Distributed and consistent multi-image feature matching via QuickMatch

In this work, we consider the multi-image object matching problem in distributed networks of robots. Multi-image feature matching is a keystone of many applications, including Simultaneous Localization and Mapping, homography, object detection, and Structure from Motion. We first review the QuickMatch algorithm for multi-image feature matching. We then present NetMatch, an algorithm for distributing sets of features across computational units (agents) that largely preserves feature match quality and minimizes communication between agents (avoiding, in particular, the need to flood all data to all agents). Finally, we present an experimental application of both QuickMatch and NetMatch on an object matching test with low-quality images. The QuickMatch and NetMatch algorithms are compared with other standard matching algorithms in terms of preservation of match consistency. Our experiments show that QuickMatch and Netmatch can scale to larger numbers of images and features, and match more accurately than standard techniques.

Identifying NCAA tournament upsets using Balance Optimization Subset Selection

The NCAA basketball tournament attracts over 60 million people who fill out a bracket to try to predict the outcome of every tournament game correctly. Predictions are often made on the basis of instinct, statistics, or a combination of the two. This paper proposes a technique to select round-of-64 upsets in the tournament using a Balance Optimization Subset Selection model. The model determines which games feature match-ups that are statistically most similar to the match-ups in historical upsets. The technique is then applied to the tournament in each of the 13 years from 2003 to 2015 in order to select two games as potential upsets each year. Of the 26 selected games, 10 (38.4%) were actual upsets, which is more than twice as many as the expected number of correct selections when using a weighted random selection method.

Methods for Comparing Simulated and Observed Satellite Infrared Brightness Temperatures and What Do They Tell Us?

In this study, the utility of dimensioned, neighborhood-based, and object-based forecast verification metrics for cloud verification is assessed using output from the experimental High Resolution Rapid Refresh (HRRRx) model over a 1-day period containing different modes of convection. This is accomplished by comparing observed and simulated Geostationary Operational Environmental Satellite (GOES) 10.7-μm brightness temperatures (BTs). Traditional dimensioned metrics such as mean absolute error (MAE) and mean bias error (MBE) were used to assess the overall model accuracy. The MBE showed that the HRRRx BTs for forecast hours 0 and 1 are too warm compared with the observations, indicating a lack of cloud cover, but rapidly become too cold in subsequent hours because of the generation of excessive upper-level cloudiness. Neighborhood and object-based statistics were used to investigate the source of the HRRRx cloud cover errors. The neighborhood statistic fractions skill score (FSS) showed that displacement errors between cloud objects identified in the HRRRx and GOES BTs increased with time. Combined with the MBE, the FSS distinguished when changes in MAE were due to differences in the HRRRx BT bias or displacement in cloud features. The Method for Object-Based Diagnostic Evaluation (MODE) analyzed the similarity between HRRRx and GOES cloud features in shape and location. The similarity was summarized using the newly defined MODE composite score (MCS), an area-weighted calculation using the cloud feature match value from MODE. Combined with the FSS, the MCS indicated if HRRRx forecast error is the result of cloud shape, since the MCS is moderately large when forecast and observation objects are similar in size.