Pub Date : 2021-12-17DOI: 10.1109/PIC53636.2021.9687023
Sotiria Karypidou, Ilias Georgousis, G. Papakostas
Computer Vision (CV) is undoubtedly one of the most popular forms of Artificial Intelligence (AI) and its implementation has gained considerable ground in all aspects of our lives, from security and automotive, to the night sky observation and astronomy. In general, CV uses pattern recognition techniques for identifying objects in visual media (both static and moving images). The current archetype in CV is largely based on supervised AI, which uses large data sets of human-labelled images for training. Machine Learning (ML) and Deep Learning (DL) models in computer vision have undergone a period of extremely rapid development in recent past years; in particular for object recognition and localisation tasks. An area of study with great interest in practical applications that concerns this essay, is astronomical images analysis. However, one of the main challenges facing researchers these days is the existence of large quantities of annotated data sets, in the appropriate resolution and scale. This challenge consequently asks for huge amounts of storage and high computational power. In this paper, we systematically review and analyze different challenges faced by astronomers and continue with state-of-the-art methodologies that were conducted over the last decade.
{"title":"Computer Vision for Astronomical Image Analysis","authors":"Sotiria Karypidou, Ilias Georgousis, G. Papakostas","doi":"10.1109/PIC53636.2021.9687023","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687023","url":null,"abstract":"Computer Vision (CV) is undoubtedly one of the most popular forms of Artificial Intelligence (AI) and its implementation has gained considerable ground in all aspects of our lives, from security and automotive, to the night sky observation and astronomy. In general, CV uses pattern recognition techniques for identifying objects in visual media (both static and moving images). The current archetype in CV is largely based on supervised AI, which uses large data sets of human-labelled images for training. Machine Learning (ML) and Deep Learning (DL) models in computer vision have undergone a period of extremely rapid development in recent past years; in particular for object recognition and localisation tasks. An area of study with great interest in practical applications that concerns this essay, is astronomical images analysis. However, one of the main challenges facing researchers these days is the existence of large quantities of annotated data sets, in the appropriate resolution and scale. This challenge consequently asks for huge amounts of storage and high computational power. In this paper, we systematically review and analyze different challenges faced by astronomers and continue with state-of-the-art methodologies that were conducted over the last decade.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121028555","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687058
Maxime Goubeaud, Nicolla Gmyrek, Farzin Ghorban, Lucas Schelkes, A. Kummert
In machine learning, data augmentation is commonly used to generate synthetic samples in order to augment datasets used to train models. The motivation behind data augmentation is to reduce the error-rate of models by increasing the diversity in the dataset. In this paper, we present a new data augmentation method for spectrograms of time series that we name Random Noise Boxes. Random Noise Boxes works by multiplying each spectrogram in a dataset with a predefined number of identical spectrograms and thereafter replacing randomly chosen square-sized parts of the resulting spectrograms with boxes of random noise pixels. We demonstrate the effectiveness of the proposed method by conducting experiments using differentsized CNN classifiers evaluated on nine well-known datasets from the UCR Time Series Classification Archive. We show that our method is beneficial in most cases, as we observe an increase of accuracy and F1-Score on most datasets.
{"title":"Random Noise Boxes: Data Augmentation for Spectrograms","authors":"Maxime Goubeaud, Nicolla Gmyrek, Farzin Ghorban, Lucas Schelkes, A. Kummert","doi":"10.1109/PIC53636.2021.9687058","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687058","url":null,"abstract":"In machine learning, data augmentation is commonly used to generate synthetic samples in order to augment datasets used to train models. The motivation behind data augmentation is to reduce the error-rate of models by increasing the diversity in the dataset. In this paper, we present a new data augmentation method for spectrograms of time series that we name Random Noise Boxes. Random Noise Boxes works by multiplying each spectrogram in a dataset with a predefined number of identical spectrograms and thereafter replacing randomly chosen square-sized parts of the resulting spectrograms with boxes of random noise pixels. We demonstrate the effectiveness of the proposed method by conducting experiments using differentsized CNN classifiers evaluated on nine well-known datasets from the UCR Time Series Classification Archive. We show that our method is beneficial in most cases, as we observe an increase of accuracy and F1-Score on most datasets.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129139164","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687028
Haotian Li, Baohang Zhang, Jiayi Li, Tao Zheng, Haichuan Yang
The water wave optimization (WWO) algorithm is a new cluster intelligence search method. It has the advantages of a small population size and simple parameter configuration. It is used to build an efficient mechanism for searching in high-dimensional solution spaces. However, it has a proclivity for becoming stuck in local optima. Coincidentally, the sparrow search algorithm (SSA) has good exploration ability. By combining WWO and SSA, we propose a hybrid algorithm, called WWOSSA. The experimental results of the WWOSSA algorithm based on 29 benchmark functions of IEEE CEC2017 have good optimization ability and a fast convergence rate.
{"title":"Using Sparrow Search Hunting Mechanism to Improve Water Wave Algorithm","authors":"Haotian Li, Baohang Zhang, Jiayi Li, Tao Zheng, Haichuan Yang","doi":"10.1109/PIC53636.2021.9687028","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687028","url":null,"abstract":"The water wave optimization (WWO) algorithm is a new cluster intelligence search method. It has the advantages of a small population size and simple parameter configuration. It is used to build an efficient mechanism for searching in high-dimensional solution spaces. However, it has a proclivity for becoming stuck in local optima. Coincidentally, the sparrow search algorithm (SSA) has good exploration ability. By combining WWO and SSA, we propose a hybrid algorithm, called WWOSSA. The experimental results of the WWOSSA algorithm based on 29 benchmark functions of IEEE CEC2017 have good optimization ability and a fast convergence rate.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116576171","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687038
Qianqian Yu, Yi-Yang Wang, Keqi Fan, Y. Zheng
Accuracy and speed have been two fundamental issues that are difficult to balance in object tracking. Trackers with high accuracy often have quite large network structures that require huge amounts of computing resources, therefore leading to a lower tracking speed. To address the problem, we propose a novel domain adaptive tracking algorithm to obtain a better balance between tracking speed and accuracy. A simple and effective domain adaptation component is employed to transfer features from the image classification domain to the object tracking domain. In addition, we construct an adaptive spatial pyramid pooling layer to substitute for the fully- connected layer connected to convolutional layers, which can significantly reduce computational complexity while achieving high tracking accuracy. Experiments on VOT2018, TrackingNet and OTB2015 shown the effectiveness of the proposed method. Compared with the state-of-the-art trackers, our tracker can obtain real-time tracking with a speed of 35 FPS.
{"title":"Domain Adaptive Visual Tracking with Multi-scale Feature Fusion","authors":"Qianqian Yu, Yi-Yang Wang, Keqi Fan, Y. Zheng","doi":"10.1109/PIC53636.2021.9687038","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687038","url":null,"abstract":"Accuracy and speed have been two fundamental issues that are difficult to balance in object tracking. Trackers with high accuracy often have quite large network structures that require huge amounts of computing resources, therefore leading to a lower tracking speed. To address the problem, we propose a novel domain adaptive tracking algorithm to obtain a better balance between tracking speed and accuracy. A simple and effective domain adaptation component is employed to transfer features from the image classification domain to the object tracking domain. In addition, we construct an adaptive spatial pyramid pooling layer to substitute for the fully- connected layer connected to convolutional layers, which can significantly reduce computational complexity while achieving high tracking accuracy. Experiments on VOT2018, TrackingNet and OTB2015 shown the effectiveness of the proposed method. Compared with the state-of-the-art trackers, our tracker can obtain real-time tracking with a speed of 35 FPS.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"1997 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116683255","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687079
Yoonsik Cheon
It is a good programming practice to include runtime checks called assertions in the code to check assumptions and invariants. Assertions are said to be often most effective when they encode design decisions and constraints. In this paper, we show our preliminary work on translating design constraints to assertions for mobile apps. Design properties and constraints are specified formally in the Object Constraint Language (OCL) and translated to executable assertions written in Dart, the language of the Flutter cross-platform framework. We consider various language and platform-specific features of OCL, Dart, and Flutter. In our approach, assertions are enabled only in debug mode and removed from the production code. It is important to reduce the memory footprint of a mobile app as the memory on a mobile device is a limited resource.
{"title":"Toward More Effective Use of Assertions for Mobile App Development","authors":"Yoonsik Cheon","doi":"10.1109/PIC53636.2021.9687079","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687079","url":null,"abstract":"It is a good programming practice to include runtime checks called assertions in the code to check assumptions and invariants. Assertions are said to be often most effective when they encode design decisions and constraints. In this paper, we show our preliminary work on translating design constraints to assertions for mobile apps. Design properties and constraints are specified formally in the Object Constraint Language (OCL) and translated to executable assertions written in Dart, the language of the Flutter cross-platform framework. We consider various language and platform-specific features of OCL, Dart, and Flutter. In our approach, assertions are enabled only in debug mode and removed from the production code. It is important to reduce the memory footprint of a mobile app as the memory on a mobile device is a limited resource.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"51 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132696061","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687011
Wei Wang, Yi Yang, Xin Xu
The image texture removal is a solution for the image content separation problem, which aim at preserving image edges and removing uninterested textures. This problem has wide applications in image feature extractions, such as texture extraction, detail enhancement and so on. A new image content decomposition approach is introduced with a minimization refinement architecture in this paper, which contains the guidance component and the total varation component. The guidance component introduces a rolling guidance filtering to iteratively update the more and more smoothing images. The total varation component uses a new total variation regularization method to remove the image texture and preserve the structural contents. The non-convex objective function is simplified into these two sub-problems, which yield a linear solution. Then the experiments demonstrate the prior performance of our method and its potential for many image processing applications.
{"title":"Image Texture Removal by Total Variantional Rolling Guidance","authors":"Wei Wang, Yi Yang, Xin Xu","doi":"10.1109/PIC53636.2021.9687011","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687011","url":null,"abstract":"The image texture removal is a solution for the image content separation problem, which aim at preserving image edges and removing uninterested textures. This problem has wide applications in image feature extractions, such as texture extraction, detail enhancement and so on. A new image content decomposition approach is introduced with a minimization refinement architecture in this paper, which contains the guidance component and the total varation component. The guidance component introduces a rolling guidance filtering to iteratively update the more and more smoothing images. The total varation component uses a new total variation regularization method to remove the image texture and preserve the structural contents. The non-convex objective function is simplified into these two sub-problems, which yield a linear solution. Then the experiments demonstrate the prior performance of our method and its potential for many image processing applications.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134490145","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687049
Yimin Peng, Rong Hu, Yiping Wen
In the big data environment, the sparsity problem of collaborative filtering recommendation algorithm becomes increasingly serious, which has a great impact on the accuracy of recommendation. In some recent researches, item categories were input into neural networks to enrich the embedded information in the process of training. However, these methods generally simultaneously use item categories and items as embedded information, which may weaken the importance of item categories. Therefore, this paper proposes a neural collaborative filtering method based on category assistance. In this method, the interaction between item category and user is first modeled by Neural Matrix Factorization ((Neu-MF)), which raises the impact of item category in the relationship extraction between items and users. Then, only the items in the trained results of categories are used in an optimized Neural Collaborative Filtering (NCF) framework for item recommendation. Based on the real ecommerce data set from Alibaba, experimental results show that this method obtains better result in the Hit Rate (HR) and the Normalized Discounted Cumulative Gain (NDCG) compared with other baseline methods.
{"title":"CA-NCF: A Category Assisted Neural Collaborative Filtering Approach for Personalized Recommendation","authors":"Yimin Peng, Rong Hu, Yiping Wen","doi":"10.1109/PIC53636.2021.9687049","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687049","url":null,"abstract":"In the big data environment, the sparsity problem of collaborative filtering recommendation algorithm becomes increasingly serious, which has a great impact on the accuracy of recommendation. In some recent researches, item categories were input into neural networks to enrich the embedded information in the process of training. However, these methods generally simultaneously use item categories and items as embedded information, which may weaken the importance of item categories. Therefore, this paper proposes a neural collaborative filtering method based on category assistance. In this method, the interaction between item category and user is first modeled by Neural Matrix Factorization ((Neu-MF)), which raises the impact of item category in the relationship extraction between items and users. Then, only the items in the trained results of categories are used in an optimized Neural Collaborative Filtering (NCF) framework for item recommendation. Based on the real ecommerce data set from Alibaba, experimental results show that this method obtains better result in the Hit Rate (HR) and the Normalized Discounted Cumulative Gain (NDCG) compared with other baseline methods.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134497971","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687026
Jema Sharin PankiRaj, A. Yassine, Salimur Choudhury
The emerging smart grid uniquely combines two-way communication and energy flow, allowing consumers to become active participants in market-based energy supply and demand strategies. In such a market, Peer-to-Peer (P2P) energy trading paradigm allows local communities and individuals who generate electricity to freely decide how and with whom they are going to trade it. The greatest challenge of P2P energy trading is how to design efficient mechanisms among rational participants that maximize their monetary benefits. Furthermore, since utility companies own the transmission lines, a key question that yet to be addressed in P2P markets is: how to match between different energy buyers and sellers while taking into account the physical constraints of the underlying grid infrastructure, e.g., capacity, congestion, and line transmission costs. This paper proposes a novel double-sided auction mechanism with a matching algorithm that addresses the aforementioned challenges. In this paper, the social welfare of the participants is modeled as an optimization problem with cost constraints incurred due to energy generation, operating and maintenance, capacity, and line transmission costs. The study provides theoretical analysis of the P2P auction model including mechanism design properties such as individual rationality, computational efficiency, and truthfulness. The results of the experiments indicate that the proposed auction model outperform existing systems and yields better economic incentives for participants.
{"title":"Double-Sided Auction Mechanism for Peer-to-Peer Energy Trading Markets","authors":"Jema Sharin PankiRaj, A. Yassine, Salimur Choudhury","doi":"10.1109/PIC53636.2021.9687026","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687026","url":null,"abstract":"The emerging smart grid uniquely combines two-way communication and energy flow, allowing consumers to become active participants in market-based energy supply and demand strategies. In such a market, Peer-to-Peer (P2P) energy trading paradigm allows local communities and individuals who generate electricity to freely decide how and with whom they are going to trade it. The greatest challenge of P2P energy trading is how to design efficient mechanisms among rational participants that maximize their monetary benefits. Furthermore, since utility companies own the transmission lines, a key question that yet to be addressed in P2P markets is: how to match between different energy buyers and sellers while taking into account the physical constraints of the underlying grid infrastructure, e.g., capacity, congestion, and line transmission costs. This paper proposes a novel double-sided auction mechanism with a matching algorithm that addresses the aforementioned challenges. In this paper, the social welfare of the participants is modeled as an optimization problem with cost constraints incurred due to energy generation, operating and maintenance, capacity, and line transmission costs. The study provides theoretical analysis of the P2P auction model including mechanism design properties such as individual rationality, computational efficiency, and truthfulness. The results of the experiments indicate that the proposed auction model outperform existing systems and yields better economic incentives for participants.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128419960","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687019
Pavél Llamocca Portella, Victoria López, Matilde Santos
People who suffer from depression or bipolar disorder have very different and complex indicators of their emotional state. The use of wearable smart devices can help to characterize the behaviour of these people and therefore allows the psychiatrist to decide the best treatment. In addition, those devices are able to extract a great amount of data from patients that can be analyzed with computer techniques. However, most patients experience fluctuations in mood according to a weekly cycle. The day of the week is a factor that influences a set of characteristics that describe the emotional state, like irritability or motivation. In this work, we analyze this factor and its influence on a set of mood variables gathered daily and their relation with the medical diagnostic of the patient. The analysis of the information is personalized since the data presents variations due to factors that affect the emotional state of each patient according to different ways and intensities. This work presents an improved mathematical model on the diagnosis by including the factor described before.
{"title":"Weighted Dependence of the Day of the Week in Patients with Emotional Disorders: A Mathematical Model","authors":"Pavél Llamocca Portella, Victoria López, Matilde Santos","doi":"10.1109/PIC53636.2021.9687019","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687019","url":null,"abstract":"People who suffer from depression or bipolar disorder have very different and complex indicators of their emotional state. The use of wearable smart devices can help to characterize the behaviour of these people and therefore allows the psychiatrist to decide the best treatment. In addition, those devices are able to extract a great amount of data from patients that can be analyzed with computer techniques. However, most patients experience fluctuations in mood according to a weekly cycle. The day of the week is a factor that influences a set of characteristics that describe the emotional state, like irritability or motivation. In this work, we analyze this factor and its influence on a set of mood variables gathered daily and their relation with the medical diagnostic of the patient. The analysis of the information is personalized since the data presents variations due to factors that affect the emotional state of each patient according to different ways and intensities. This work presents an improved mathematical model on the diagnosis by including the factor described before.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"35 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127557613","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 : 2021-12-17DOI: 10.1109/PIC53636.2021.9687055
Jishan Sun, Yaojie Chen, Wei Wang
When the intelligent water cannon strikes a surface target, it needs to know the distance to the strike target and automatically adjust the strike angle to complete the accurate strike mission. Based on this estimation, the control system of the water cannon would automatically achieve the strike mission. For a universal usage, a monocular image depth estimation method based on SC-SfMLearner is used, which first estimates the depth information of the image from one sample of the real-time video frames and then uses a polynomial fitting model to transfer a depth map into the physical distance in the real world. The experimental results show that the mean square deviation of the predicted distance results in the practical environment for shore-side water targets is between 0.02 and 0.03, and the accuracy rate is above 95 %, which is a good prediction and effectively addresses the accuracy of striking water targets in practical applications.
{"title":"Single Image Based Depth Estimation for Maritime Surface Targets","authors":"Jishan Sun, Yaojie Chen, Wei Wang","doi":"10.1109/PIC53636.2021.9687055","DOIUrl":"https://doi.org/10.1109/PIC53636.2021.9687055","url":null,"abstract":"When the intelligent water cannon strikes a surface target, it needs to know the distance to the strike target and automatically adjust the strike angle to complete the accurate strike mission. Based on this estimation, the control system of the water cannon would automatically achieve the strike mission. For a universal usage, a monocular image depth estimation method based on SC-SfMLearner is used, which first estimates the depth information of the image from one sample of the real-time video frames and then uses a polynomial fitting model to transfer a depth map into the physical distance in the real world. The experimental results show that the mean square deviation of the predicted distance results in the practical environment for shore-side water targets is between 0.02 and 0.03, and the accuracy rate is above 95 %, which is a good prediction and effectively addresses the accuracy of striking water targets in practical applications.","PeriodicalId":297239,"journal":{"name":"2021 IEEE International Conference on Progress in Informatics and Computing (PIC)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125953686","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：