Pub Date : 2019-07-01DOI: 10.1109/NAECON46414.2019.9057909
C. Yakopcic, Nayim Rahman, Tanvir Atahary, T. Taha, Alex Beigh, Scott Douglass
Cognitive agents are typically utilized in autonomous systems for automated decision making. These systems interact at real time with their environment and are generally heavily power constrained. Thus, there is a strong need for a real time agent running on a low power platform. The agent examined is the Cognitively Enhanced Complex Event Processing (CECEP) architecture. This is an autonomous decision support tool that reasons like humans and enables enhanced agent-based decision-making. It has applications in a large variety of domains including autonomous systems, operations research, intelligence analysis, and data mining. One of the most time consuming and key components of CECEP is the mining of knowledge from a repository described as a Cognitive Domain Ontology (CDO). One problem that is often tasked to CDOs is asset allocation. Given the number of possible solutions in this allocation problem, determining the optimal solution via CDO can be very time and energy consuming. A grid of isolated spiking neurons is capable of generating solutions to this problem very quickly, although some degree approximation is required to achieve the speedup. The approximate spiking approach presented in this work was able to complete nearly all allocation simulations with greater than 98% accuracy. Our results in this work show that constraining the possible solution space by creating specific rules for a scenario can alter the quality of the allocation result. We present a study compares allocation score and computation time for three different constraint implementation cases. Given the vast increase in speed, as well as the reduction computational requirements, the presented algorithm is ideal for moving asset allocation to low power embedded hardware.
{"title":"High Speed Approximate Cognitive Domain Ontologies for Constrained Asset Allocation based on Spiking Neurons","authors":"C. Yakopcic, Nayim Rahman, Tanvir Atahary, T. Taha, Alex Beigh, Scott Douglass","doi":"10.1109/NAECON46414.2019.9057909","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9057909","url":null,"abstract":"Cognitive agents are typically utilized in autonomous systems for automated decision making. These systems interact at real time with their environment and are generally heavily power constrained. Thus, there is a strong need for a real time agent running on a low power platform. The agent examined is the Cognitively Enhanced Complex Event Processing (CECEP) architecture. This is an autonomous decision support tool that reasons like humans and enables enhanced agent-based decision-making. It has applications in a large variety of domains including autonomous systems, operations research, intelligence analysis, and data mining. One of the most time consuming and key components of CECEP is the mining of knowledge from a repository described as a Cognitive Domain Ontology (CDO). One problem that is often tasked to CDOs is asset allocation. Given the number of possible solutions in this allocation problem, determining the optimal solution via CDO can be very time and energy consuming. A grid of isolated spiking neurons is capable of generating solutions to this problem very quickly, although some degree approximation is required to achieve the speedup. The approximate spiking approach presented in this work was able to complete nearly all allocation simulations with greater than 98% accuracy. Our results in this work show that constraining the possible solution space by creating specific rules for a scenario can alter the quality of the allocation result. We present a study compares allocation score and computation time for three different constraint implementation cases. Given the vast increase in speed, as well as the reduction computational requirements, the presented algorithm is ideal for moving asset allocation to low power embedded hardware.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126448746","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9058239
B. Alemayehu, Akash Kota, Amy T. Neidhard-Doll, V. Chodavarapu, G. Subramanyam
Dissolved gas analysis and oil sample analysis have been established as effective ways of determining the transformer oil health. In this paper, we present a new approach to diagnose the oil condition towards utility transformer health monitoring based on using the AD5933 single chip impedance analyzer from Analog Devices. We propose an integrated smart infrastructure monitoring with the results from the impedance analyzer transmitted, logged, and processed via a cloud computing interface. The transformer oils are characterized and monitored by analyzing their impedance values over a range of frequencies from 1 kHz to 100 kHz.
{"title":"Utility Transformer Health Monitoring using a Single Chip Impedance Analyzer","authors":"B. Alemayehu, Akash Kota, Amy T. Neidhard-Doll, V. Chodavarapu, G. Subramanyam","doi":"10.1109/NAECON46414.2019.9058239","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9058239","url":null,"abstract":"Dissolved gas analysis and oil sample analysis have been established as effective ways of determining the transformer oil health. In this paper, we present a new approach to diagnose the oil condition towards utility transformer health monitoring based on using the AD5933 single chip impedance analyzer from Analog Devices. We propose an integrated smart infrastructure monitoring with the results from the impedance analyzer transmitted, logged, and processed via a cloud computing interface. The transformer oils are characterized and monitored by analyzing their impedance values over a range of frequencies from 1 kHz to 100 kHz.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"156 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131778606","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9058307
Anthony Reiling, William Mitchell, Stefan Westberg, E. Balster, T. Taha
Hand tuning convolutional neural networks (CNN) for performance optimization can be tedious. A novel approach using a genetic algorithm to automate CNN hyper-parameter adjustment is proposed. This automated approach shows a 5% accuracy improvement over hand tuned methods and highly energy efficient networks on the Intel Movidius Compute Stick.
{"title":"CNN Optimization with a Genetic Algorithm","authors":"Anthony Reiling, William Mitchell, Stefan Westberg, E. Balster, T. Taha","doi":"10.1109/NAECON46414.2019.9058307","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9058307","url":null,"abstract":"Hand tuning convolutional neural networks (CNN) for performance optimization can be tedious. A novel approach using a genetic algorithm to automate CNN hyper-parameter adjustment is proposed. This automated approach shows a 5% accuracy improvement over hand tuned methods and highly energy efficient networks on the Intel Movidius Compute Stick.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"23 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127967888","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9058257
Ross D. Arnold, Benjamin Abruzzo, C. Korpela
Object classification capabilities and associated reactive swarm behaviors are implemented in a decentralized swarm of autonomous, heterogeneous unmanned aerial vehicles (UAVs). Each UAV possesses a separate capability to recognize and classify objects using the You Only Look Once (YOLO) neural network model. The UAVs communicate and share data through a swarm software architecture using an adhoc wireless network. When one UAV recognizes a particular object of interest, the entire swarm reacts with a pre-programmed behavior. Classification results of people and backpacks using our modified UAV detection platforms are provided, as well as a simulated demonstration of the reactive swarm behaviors with actual hardware and swarm software in the loop.
目标分类能力和相关的反应性群体行为是在分散的自主异构无人机群中实现的。每架无人机拥有使用You Only Look Once (YOLO)神经网络模型识别和分类物体的独立能力。无人机通过使用自组织无线网络的群软件架构进行通信和共享数据。当一架无人机识别出感兴趣的特定目标时，整个蜂群会以预先编程的行为做出反应。给出了改进后的无人机检测平台对人员和背包的分类结果，并通过实际硬件和群软件在回路中对反应性群体行为进行了仿真演示。
{"title":"Towards a Heterogeneous Swarm for Object Classification","authors":"Ross D. Arnold, Benjamin Abruzzo, C. Korpela","doi":"10.1109/NAECON46414.2019.9058257","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9058257","url":null,"abstract":"Object classification capabilities and associated reactive swarm behaviors are implemented in a decentralized swarm of autonomous, heterogeneous unmanned aerial vehicles (UAVs). Each UAV possesses a separate capability to recognize and classify objects using the You Only Look Once (YOLO) neural network model. The UAVs communicate and share data through a swarm software architecture using an adhoc wireless network. When one UAV recognizes a particular object of interest, the entire swarm reacts with a pre-programmed behavior. Classification results of people and backpacks using our modified UAV detection platforms are provided, as well as a simulated demonstration of the reactive swarm behaviors with actual hardware and swarm software in the loop.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"144 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133728876","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9058251
Hamed Elwarfalli, Dimitri Papazoglou, D. Erdahl, Amy Doll, J. Speltz
Additive Manufacturing (AM) is a growing field for various industries of avionics, biomedical, automotive and manufacturing. The onset of Laser Powder Bed Fusion (LPBF) technologies for metal printing has shown exceptional growth in the past 15 years. Quality of parts for LPBF is a concern for the industry, as many parts produced are high risk, such as biomedical implants. To address these needs, a LPBF machine was designed with in-situ sensors to monitor the build process. Image processing and machine learning algorithms provide an efficient means to take bulk data and assess part quality, validating specific internal geometries and build defects. This research will analyze infrared (IR) images from a Selective Laser Melting (SLM) machine using a Computer Aided Design (CAD) designed part, featuring specific geometries (squares, circles, and triangles) of varying sizes (0.75–3.5 mm) on multiple layers for feature detection. Applying image processing to denoise, then Principal Component Analysis (PCA) for further denoising and applying Convolution Neural Networks (CNN) to identify the features and identifying a class which does not belong to a dataset, where a dataset are created from CAD images. Through this automated process, 300 geometric elements detected, classified, and validated against the build file through CNN. In addition, several build anomalies were detected and saved for end-user inspection.
{"title":"In Situ Process Monitoring for Laser-Powder Bed Fusion using Convolutional Neural Networks and Infrared Tomography","authors":"Hamed Elwarfalli, Dimitri Papazoglou, D. Erdahl, Amy Doll, J. Speltz","doi":"10.1109/NAECON46414.2019.9058251","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9058251","url":null,"abstract":"Additive Manufacturing (AM) is a growing field for various industries of avionics, biomedical, automotive and manufacturing. The onset of Laser Powder Bed Fusion (LPBF) technologies for metal printing has shown exceptional growth in the past 15 years. Quality of parts for LPBF is a concern for the industry, as many parts produced are high risk, such as biomedical implants. To address these needs, a LPBF machine was designed with in-situ sensors to monitor the build process. Image processing and machine learning algorithms provide an efficient means to take bulk data and assess part quality, validating specific internal geometries and build defects. This research will analyze infrared (IR) images from a Selective Laser Melting (SLM) machine using a Computer Aided Design (CAD) designed part, featuring specific geometries (squares, circles, and triangles) of varying sizes (0.75–3.5 mm) on multiple layers for feature detection. Applying image processing to denoise, then Principal Component Analysis (PCA) for further denoising and applying Convolution Neural Networks (CNN) to identify the features and identifying a class which does not belong to a dataset, where a dataset are created from CAD images. Through this automated process, 300 geometric elements detected, classified, and validated against the build file through CNN. In addition, several build anomalies were detected and saved for end-user inspection.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"323 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133826740","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9057822
Hesham Alghodhaifi, Abdulmajeed Alghodhaifi, Mohammed Alghodhaifi
Over the past ten years, there has been a rise in using deep learning for medical image analysis such as CNN. Deep learning is used extensively in the field of healthcare to identify patterns, classify and segment tumors and so on. The classification of breast cancer is a well-known problem that attracts the attention of many researchers in the field of healthcare because breast cancer is the second major cause of cancer-related deaths in women. The most common subtype of all breast cancers is the Invasive Ductal Carcinoma (IDC). There are many ways to identify this type of breast cancer such as a biopsy where tissue is removed from patient and studied under microscope. The biopsy is followed by a diagnosis which is based on the qualification of the pathologists, who will look for abnormal cells. The next task for pathologists is to assign an aggressiveness grade to a whole mount sample. To do this, pathologists focus on the region of interest which contain the IDC. Therefore, one of the popular pre-processing steps for automatic aggressiveness grading is to delineate the exact regions of IDC inside of a whole mount slide. In this paper, we have experimentally tested two CNN models using depthwise separable convolution and standard convolution to enhance the accuracy of the convolutional neural network. We tested different types of activation functions such as ReLU, Sigmoid, and Tanh. As well as applying gaussian noise to test the robustness of the two models. The results show convolutional neural networks outperformed the softmax classifier, with standard convolution and ReLU where we achieved ~87.5% classification accuracy, ~93.5% sensitivity, and ~71.5% specificity.
{"title":"Predicting Invasive Ductal Carcinoma in breast histology images using Convolutional Neural Network","authors":"Hesham Alghodhaifi, Abdulmajeed Alghodhaifi, Mohammed Alghodhaifi","doi":"10.1109/NAECON46414.2019.9057822","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9057822","url":null,"abstract":"Over the past ten years, there has been a rise in using deep learning for medical image analysis such as CNN. Deep learning is used extensively in the field of healthcare to identify patterns, classify and segment tumors and so on. The classification of breast cancer is a well-known problem that attracts the attention of many researchers in the field of healthcare because breast cancer is the second major cause of cancer-related deaths in women. The most common subtype of all breast cancers is the Invasive Ductal Carcinoma (IDC). There are many ways to identify this type of breast cancer such as a biopsy where tissue is removed from patient and studied under microscope. The biopsy is followed by a diagnosis which is based on the qualification of the pathologists, who will look for abnormal cells. The next task for pathologists is to assign an aggressiveness grade to a whole mount sample. To do this, pathologists focus on the region of interest which contain the IDC. Therefore, one of the popular pre-processing steps for automatic aggressiveness grading is to delineate the exact regions of IDC inside of a whole mount slide. In this paper, we have experimentally tested two CNN models using depthwise separable convolution and standard convolution to enhance the accuracy of the convolutional neural network. We tested different types of activation functions such as ReLU, Sigmoid, and Tanh. As well as applying gaussian noise to test the robustness of the two models. The results show convolutional neural networks outperformed the softmax classifier, with standard convolution and ReLU where we achieved ~87.5% classification accuracy, ~93.5% sensitivity, and ~71.5% specificity.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132527102","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9057977
Mounica Patnala, T. Ytterdal, M. Rizkalla
In this papr, A 2-bit, 3-bit, and 4-bit DACs using newly emerged transistor technology known as Graphene Nano Ribbon Field Effect Transistor (GNRFET) technology were developed. A channel length of 10nm for the GNRFET device with supply voltage of 0.7V was incorporated in the design and simulated via ADS (Advanced Digital System) platform. Biasing with current mirror topology was used for highly efficient small size implementation. The power consumption was analyzed for all three devices. The design showed a full range linear input region within the 0.7 V supply. The signal to noise distortion ratio (SNDR) was 25.8 for the 4-bit DAC. The findings of this design conclude that the proposed DAC is more suitable for high speed nano electromechanical systems (NEMs), computer architecture and memory cells, among other applications.
本文采用新型晶体管技术石墨烯纳米带场效应晶体管(GNRFET)技术开发了2位、3位和4位dac。设计中引入了电源电压为0.7V、通道长度为10nm的GNRFET器件，并通过ADS (Advanced Digital System)平台进行了仿真。采用电流镜像拓扑进行偏置，实现了高效率的小尺寸实现。对这三种设备的功耗进行了分析。该设计显示了0.7 V电源内的全范围线性输入区域。4位DAC的信噪比(SNDR)为25.8。本设计的研究结果表明，所提出的DAC更适合高速纳米机电系统(nem)、计算机体系结构和存储单元等应用。
{"title":"High Speed-Low Power GNRFET based Digital to Analog Converters for ULSI applications","authors":"Mounica Patnala, T. Ytterdal, M. Rizkalla","doi":"10.1109/NAECON46414.2019.9057977","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9057977","url":null,"abstract":"In this papr, A 2-bit, 3-bit, and 4-bit DACs using newly emerged transistor technology known as Graphene Nano Ribbon Field Effect Transistor (GNRFET) technology were developed. A channel length of 10nm for the GNRFET device with supply voltage of 0.7V was incorporated in the design and simulated via ADS (Advanced Digital System) platform. Biasing with current mirror topology was used for highly efficient small size implementation. The power consumption was analyzed for all three devices. The design showed a full range linear input region within the 0.7 V supply. The signal to noise distortion ratio (SNDR) was 25.8 for the 4-bit DAC. The findings of this design conclude that the proposed DAC is more suitable for high speed nano electromechanical systems (NEMs), computer architecture and memory cells, among other applications.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"2008 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125612934","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9058319
Dajung Lee, Colman Cheung, Dan Pritsker
Radar-based object detection becomes a more important problem as such sensor technology is broadly adopted in many applications including military, robotics, space exploring, and autonomous vehicles. However, the existing solution for classification of radar echo signal is limited because its deterministic analysis is too complicated to describe various object features. It needs a more sophisticated approach that is capable of identifying them. In this paper, we intend to solve this problem using a state-of-art machine learning approach to read object features or patterns in their micro-Doppler signatures in radar reflection data. In spectrogram analysis, we observe unique patterns of objects, which should be recognizable by a well-trained machine learning algorithm. We train our AlexNet-inspired convolutional neural network model to see these patterns over their radar signal spectrogram and design an intelligent waveform detection system. We demonstrate our proposed system on an Intel® Xeon CPU and an Intel® Arria 10 FPGA using Intel® Open VINO toolkit, a unified framework to import deep learning algorithms in different platforms and achieve a real-time system for automated object classification with over 90% of accuracy on a given radar dataset.
{"title":"Radar-based Object Classification Using An Artificial Neural Network","authors":"Dajung Lee, Colman Cheung, Dan Pritsker","doi":"10.1109/NAECON46414.2019.9058319","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9058319","url":null,"abstract":"Radar-based object detection becomes a more important problem as such sensor technology is broadly adopted in many applications including military, robotics, space exploring, and autonomous vehicles. However, the existing solution for classification of radar echo signal is limited because its deterministic analysis is too complicated to describe various object features. It needs a more sophisticated approach that is capable of identifying them. In this paper, we intend to solve this problem using a state-of-art machine learning approach to read object features or patterns in their micro-Doppler signatures in radar reflection data. In spectrogram analysis, we observe unique patterns of objects, which should be recognizable by a well-trained machine learning algorithm. We train our AlexNet-inspired convolutional neural network model to see these patterns over their radar signal spectrogram and design an intelligent waveform detection system. We demonstrate our proposed system on an Intel® Xeon CPU and an Intel® Arria 10 FPGA using Intel® Open VINO toolkit, a unified framework to import deep learning algorithms in different platforms and achieve a real-time system for automated object classification with over 90% of accuracy on a given radar dataset.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130769058","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9057988
Dewant Katare, M. El-Sharkawy
Present day autonomous vehicle relies on several sensor technologies for it’s autonomous functionality. The sensors based on their type and mounted-location on the vehicle, can be categorized as: line of sight and non-line of sight sensors and are responsible for the different level of autonomy. These line of sight sensors are used for the execution of actions related to localization, object detection and the complete environment understanding. The surrounding or environment understanding for an autonomous vehicle can be achieved by segmentation. Several traditional and deep learning related techniques providing semantic segmentation for an input from camera is already available, however with the advancement in the computing processor, the progression is on developing the deep learning application replacing traditional methods. This paper presents an approach to combine the input of camera and lidar for semantic segmentation purpose. The proposed model for outdoor scene segmentation is based on the frustum pointnet, and ResNet which utilizes the 3d point cloud and camera input for the 3d bounding box prediction across the moving and non-moving object and thus finally recognizing and understanding the scenario at the point-cloud or pixel level. For real time application the model is deployed on the RTMaps framework with Bluebox (an embedded platform for autonomous vehicle). The proposed architecture is trained with the CITYScpaes and the KITTI dataset.
{"title":"Real-Time 3-D Segmentation on An Autonomous Embedded System: using Point Cloud and Camera","authors":"Dewant Katare, M. El-Sharkawy","doi":"10.1109/NAECON46414.2019.9057988","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9057988","url":null,"abstract":"Present day autonomous vehicle relies on several sensor technologies for it’s autonomous functionality. The sensors based on their type and mounted-location on the vehicle, can be categorized as: line of sight and non-line of sight sensors and are responsible for the different level of autonomy. These line of sight sensors are used for the execution of actions related to localization, object detection and the complete environment understanding. The surrounding or environment understanding for an autonomous vehicle can be achieved by segmentation. Several traditional and deep learning related techniques providing semantic segmentation for an input from camera is already available, however with the advancement in the computing processor, the progression is on developing the deep learning application replacing traditional methods. This paper presents an approach to combine the input of camera and lidar for semantic segmentation purpose. The proposed model for outdoor scene segmentation is based on the frustum pointnet, and ResNet which utilizes the 3d point cloud and camera input for the 3d bounding box prediction across the moving and non-moving object and thus finally recognizing and understanding the scenario at the point-cloud or pixel level. For real time application the model is deployed on the RTMaps framework with Bluebox (an embedded platform for autonomous vehicle). The proposed architecture is trained with the CITYScpaes and the KITTI dataset.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132407118","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 : 2019-07-01DOI: 10.1109/NAECON46414.2019.9058210
Ben Purman, J. Messing, J. Crossman
We present the development of a cognitive agent for real-time, wide-area search applications. Wide-area search problems present specific challenges driven by limited resources, large search areas, and limited time to conduct a search or inspection. Cognitive agents present an opportunity to incorporate a range of reasoning and sensor processing approaches to more effectively focus attention and make decisions about how to interpret data.We developed a design for a cognitive agent and implemented supporting reasoning algorithms to interact with sensor data. The agent encodes knowledge about objects of interest, and how they present themselves in the environment. This allows object detection algorithms to focus on detecting single objects, keeping training data requirements manageable. The cognitive agent provides external reasoning to reduce false alarm rates and make additional inferences. In this paper, we describe the cognitive agent design, conduct feasibility studies to establish reasoning strategies, and identify areas for future agent contributions.
{"title":"Toward the Development of a Cognitive Agent for Wide-Area Search","authors":"Ben Purman, J. Messing, J. Crossman","doi":"10.1109/NAECON46414.2019.9058210","DOIUrl":"https://doi.org/10.1109/NAECON46414.2019.9058210","url":null,"abstract":"We present the development of a cognitive agent for real-time, wide-area search applications. Wide-area search problems present specific challenges driven by limited resources, large search areas, and limited time to conduct a search or inspection. Cognitive agents present an opportunity to incorporate a range of reasoning and sensor processing approaches to more effectively focus attention and make decisions about how to interpret data.We developed a design for a cognitive agent and implemented supporting reasoning algorithms to interact with sensor data. The agent encodes knowledge about objects of interest, and how they present themselves in the environment. This allows object detection algorithms to focus on detecting single objects, keeping training data requirements manageable. The cognitive agent provides external reasoning to reduce false alarm rates and make additional inferences. In this paper, we describe the cognitive agent design, conduct feasibility studies to establish reasoning strategies, and identify areas for future agent contributions.","PeriodicalId":193529,"journal":{"name":"2019 IEEE National Aerospace and Electronics Conference (NAECON)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-07-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128302570","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：