doi.org/10.1145/3581807.3581858","url":null,"abstract":"To address the problems of sparse data, low recommendation accuracy and poor recommendation effect in recommendation systems. In this paper, we propose a recommendation algorithm that fuses the self-attention mechanism and knowledge graph. The algorithm mainly includes recommendation module, knowledge graph feature learning, and self-attention. In this algorithm recommendation system module, a user and an item are input, and the input item vector and entity vector are embedded in the self-attention module, so that the feature representation of these two vectors is enhanced. The knowledge graph feature representation module maps the head entities and relations in the triad into a continuous vector space, and calculates the corresponding values through the score function. The recommendation module and the knowledge graph representation model are connected through the cross-compression unit embedded in the self-attentive mechanism. Finally, the loss of each module is calculated by a loss function. Experiments on three different publicly available datasets show that: the embedded attention mechanism module introduced can well solve the accuracy problem of the recommendation system; Secondly, the embedded attention mechanism cross-compression unit module enhances the recommendation system in which vectors are compressed in horizontal and vertical directions. Finally, through experiments comparing other algorithms, the proposed method improves the recommendation accuracy and effectiveness in the recommendation system.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129987707","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Brief Comparison of Deep Learning Methods for Semantic Segmentation","authors":"Chang-Bin Zhang, Xiangyun Bai","doi":"10.1145/3581807.3581830","DOIUrl":"https://doi.org/10.1145/3581807.3581830","url":null,"abstract":"Semantic segmentation is widely used in many fields, which is the key and difficult point in computer vision field.In recent years, with the rapid development of deep learning, deep learning has greatly improved the performance of semantic segmentation. Many methods have been proposed.This paper mainly reviews the research progress of semantic segmentation model based on convolutional neural network (CNN), compares the methods of the same class, and analyzes the connection and difference of each method. In this paper, we mainly discuss the recent semantic segmentation models to improve segmentation accuracy from different strategies, and compare and analyze the relationships and differences between these methods. we also prospected the future development direction of semantic segmentation methods.This paper hopes to give readers an understanding of the progress and challenges of CNN based semantic segmentation research.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122443480","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"DOA Estimation of Propagator Method Based on FFT and Golden Section","authors":"Huijing Dou, W. Guo, Dongxu Xie","doi":"10.1145/3581807.3581865","DOIUrl":"https://doi.org/10.1145/3581807.3581865","url":null,"abstract":"In view of the huge amount of calculation of the spectral peak search module in the propagation operator (PM) algorithm, it is often necessary to sacrifice direction finding accuracy in exchange for direction finding real-time performance in engineering practice. A propagation operator angle of arrival (DOA) estimation algorithm combining Fast Fourier Transform (FFT) and golden section method is proposed. The proposed algorithm firstly divides the covariance matrix and then performs linear operation to obtain the noise vector matrix, and then performs fast Fourier transform on it to roughly estimate the direction of arrival angle, Then, the golden section method is used to iteratively obtain the extreme value of the pseudospectral function within the estimated angle range to achieve accurate estimation of DOA.Theoretical analysis and simulation results show that the improved algorithm greatly reduces the computational complexity and execution time of the algorithm, and the direction finding accuracy is improved to a certain extent compared with the classical PM algorithm with 1° scanning step.In the environment of multi-signal coexistence and low signal-to-noise ratio, the direction finding stability is good.The improved algorithm can replace the spectral peak search module in the classic PM algorithm, and realize high-precision real-time direction finding while reducing the complexity of the PM algorithm, which has good practical value.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132941188","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Coarse-to-fine Palmprint-Specific Quality Feature Learning for Palmprint Image Quality Assessment","authors":"Xiao Sun, Lunke Fei, Zhi-xiang Liu, Zhenkai Tang, Jijia Chen, Jiangpeng Su, Shiqiao Zhang","doi":"10.1145/3581807.3581845","DOIUrl":"https://doi.org/10.1145/3581807.3581845","url":null,"abstract":"Palmprint recognition has aroused broad concern recently due to its several advantages, such as contactless, hygienic, and less-invasive properties. However, most existing palmprint recognition methods focus on feature extraction and matching without assessing the quality of palmprint images, making the recognition result sensitive []to low-quality images. To the best of our knowledge, there is still no literature with an attempt to specially study the problem of palmprint image quality assessment. To address this, in this paper, we propose an end-to-end palmprint-specific quality feature learning and assessment framework, which consists of an attention-embedded coarse feature learning network and a fine quality feature learning network. The coarse feature learning network aims to extensively explore the quality-related information from palmprint images by embedding texture maps into the palmprint images via an attention-embedded CNN network. Then, the fine quality feature learning network is learned to extract the latent quality-specific features of palmprint images. Moreover, we established a new quality-labeled palmprint image benchmark database based on an automatic quality labeling scheme. Experimental results on the new palmprint image benchmark database demonstrate that the proposed method consistently outperforms the state-of-the-art methods.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133567059","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-Channel Text Classification Model Based on ERNIE","authors":"Dongxue Bao, Donghong Qin, Lila Hong, Siqi Zhan","doi":"10.1145/3581807.3581853","DOIUrl":"https://doi.org/10.1145/3581807.3581853","url":null,"abstract":"Aiming at the large amount of news and review text data, sparse features, and the inability of traditional text feature representation to dynamically obtain grammatical structure, semantic information, and multi-dimensional rich feature representation of entity phrases. This paper proposes to obtain more generalized knowledge semantic feature information such as rich context phrases, entity words and so on by integrating knowledge enhanced semantic representation (Enhanced Representation Through Knowledge Integration, ERNIE). The pre-trained language model ERNIE hides words and entities by random Semantic unit prediction context realizes word vector language representation, and the output vector representation of ERNIE is input to BiLSTM, Attention mechanism and DPCNN network model to generate high-order text feature vectors, and each channel vector is processed by BatchNormalization and ReLU activation functions respectively.Thus, the semantic description information of the multi-channel word vector is fused. The model proposed in this paper can not only improve the training speed and prevent overfitting, but also enhance the feature information such as semantics and grammatical structure, thereby improving the text classification effect. By comparing the two datasets with other improved ERNIE models in terms of accuracy, precision, recall, and F1 value, the experimental results show that the model proposed in this paper can obtain multi-dimensional rich semantic grammatical structure features for text classification, and then improve the Text classification effect.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129189570","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vehicle and pedestrian detection method based on improved YOLOX in foggy environment","authors":"Li-zong Lin, Zhaohui Liu, Shiji Zhao, Jinzhao Zhang","doi":"10.1145/3581807.3581819","DOIUrl":"https://doi.org/10.1145/3581807.3581819","url":null,"abstract":"Most of the current vision sensor-based target detection is suitable for good weather conditions. Adverse weather conditions, especially foggy environments, significantly reduce visibility, which seriously affects the target detection performance. To improve driving safety in foggy environments, this paper proposes an improved YOLOX-based vehicle and pedestrian detection method in foggy environments. The method is based on the advanced YOLOX network model and introduces an attention mechanism in the feature extraction network to enhance the network's extraction of target features in foggy images. Some images in the training dataset are fogged to supplement the target-specific features in foggy environments and improve the robustness of the target detection network in foggy environments. The idea of migration learning is used in the training process to save training time and optimize the training effect. The experimental results show that the target detection method proposed in this paper has significantly improved the detection performance of vehicles and pedestrians in the foggy environment, with an 11.35% improvement in mAP, and the detection effect is better than the GCANet image defogging method. The effectiveness of the method improvement is proved.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127590346","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Brain Network Analysis Results of Finger Motor Execution and Motor Imagery","authors":"Fengyue Liu, Lin Lu, Abdelkader Nasreddine Belkacem, Jiaxin Zhang, Penghai Li, Jun-bing Liang, Changming Wang, Chao Chen","doi":"10.1145/3581807.3581873","DOIUrl":"https://doi.org/10.1145/3581807.3581873","url":null,"abstract":"In recent years, finger movement execution and motor imagery have become a new method in the field of motor function rehabilitation after stroke, which has important reference value for the rehabilitation of patients with motor dysfunction. The ability of motor imagery retained by stroke patients makes it possible to recover the function of neural plasticity motor system. In this paper, conditional Granger causality analysis method is used. On the premise that the residuals of the subjects meet the Durbin White test criterion (P>0.60), five network nodes directly related to motor function are selected to analyze and compare the conditional Granger causality connectivity and connection strength between these regions. Drawing the whole brain network diagram of left-hand finger movement execution and movement imagination, getting the connection direction and connection strength of each brain network node, looking for objective quantitative indicators, which can be used as indicators of rehabilitation of stroke patients. The whole brain network diagram clearly shows that the left pre-motor area (PMA) has a stimulating connection to the other three regions during the execution of the left-hand finger movement. In the process of left-handed finger motion imagination, the left-PMA has stimulating connections to the other two regions, and the stimulating connections to other regions are the most in the five regions. Through the study, it was found that in the exercise execution and motor imagination of the left hand, the stimulation of the left-PMA on other regions was more obvious than that of other regions, and the connection between this region and other regions was also stronger, which could accelerate the recovery process of patients by strengthening the treatment of the left-PMA in the late treatment of stroke patients.","PeriodicalId":292813,"journal":{"name":"Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-11-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127686170","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}