Pub Date : 2021-08-01DOI: 10.1109/ISCEIC53685.2021.00051
Kuan-sheng Zou, Zhenbang Jiang, Qian Zhang
Power Line Extraction (PLE) is useful for low-altitude aircraft avoiding the high-voltage power line, and it also can be used in the power line autonomous inspection. PLE based on aerial images has caused many researchers to study with enthusiasm, because machine learning methods play an important role in PLE. The PLE methods based on machine learning are summarized in this paper, and then the research progresses of PLE methods based on traditional image processing, machine learning and deep learning are analyzed; then the future research trends of PLE are predicted based on the survey of novel methods proposed within the pasted two years. The PLE belongs to the interdisciplinary research direction, and it has certain reference value for researchers with research fields such as power fault diagnosis, image processing, and machine learning.
{"title":"Research Progresses and Trends of Power Line Extraction based on Machine Learning","authors":"Kuan-sheng Zou, Zhenbang Jiang, Qian Zhang","doi":"10.1109/ISCEIC53685.2021.00051","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00051","url":null,"abstract":"Power Line Extraction (PLE) is useful for low-altitude aircraft avoiding the high-voltage power line, and it also can be used in the power line autonomous inspection. PLE based on aerial images has caused many researchers to study with enthusiasm, because machine learning methods play an important role in PLE. The PLE methods based on machine learning are summarized in this paper, and then the research progresses of PLE methods based on traditional image processing, machine learning and deep learning are analyzed; then the future research trends of PLE are predicted based on the survey of novel methods proposed within the pasted two years. The PLE belongs to the interdisciplinary research direction, and it has certain reference value for researchers with research fields such as power fault diagnosis, image processing, and machine learning.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115055861","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}
Pub Date : 2021-08-01DOI: 10.1109/ISCEIC53685.2021.00008
Mingfei Li, Haibin Liu, Donglai Xu, Chaowei Lu
The steering gear is a very critical component in the control system of the aircraft with steering gear. In order to achieve a high-precision rudder deviation command, it is necessary to detect the mechanical zero position of the steering gear. Based on machine vision, a non-contact method for capturing the mechanical zero position of the gas steering gear was proposed, which aimed to transmit the position of the mechanical zero position to the encoder, and saved it to the system. The method includes three parts: identification, capture and transmission of the center of the mechanical zero position. The edge detection method was used to identify the center of the mechanical zero position which was captured by overlapping the center of the mechanical zero position with the pixel reference point in the pixel coordinate system. To complete the transmission of the reference, according to the relative position of the pixel reference point and the mechanical zero position to be tested, the angle inverse method was used to calculate the encoder value corresponding to the mechanical zero position. This study provides a reliable basis for the calculation of rudder deflection angle and electrical compensation value.
{"title":"Research on the Mechanical Zero Position Capture and Transfer of Steering Gear Based on Machine Vision","authors":"Mingfei Li, Haibin Liu, Donglai Xu, Chaowei Lu","doi":"10.1109/ISCEIC53685.2021.00008","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00008","url":null,"abstract":"The steering gear is a very critical component in the control system of the aircraft with steering gear. In order to achieve a high-precision rudder deviation command, it is necessary to detect the mechanical zero position of the steering gear. Based on machine vision, a non-contact method for capturing the mechanical zero position of the gas steering gear was proposed, which aimed to transmit the position of the mechanical zero position to the encoder, and saved it to the system. The method includes three parts: identification, capture and transmission of the center of the mechanical zero position. The edge detection method was used to identify the center of the mechanical zero position which was captured by overlapping the center of the mechanical zero position with the pixel reference point in the pixel coordinate system. To complete the transmission of the reference, according to the relative position of the pixel reference point and the mechanical zero position to be tested, the angle inverse method was used to calculate the encoder value corresponding to the mechanical zero position. This study provides a reliable basis for the calculation of rudder deflection angle and electrical compensation value.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"74 4","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"113983526","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}
Image recognition algorithms based on deep learning techniques have played an important role in the military, medical, industrial, and many other applications. However, most existing deep neural networks consume excessive computational resources which are unaffordable for the widely used edge devices, such as mobile phones. In this paper, we propose a lightweight network CResNet based on ResNet-50 by combining efficient channel pruning with depthwise decomposition. Ablation experiments are carried out based on the Animals-10 dataset for measuring the impact of each adopted technique. Great compression performance of the model parameters can be achieved at the price of slightly lower accuracy. Eventually, CResNet results in 4.08 M parameters, which is only one-fifth of the parameter size of the original ResNet-50, sufficiently reducing resource consumption. Approximately 90.2% Top-1 classification accuracy estimated on Animals-10 can be achieved by our lightweight CResNet. Compared to ResNet-50 and many existing lightweight networks, this work achieves a better tradeoff between segmentation accuracy and computing complexity by optimizing the computational efficiency, resulting in a small model size and a decent accuracy.
{"title":"Efficient Residual Network Compression for Optimizing the Accuracy-Complexity Tradeoff","authors":"A. Luo, Beibei Huang, Yuan Li, Chang Lu, Rui Wang, Zunkai Huang, Yicong Zhou","doi":"10.1109/ISCEIC53685.2021.00022","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00022","url":null,"abstract":"Image recognition algorithms based on deep learning techniques have played an important role in the military, medical, industrial, and many other applications. However, most existing deep neural networks consume excessive computational resources which are unaffordable for the widely used edge devices, such as mobile phones. In this paper, we propose a lightweight network CResNet based on ResNet-50 by combining efficient channel pruning with depthwise decomposition. Ablation experiments are carried out based on the Animals-10 dataset for measuring the impact of each adopted technique. Great compression performance of the model parameters can be achieved at the price of slightly lower accuracy. Eventually, CResNet results in 4.08 M parameters, which is only one-fifth of the parameter size of the original ResNet-50, sufficiently reducing resource consumption. Approximately 90.2% Top-1 classification accuracy estimated on Animals-10 can be achieved by our lightweight CResNet. Compared to ResNet-50 and many existing lightweight networks, this work achieves a better tradeoff between segmentation accuracy and computing complexity by optimizing the computational efficiency, resulting in a small model size and a decent accuracy.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"129 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123394150","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}
Pub Date : 2021-08-01DOI: 10.1109/ISCEIC53685.2021.00089
Ping Hu, Kai Liu, Lin-Sen Zhang, Xue-Xia Liu
To realize real-time weapon trajectory in VC, the paper compares various methods and selects NI Measurement Studio with Visual C++ tool kit. The using ActiveX technology not only including the combine of NI graph control with VC, the development of high-efficiency and practical trajectory graph realization method were also included. Applying the method into practical project it gains operation of VI graph control, it changesextreme values of graph control to realize trajectory’s zooming and moving.Practical application verifies validity and advantage of the method.
{"title":"Implementation of Virtual Instrument Graphics Control in VC","authors":"Ping Hu, Kai Liu, Lin-Sen Zhang, Xue-Xia Liu","doi":"10.1109/ISCEIC53685.2021.00089","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00089","url":null,"abstract":"To realize real-time weapon trajectory in VC, the paper compares various methods and selects NI Measurement Studio with Visual C++ tool kit. The using ActiveX technology not only including the combine of NI graph control with VC, the development of high-efficiency and practical trajectory graph realization method were also included. Applying the method into practical project it gains operation of VI graph control, it changesextreme values of graph control to realize trajectory’s zooming and moving.Practical application verifies validity and advantage of the method.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121532321","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}
Pub Date : 2021-08-01DOI: 10.1109/ISCEIC53685.2021.00058
Yumeng Jiang
At present, inter-beam interference elimination has great development needs in multi-beam mobile satellite communication. In this paper, based on the precoding technology research and application, which is a reliable way, and through the analysis of the forward link mode of multi-beam satellite communication system, the principle of realization of the linear technology represented by forced zero ZF precoding and the nonlinear technology represented by dirty paper DPC precoding was introduced, and then the advantages and disadvantages of the two precoding techniques and their application scope were summarized. Finally, an improved DPC-RZF algorithm was proposed, and its performance was compared with that of ZF and RZF by MATLAB simulation. When the number of beams was large, the system throughput increased by nearly 30%, which proved the effectiveness.
{"title":"Research on precoding techniques of multi-beam satellite communication","authors":"Yumeng Jiang","doi":"10.1109/ISCEIC53685.2021.00058","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00058","url":null,"abstract":"At present, inter-beam interference elimination has great development needs in multi-beam mobile satellite communication. In this paper, based on the precoding technology research and application, which is a reliable way, and through the analysis of the forward link mode of multi-beam satellite communication system, the principle of realization of the linear technology represented by forced zero ZF precoding and the nonlinear technology represented by dirty paper DPC precoding was introduced, and then the advantages and disadvantages of the two precoding techniques and their application scope were summarized. Finally, an improved DPC-RZF algorithm was proposed, and its performance was compared with that of ZF and RZF by MATLAB simulation. When the number of beams was large, the system throughput increased by nearly 30%, which proved the effectiveness.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125995497","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}
Pub Date : 2021-08-01DOI: 10.1109/ISCEIC53685.2021.00085
L. Guan, Yubing Han, Pandong Zhang
Hyperspectral image (HSI) is a kind of special remote sensing image, which provides rich spatial information as well as spectral information of ground objects. 3D-CNN can extract the spectral and spatial features of hyperspectral image based on this characteristic of hyperspectral image. Firstly, the hyperspectral image data were normalized to accelerate the convergence of the network in the training. Then, a three-dimensional multi-scale residual block similar to Resnet block is designed in the network, and BN (batch normalization) layer is added to alleviate over fitting. Finally, a softmax layer outputs the classification results. The experimental results were compared with SVM and several mainstream CNN methods. In the Indian Pines dataset, compared with the performance of second model, the overall classification accuracy is increased by 1.29%, and the model parameters are around one third of the of second model; in the Pavia University dataset, the overall classification accuracy is increased by 2.1%, and the model parameters are also about one third of the performance the second model. The effects of skip-connection, pixel block size, and different spectral step of first convolution layer are also discussed. Experiments show that the network model proposed in this paper can extract better classification features and has less parameters than the traditional hyperspectral image classification algorithm, and make the hyperspectral remote sensing image classification more accurate and efficient.
{"title":"Hyperspectral image classification based on improved multi-scale residual network structure","authors":"L. Guan, Yubing Han, Pandong Zhang","doi":"10.1109/ISCEIC53685.2021.00085","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00085","url":null,"abstract":"Hyperspectral image (HSI) is a kind of special remote sensing image, which provides rich spatial information as well as spectral information of ground objects. 3D-CNN can extract the spectral and spatial features of hyperspectral image based on this characteristic of hyperspectral image. Firstly, the hyperspectral image data were normalized to accelerate the convergence of the network in the training. Then, a three-dimensional multi-scale residual block similar to Resnet block is designed in the network, and BN (batch normalization) layer is added to alleviate over fitting. Finally, a softmax layer outputs the classification results. The experimental results were compared with SVM and several mainstream CNN methods. In the Indian Pines dataset, compared with the performance of second model, the overall classification accuracy is increased by 1.29%, and the model parameters are around one third of the of second model; in the Pavia University dataset, the overall classification accuracy is increased by 2.1%, and the model parameters are also about one third of the performance the second model. The effects of skip-connection, pixel block size, and different spectral step of first convolution layer are also discussed. Experiments show that the network model proposed in this paper can extract better classification features and has less parameters than the traditional hyperspectral image classification algorithm, and make the hyperspectral remote sensing image classification more accurate and efficient.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133341281","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}
Pub Date : 2021-08-01DOI: 10.1109/ISCEIC53685.2021.00064
Meng’An Shi, Huimin Cai, Yang Gao
This paper briefly describes the similarities and differences of the mainstream models of deep learning target detection box, analyzes the characteristics and advantages of Mask RCNN, a universal target detection box, and focuses on the application of Mask RCNN in human posture detection in multi- person human posture task. Through the analysis, it is considered that the advantage of Mask RCNN in multi-person human posture detection task is the accuracy, while the bottleneck is the detection speed. To solve this problem, an optimization of Mask RCNN model based on MobileNet was proposed to accelerate the inference calculation speed of Mask RCNN. At the same time, in order to further improve the detection accuracy of Mask RCNN, a method of using pixel segmentation results to assist the detection of human body key points is proposed. Experimental results show that compared with the original algorithm, it improves the reasoning speed and reduces the false detection rate caused by the environment.
{"title":"Optimization of Human Pose Detection Based on Mask RCNN","authors":"Meng’An Shi, Huimin Cai, Yang Gao","doi":"10.1109/ISCEIC53685.2021.00064","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00064","url":null,"abstract":"This paper briefly describes the similarities and differences of the mainstream models of deep learning target detection box, analyzes the characteristics and advantages of Mask RCNN, a universal target detection box, and focuses on the application of Mask RCNN in human posture detection in multi- person human posture task. Through the analysis, it is considered that the advantage of Mask RCNN in multi-person human posture detection task is the accuracy, while the bottleneck is the detection speed. To solve this problem, an optimization of Mask RCNN model based on MobileNet was proposed to accelerate the inference calculation speed of Mask RCNN. At the same time, in order to further improve the detection accuracy of Mask RCNN, a method of using pixel segmentation results to assist the detection of human body key points is proposed. Experimental results show that compared with the original algorithm, it improves the reasoning speed and reduces the false detection rate caused by the environment.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"35 1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131695032","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}
Pub Date : 2021-08-01DOI: 10.1109/ISCEIC53685.2021.00087
Yonghua Chen, Jiawei Chen
In view of the intelligent requirements of industrial production, this paper proposes the construction of industrial Internet system, firstly, it starts from the problems that need to be solved by industrial Internet, the needs of the industrial Internet are analyzed. The industrial Internet architecture is proposed. The application scenarios of the industrial Internet are discussed in detail. Subsequently, the architecture of the industrial Internet is described, Finally, a technology company is taken as an example, and the implementation goals and effects of the industrial Internet project is proved.
{"title":"Industrial Internet System Construction and Engineering Practice","authors":"Yonghua Chen, Jiawei Chen","doi":"10.1109/ISCEIC53685.2021.00087","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00087","url":null,"abstract":"In view of the intelligent requirements of industrial production, this paper proposes the construction of industrial Internet system, firstly, it starts from the problems that need to be solved by industrial Internet, the needs of the industrial Internet are analyzed. The industrial Internet architecture is proposed. The application scenarios of the industrial Internet are discussed in detail. Subsequently, the architecture of the industrial Internet is described, Finally, a technology company is taken as an example, and the implementation goals and effects of the industrial Internet project is proved.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"93 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134288712","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}
In the soft switch inverter system, the core part is the lower computer of the main chip with the TMS320F28335 DSP. In order for the operator to issue the command control system operation and intuitively reflect the operating performance of the system, this paper applies virtual instrument technology to soft switch inverter system and designs the software of the upper computer of soft switch inverter system with the help of the powerful graphics control of virtual instrument. The upper computer and the lower computer realize information interaction through CAN communication. On the upper computer, the motor operation status is displayed using control software written by Lah/Windows CVI. The upper computer sends instructions such as rotation speed setting according to the system’s needs. Through experiments, the correctness and reliability of the upper computer software are verified.
{"title":"Software Design of the Upper Computer of Soft Switching Inverter System Based on Virtual Instrument Technology","authors":"Ping Hu, Kai Liu, Lin-Sen Zhang, Xue-Xia Liu, Hao Chen, Chao Zhang","doi":"10.1109/ISCEIC53685.2021.00083","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00083","url":null,"abstract":"In the soft switch inverter system, the core part is the lower computer of the main chip with the TMS320F28335 DSP. In order for the operator to issue the command control system operation and intuitively reflect the operating performance of the system, this paper applies virtual instrument technology to soft switch inverter system and designs the software of the upper computer of soft switch inverter system with the help of the powerful graphics control of virtual instrument. The upper computer and the lower computer realize information interaction through CAN communication. On the upper computer, the motor operation status is displayed using control software written by Lah/Windows CVI. The upper computer sends instructions such as rotation speed setting according to the system’s needs. Through experiments, the correctness and reliability of the upper computer software are verified.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131001863","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}
Pub Date : 2021-08-01DOI: 10.1109/ISCEIC53685.2021.00046
Weiqiang Lv, Hanjie Wen, Hu Chen
Point clouds are a very popular way of representing 3D data. The development of various advanced devices has made it easily accessible. However, the acquired point cloud data usually has the following characteristics: the number of points is not easily controlled, sparse and non-uniform. These characteristics make it difficult to apply the acquired point cloud data directly to various tasks. To effectively address these issues, we propose a new method based on generative adversarial networks to implement upsampling pre-processing of point clouds. It is possible to easily upsample the number of points in the point cloud to our desired value and the obtained point cloud data can be very homogeneous while maintaining the original contours. In detail, we have introduced Skip-attention to our generator, which allows the network to effectively fuse the local and global features of the point cloud, and in addition to this, we have used PointNet-Mix as our discriminator, a simple and lightweight structure that works well with our generator. Extensive qualitative and quantitative experiments have demonstrated that the upsampling data obtained using our method can achieve equally competitive results.
{"title":"Point Cloud Upsampling by Generative Adversarial Network with Skip-attention","authors":"Weiqiang Lv, Hanjie Wen, Hu Chen","doi":"10.1109/ISCEIC53685.2021.00046","DOIUrl":"https://doi.org/10.1109/ISCEIC53685.2021.00046","url":null,"abstract":"Point clouds are a very popular way of representing 3D data. The development of various advanced devices has made it easily accessible. However, the acquired point cloud data usually has the following characteristics: the number of points is not easily controlled, sparse and non-uniform. These characteristics make it difficult to apply the acquired point cloud data directly to various tasks. To effectively address these issues, we propose a new method based on generative adversarial networks to implement upsampling pre-processing of point clouds. It is possible to easily upsample the number of points in the point cloud to our desired value and the obtained point cloud data can be very homogeneous while maintaining the original contours. In detail, we have introduced Skip-attention to our generator, which allows the network to effectively fuse the local and global features of the point cloud, and in addition to this, we have used PointNet-Mix as our discriminator, a simple and lightweight structure that works well with our generator. Extensive qualitative and quantitative experiments have demonstrated that the upsampling data obtained using our method can achieve equally competitive results.","PeriodicalId":342968,"journal":{"name":"2021 2nd International Symposium on Computer Engineering and Intelligent Communications (ISCEIC)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-08-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133800237","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：