Search

Recent years have shown that deep learned neural networks are a valuable tool in the field of computer vision. This paper addresses the use of two different kinds of network architectures, namely LeNet and Network in Network (NiN). They will be compared in terms of both performance and computational efficiency by addressing the classification and detection problems. In this paper, multiple databases will be used to test the networks. One of them contains images depicting burn wounds from pediatric cases, another one contains an extensive number of art images and other facial databases were used for facial keypoints detection.

The ESO/MPG WFI and the INT WFC wide field archives comprising 330,000 images were mined to search for serendipitous encounters of known Near Earth Asteroids (NEAs) and Potentially Hazardous Asteroids (PHAs). A total of 152 asteroids (44 PHAs and 108 other NEAs) were identified using the PRECOVERY software, their astrometry being measured on 761 images and sent to the Minor Planet Centre. Both recoveries and precoveries were reported, including prolonged orbital arcs for 18 precovered objects and 10 recoveries. We analyze all new opposition data by comparing the orbits fitted before and after including our contributions. We conclude the paper presenting Mega-Precovery, a new online service focused on data mining of many instrument archives simultaneously for one or a few given asteroids. A total of 28 instrument archives have been made available for mining using this tool, adding together about 2.5 million images forming the Mega-Archive., Comment: Accepted for publication in Astronomische Nachrichten (Sep 2012)

We report on the follow-up and recovery of 100 program NEAs, PHAs and VIs using the ESO/MPG 2.2m, Swope 1m and INT 2.5m telescopes equipped with large field cameras. The 127 fields observed during 11 nights covered 29 square degrees. Using these data, we present the incidental survey work which includes 558 known MBAs and 628 unknown moving objects mostly consistent with MBAs from which 58 objects became official discoveries. We planned the runs using six criteria and four servers which focus mostly on faint and poorly observed objects in need of confirmation, follow-up and recovery. We followed 62 faint NEAs within one month after discovery and we recovered 10 faint NEAs having big uncertainties at their second or later opposition. Using the INT we eliminated 4 PHA candidates and VIs. We observed in total 1,286 moving objects and we reported more than 10,000 positions. All data were reduced by the members of our network in a team effort, and reported promptly to the MPC. The positions of the program NEAs were published in 27 MPC and MPEC references and used to improve their orbits. The O-C residuals for known MBAs and program NEAs are smallest for the ESO/MPG and Swope and about four times larger for the INT whose field is more distorted. The incidental survey allowed us to study statistics of the MBA and NEA populations observable today with 1--2m facilities. We calculate preliminary orbits for all unknown objects, classifying them as official discoveries, later identifications and unknown outstanding objects. The orbital elements a, e, i calculated by FIND_ORB software for the official discoveries and later identified objects are very similar with the published elements which take into account longer observational arcs; thus preliminary orbits were used in statistics for the whole unknown dataset. (CONTINUED), Comment: Accepted in Planetary and Space Science (Aug 2011)

The Canada-France-Hawaii Legacy Survey (CFHTLS) comprising about 25 000 MegaCam images was data mined to search for serendipitous encounters of known Near Earth Asteroids (NEAs) and Potentially Hazardous Asteroids (PHAs). A total of 143 asteroids (109 NEAs and 34 PHAs) were found on 508 candidate images which were field corrected and measured carefully, and their astrometry was reported to Minor Planet Centre. Both recoveries and precoveries (apparitions before discovery) were reported, including data for 27 precovered asteroids (20 NEAs and 7 PHAs) and 116 recovered asteroids (89 NEAs and 27 PHAs). Our data prolonged arcs for 41 orbits at first or last opposition, refined 35 orbits by fitting data taken at one new opposition, recovered 6 NEAs at their second opposition and allowed us to ameliorate most orbits and their Minimal Orbital Intersection Distance (MOID), an important parameter to monitor for potential Earth impact hazard in the future., Comment: Accepted in Astron. Nachr. (10 pags)