Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery最新文献

{"title":"Combining Satellite Imagery and GPS Data for Road Extraction","authors":"Tao Sun, Zonglin Di, Yin Wang","doi":"10.1145/3281548.3281550","DOIUrl":"https://doi.org/10.1145/3281548.3281550","url":null,"abstract":"Road extraction is a fundamental problem in remote sensing and mapping. Recent advances in Convolution Neural Network (CNN) have contributed significant improvements in automatic road extraction from satellite imagery, albeit prediction gaps challenge post-processing. Some of the gaps are hard to bridge by satellite imagery alone due to dense vegetation, road construction, and building shadows. In this paper, we combine satellite imagery with GPS data to improve road extraction quality. Our dataset includes 100cm pixel resolution satellite imagery and 192-hour taxi GPS traces from the urban area of Beijing. Experimenting with various layers to combine GPS data, our CNN model outperforms the RGB-only model by nearly 13% on mean IoU.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"49 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123901823","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Enhancing Trip Distribution Prediction with Twitter Data: Comparison of Neural Network and Gravity Models","authors":"Nastaran Pourebrahim, Selima Sultana, J. Thill, S. Mohanty","doi":"10.1145/3281548.3281555","DOIUrl":"https://doi.org/10.1145/3281548.3281555","url":null,"abstract":"Predicting human mobility within cities is an important task in urban and transportation planning. With the vast amount of digital traces available through social media platforms, we investigate the potential application of such data in predicting commuter trip distribution at small spatial scale. We develop back propagation (BP) neural network and gravity models using both traditional and Twitter data in New York City to explore their performance and compare the results. Our results suggest the potential of using social media data in transportation modeling to improve the prediction accuracy. Adding Twitter data to both models improved the performance with a slight decrease in root mean square error (RMSE) and an increase in R-squared (R2) value. The findings indicate that the traditional gravity models outperform neural networks in terms of having lower RMSE. However, the R2 results show higher values for neural networks suggesting a better fit between the real and predicted outputs. Given the complex nature of transportation networks and different reasons for limited performance of neural networks with the data, we conclude that more research is needed to explore the performance of such models with additional inputs.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"13 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123586859","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Aconcagua","authors":"Karima Elgarroussi, Sujing Wang, Romita Banerjee, C. Eick","doi":"10.1145/3281548.3281552","DOIUrl":"https://doi.org/10.1145/3281548.3281552","url":null,"abstract":"In this paper, we introduce Aconcagua, a novel spatio-temporal emotion change analysis framework. Our current research uses Twitter tweets as the knowledge source for emotion analysis. The inputs for the emotion mapping and change analysis system, we are currently developing, are the location and time of the tweets and their corresponding emotion assessment score falling in the range [-1, +1], with +1 representing a very positive emotion and -1 representing a very negative emotion. We start by identifying spatial clusters that capture positive and negative emotion regions for batches of the dataset with each batch corresponding to a specific time interval, e.g. a single day. These obtained spatial clusters and their statistical summaries are then used as the input for Aconcagua which monitors change of emotions with respect to a set of unary and binary change predicates that are evaluated with respect to the set of spatial clusters; as the result of this process an emotion change graph is obtained whose nodes are spatial clusters and whose edges capture different types of temporal relationships between spatial clusters. An implementation of the change monitoring process is discussed which operates on top of a relational database and uses SQL queries to specify change predicates. To obtain more aggregated change summaries and ultimately change stories, the change graph further must be mined and summarized based on what aspects of change the analyst is interested in. To support such capabilities, our approach supports several types of change analysis templates called story types. We demo our approach using tweets collected in the state of New York in June 2014.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"69 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129613001","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"How Good is Good Enough?: Quantifying the Effects of Training Set Quality","authors":"B. Swan, M. Laverdiere, Hsiuhan Lexie Yang","doi":"10.1145/3281548.3281557","DOIUrl":"https://doi.org/10.1145/3281548.3281557","url":null,"abstract":"There is a general consensus in the neural network community that noise in training data has a negative impact on model output; however, efforts to quantify the impact of varying levels have been limited, particularly for semantic segmentation tasks. This is a question of particular importance for remote sensing applications where the cost of producing a large training set can lead to reliance on publicly available data with varying degrees of noise. This work explores the effects of different degrees and types of training label noise on a pre-trained building extraction deep learner. Quantitative and qualitative evaluations of these effects can help inform decisions about trade-offs between the cost of producing training data and the quality of model outputs. We found that, relative to the base model, models trained with small amounts of noise showed little change in precision but achieved considerable increases in recall. Conversely, as noise levels increased, both precision and recall decreased. Precision and recall both lagged behind a model trained with pristine data. These exploratory results indicate the importance of quality control for training and, more broadly, that the relationship between degrees and types of training data noise and model performance is more complex than trade-offs between precision and recall.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"19 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126766293","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vision-based UAVs Aerial Image Localization: A Survey","authors":"Yingxiao Xu, Long Pan, C. Du, Jun Li, N. Jing, Jiangjiang Wu","doi":"10.1145/3281548.3281556","DOIUrl":"https://doi.org/10.1145/3281548.3281556","url":null,"abstract":"Unmanned aerial vehicles (UAVs) have been increasingly used in earth observation, public safety, military and civilian applications due to its portability, high mobility and flexibility. In some GPS-denied environments, accurate drone position cannot be obtained due to occlusion, multi-path interference and other factors. While understanding and localization the content of the images is vital for earth observation, map revision, multi-source image fusion, disaster relief, smart city and other applications. The progress of computer vision and convolutional neural networks(CNNs) in image processing provide a promising solution to locate UAVs aerial image and mapping to the large-scale reference image. Firstly, key localization techniques based on image retrieval-----image description, image matching and position mapping are summarized considering the characteristics of UAVs aerial images. And then, image localization based on extracting deep semantic features and image localization based on classification method by subdividing areas are recommended. Throughout this paper, we will have an insight into the prospect of the UAVs image localization and the challenges to be faced.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"119 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133732168","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"SKATER-CON: Unsupervised Regionalization via Stochastic Tree Partitioning within a Consensus Framework Using Random Spanning Trees: Research Paper","authors":"Orhun Aydin, Mark V. Janikas, R. Assunção, Ting-Hwan Lee","doi":"10.1145/3281548.3281554","DOIUrl":"https://doi.org/10.1145/3281548.3281554","url":null,"abstract":"Spatially constrained clustering, also known as regionalization, aims to group spatial objects into spatially contiguous clusters also known as regions. Among different approaches, tree-based partitioning is reported to define homogeneous regions rigorously, without ad-hoc adjustments, in a computationally efficient manner. One of the shortcomings of tree-based partitioning is the so-called chaining problem that results in sub-optimal regions. We propose a consensus-based regionalization approach to address the chaining problem associated with a single tree, in particular the minimum spanning tree, by exploring a wide range of partitions via a set of random spanning trees (RST). We propose an algorithm, namely SKATER-CON, that partitions spatial data via a consensus-based framework from an ensemble of regionalizations defined by its deterministic counter-part, the SKATER algorithm applied along stochastic search paths defined by RSTs. SKATER-CON utilizes evidence accumulation to represent an ensemble of regionalizations as a similarity graph. The similarity graph represents spatial objects as vertexes and frequency at which objects are assigned to the same region in the ensemble as edge weights. Proposed algorithm determines consensus among different regionalization by partitioning the similarity graph using a multi-level graph partitioning algorithm (METIS). Spatial constraints are imposed on the similarity graph prior to partitioning to ensure spatial constraints are reflected in the consensus result. We rigorously test the quality of regions produced by SKATER-CON on a large, synthetically generated dataset. The synthetic dataset is the result of full-factorial experiments designed on number, fuzziness, geometry and size of regions. Same dataset is also used compare our approach against state-of-the-art regionalization algorithms (SKATER and ARISEL). Lastly, we show the value added by SKATER-CON compared to SKATER on a real-world dataset based on Ecological Marine Units (EMU) dataset.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"168 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122045707","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Vehicle Point of Interest Detection Using In-Car Data","authors":"James Van Hinsbergh, N. Griffiths, Phillip Taylor, Alasdair Thomason, Zhou Xu, A. Mouzakitis","doi":"10.1145/3281548.3281549","DOIUrl":"https://doi.org/10.1145/3281548.3281549","url":null,"abstract":"Intelligent transportation systems often identify and make use of locations extracted from GPS trajectories to make informed decisions. However, many of the locations identified by existing systems are false positives, such as those in heavy traffic. Signals from the vehicle, such as speed and seatbelt status, can be used to identify these false positives. In this paper, we (i) demonstrate the utility of the Gradient-based Visit Extractor (GVE) in the automotive domain, (ii) propose a classification stage for removing false positives from the location extraction process, and (iii) evaluate the effectiveness of these techniques in a high resolution vehicular dataset.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"405 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125170840","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multilabel Building Functions Classification from Ground Pictures using Convolutional Neural Networks","authors":"Shivangi Srivastava, John E. Vargas-Muñoz, David Swinkels, D. Tuia","doi":"10.1145/3281548.3281559","DOIUrl":"https://doi.org/10.1145/3281548.3281559","url":null,"abstract":"We approach the problem of multi building function classification for buildings from the city of Amsterdam using a collection of Google Street View (GSV) pictures acquired at multiple zoom levels (field of views, FoV) and the corresponding governmental census data per building. Since buildings can have multiple usages, we cast the problem as multilabel classification task. To do so, we trained a CNN model end-to-end with the task of predicting multiple co-occurring building function classes per building. We fuse the individual features of three FoVs by using volumetric stacking. Our proposed model outperforms baseline CNN models that use either single or multiple FoVs.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126438924","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"When GeoAI Meets the Crowd","authors":"T. E. Chow","doi":"10.1145/3281548.3281551","DOIUrl":"https://doi.org/10.1145/3281548.3281551","url":null,"abstract":"Estimating a moving crowd, such as the head count of a presidential inauguration or a football game, presents a practical and intellectual challenge that is often politically and emotionally charged. The objectives of this paper are to discuss the integration of artificial intelligence and agent-based model (ABM) to simulate and estimate a moving crowd and outline some key issues and research agenda. To simulate individual movements of a moving crowd, Genetic Algorithm (GA) can be employed to fine-tune agent parameters in wayfinding (e.g. direction, speed, etc.) through mutation, crossover, elitism and extinction. Besides individual-based wayfinding parameters, GA can also be employed to optimize population-wide model parameters as well, such as the maximum walking speed, maximum crowd capacity, early departure and late arrival rates. These individual and global model parameters present different bottom-up and top-down forces in shaping and precipitating diverse crowd behaviors and movements to match empirical pattern. Besides spatial optimization, convolutional NN can also be trained to derive snapshots of crowd count and crowd density from still-frame pictures and videos to better provide feedbacks to the fitness function of GA. However, more researches are needed to better understand and overcome various technical issues in crowd simulation, including but not limited to overtraining in optimization, feature extraction of objects moving in multi- and random directions, ontological separation of protesters from pedestrians and spectators, reconciliation of a single/multiple crowds over time and space.","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131323874","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","authors":"","doi":"10.1145/3281548","DOIUrl":"https://doi.org/10.1145/3281548","url":null,"abstract":"","PeriodicalId":231184,"journal":{"name":"Proceedings of the 2nd ACM SIGSPATIAL International Workshop on AI for Geographic Knowledge Discovery","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2018-11-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128209015","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}