Semantic Geometric Features: A Preliminary Investigation of ...

Semantic Geometric Features: A Preliminary Investigation of ...

Semantic Geometric Features: A Preliminary Investigation of Automobile Identification Carl E. Abrams Sung-Hyuk Cha, Michael Gargano, and Charles Tappert Pace DPS Agenda Overview of the Problem The Experiments Results Going Forward Pace DPS Overview Object recognition remains a hard problem The human mind uses shapes to recognize objects Can semantic features defined by their shapes be more effective in the recognition and identification of objects? Pace DPS The Experiments 10 test images of cars Directly form the manufactures websites Images were restricted to side views of the cars taken from 90 degrees All 2005 models Feature vectors calculated/measured from the images Pace DPS The Vehicles Honda Accord Sedan 2005 Honda Civic Coupe 2005 Mazda 3 2005 Mazda 6 2005 Porsche Carrera Toyota Camry 2005 Toyota Corolla CE 2005 Toyota Celica GT 2005 Toyota Echo 2005

VW Passat 2005 Pace DPS Experiments used Euclidean Distance as the Measure L d (x t ) i 2 i i 1 the xi and ti are measurements from two different vehicles Pace DPS Experiments used Euclidean Distance as the Measure (x2,y2) c (x1,y1) b a c = (a2+b2)1/2 c = ((x1-x2)2+(y1-y2)2)1/2 Pace DPS Manufacturers Specifications First Experiment Vehicle Description Wheelbase in inchesWidth Honda Accord Civic Coupe 2005 103.1 66.7 Honda Accord Sedan 2005 107.9 71.5 Mazda 3 2005

103.9 69.9 Mazda 6 2005 105.3 70.1 Porsche Carerra 2005 92.52 71.18 Toyota Camry 2005 107.1 70.7 Toyota Celica GT 2005 102.4 68.3 Toyota Corolla CE 2005 102.4 66.9 Toyota Echo 2005 92.3 65.4 VW Passat 2005 106.4 68.7 Pace DPS Length 175.4 189.5 178.3 186.8 174.29 189.2 170.5 178.3 164.8 185.2 Height 55.1 57.1 57.7 56.7 51.57 58.7 51.4 58.5 59.4 57.6 Boundary Description using Rays Second Experiment Vehicle Description

Honda Accord Civic Coupe 2005 Honda Accord Sedan 2005 Mazda 3 2005 Mazda 6 2005 Mercedes Benz C230 Sports Coupe 2005 Nissan Altima 2005 Porsche Carerra 2005 Toyota Camry 2005 Toyota Celica GT 2005 Toyota Corolla CE 2005 Toyota Echo 2005 VW Passat 2005 Wheelbase in mm 148 139 146 140 138 136 128 136 134 144 154 145 Ray length in mm at angles in degress 5 10 15 20 125 121 123 110 120 116 118 110 118 113 115 113 118 104 115 109 102 115 112 115 115 110

110 110 112 112 115 100 126 127 126 120 102 104 111 113 119 117 120 113 125 124 127 112 122 120 120 110 Pace DPS 25 95 98 100 95 100 91 90 101 87 96 113 96 30 90 90 94 88 94 86 82 94 82 93

108 90 35 86 85 89 84 89 81 76 88 75 89 100 86 40 81 80 84 80 82 76 71 86 68 84 99 82 45 78 78 80 76 79 71 67 73 65 82 96 74 50 75 73 78 72 73 66 66 73

63 77 92 72 55 70 70 71 69 71 63 62 71 60 73 87 69 60 66 67 67 63 67 62 60 68 57 69 84 66 Semantic Features Third Experiment Vehicle Description Honda Accord Civic Coupe 2005 Honda Accord Sedan 2005 Mazda 3 2005 Mazda 6 2005 Mercedes Benz C230 Sports Coupe 2005 Wheelbase in mm Angle FD Length FD Angle RD Length RD Angle Mirror Length Mirror 148 123

33 0 0 51 50 139 146 140 98 92 93 28 30 29 154 150 151 63 58 64 51 51 47 47 50 48 138 126 37 0 0 52 43 Nissan Altima 2005

Porsche Carerra 2005 136 98 25 153 58 47 45 128 144 33 0 0 69 34 Toyota Camry 2005 Toyota Celica GT 2005 Toyota Corolla CE 2005 Toyota Echo 2005 VW Passat 2005 136 85 24 151 62 42 53 134

125 31 0 0 49 44 144 154 145 90 97 97 34 34 36 148 0 154 63 0 59 42 49 47 58 65 50 Pace DPS Challenge: Determine the qualitative ability of the feature vectors to separate the vehicles Within each experiment compute the distance of each vehicle from all the others Evenly divide the measures into 5 bins Observe the distribution of the measures Pace DPS The Results

50 45 40 35 30 25 20 15 10 5 0 HWD Semantic Rays Bin1 Bin2 Bin3 Bin4 Bin5 Pace DPS Distance Matrix Semantic Features Honda Civic Honda Accord Mazda 3 Mazda 6 Porsche Carerra Toyota Camry Toyota Celica Toyota Corolla Toyota Echo VW Passat Honda Civic 0

156.02 153.17 154.00 27.66 155.96 2.82 151.90 26.07 156.23 Honda Accord 156.02 0 7.21 7.07 161.72 16.09 156.36 13.45 154.01 4.12 153.17 7.21 0 4.24 159.78 11.44 153.60

9.43 150.09 7.55 154.00 7.07 4.24 0 160.89 9.43 154.37 6.55 151.06 5.00 Porsche Carerra 27.65 161.73 159.77 160.89 0 164/35 27.58 159.84 51.08 162.50 Toyota Camry 155.96

16.09 11.44 9.43 164.35 0 156.36 5.83 151.63 13.34 Toyota Celica 2.82 156.36 153.60 154.36 27.58 156.36 0 152.24 28.00 156.53 Toyota Corolla 151.90 13,45 9.43 6.55 159.84

5.83 152.24 0 148.33 10.48 Toyota Echo 26.07 154.01 150.09 151.06 51.07 151.63 28.00 148.33 0 154.01 VW Passat 156.23 4.12 7.55 5.00 162.50 13.34 156.53 10.48 154.01

0 Mazda 3 Mazda 6 Pace DPS Going Forward Extend techniques to encompass semantic shapes within an object (shape contexts) Compare the extended semantic methods to existing methods in multiple domains Pace DPS Going Forward Shape Contexts Pace DPS References [1] R. D. Acqua and R. Job, "Is global shape sufficient for automatic object identification?" Congitive Science, vol. 8, pp. 801821, 2001. [2] A. K. Jain, A. Ross, and S. Pankanti, "A Prototype Hand Geomtery-based Verification System," presented at Proceedings of 2nd International conference on Audio and Video-based Biometric Person Authentication, Wahington D.C., 1999. [3] H. Schneiderman and T. Kanade, "A Statistical Model for 3D Object Detection Applied to Faces and Cars," presented at IEEE Conference on Computer Vision and Pattern Recognition, IEEE, 2000 [4] S. Belongie,J Malik, J Puzicha, Matching Shapes ,presented at the International Conference on Computer Vision (ICCV 01) Vol 1, Jan 2001 Pace DPS

Recently Viewed Presentations

  • Peer-Led Team Learning in General Chemistry Melissa Otis

    Peer-Led Team Learning in General Chemistry Melissa Otis

    Faculty Advisor: Dr. Chris Bauer. Department of Chemistry, University of New Hampshire. Peer-Led Team Learning in General Chemistry. Background. History and Structure of PLTL at UNH. Peer-Led Team Learning (PLTL) is a nationally recognized instructional model in which students engage...
  • Sophocles and Antigone - English 4

    Sophocles and Antigone - English 4

    SOPHOCLES AND ANTIGONE A Look at Ancient Greek Tragedy ANCIENT GREEK THEATER A Review PERFORMANCE OF GREEK TRAGEDY Set in an outdoor auditorium or amphitheater, and, because of the lack of artificial light, set mostly in the daytime.
  • Maya Angelou - Ram Pages

    Maya Angelou - Ram Pages

    THE FUDGE BOOKS. Tales of a fourth grade nothing (1972) Otherwise known as Sheila the great (1972) Superfudge(1980) Fudge-a-mania (1990) Double fudge (2002) Imagination and personal experience base for stories. Received BS in Education from NYU in 1961.
  • Agenda - National Oceanic and Atmospheric Administration

    Agenda - National Oceanic and Atmospheric Administration

    The current BPSA LUT implemented in CLASS doesn't consider surface BRDF. This leads to temporal variability of LSA retrievals over stable surfaces. The quality of BPSA retrievals rely on the accurate detection of cloud and cloud shadow. Undetected cloud/shadow will...
  • Mapping Capability and Connections Northern Tasmania Manufacturing Sarah

    Mapping Capability and Connections Northern Tasmania Manufacturing Sarah

    Data analysts Doggers/riggers training Engineers (experienced & graduates) HV Specialists ICT skills Industrial blasting and spraying Lean/continuous improvement Production/process workers Supervisory/team leadership skills Technical (equipment) specialists Trades (automation,mech.) Trades (electrical) Trades (other) 2 1 3 1 1 1 3 1...
  • INSOLVENCY AND BANKRUPTCY CODE, 2016 1 The Insolvency

    INSOLVENCY AND BANKRUPTCY CODE, 2016 1 The Insolvency

    The non-obstante clause contained in IBC, which is a later statute shall prevail over the non-obstante clause contained in Section 71 of PMLA and the proceedings before the Adjudicating Authority under PMLA is civil in nature and hence, in view...
  • Evolution  Darwin Studied finches in the Galapagos Islands

    Evolution Darwin Studied finches in the Galapagos Islands

    Kingdom. Phylum. Class. Order. Family . Genus. Species . BROADEST TAXON. Know in order, how the taxa are arranged from broadest to most specific. Genus and species are the two names used to identify specific organisms in the binomial system...
  • Teen Drinking Survey

    Teen Drinking Survey

    Have you ever been in the car with someone who was drinking and driving? Do you feel in your high school environment that more kids are drinking than staying above the influence? Do you feel that in school you are...