[Studies in Computational Intelligence 481] Artur Babiarz, Robert Bieda, Karol Jędrasiak, Aleksander Nawrat (auth.), Aleksander Nawrat, Zygmunt Kuś (eds.) - Vision Based Systemsfor UAV Applications (2013, Sprin

Recommendations

Info

268 H. Josiński et al. selected attributes to number of attributes in an input space. It can be noticed that greedy hill climbing gives much more compact descriptors with lower number of features in comparison to genetic search. For instance in case of variation cover Q=0.50 which gives 504 MPCA components, greedy hill climbing selects 7 and 34 features with wrapper and CFS evaluators respectively and genetic search selects 194 and 78 components. The difference rises with increasing variation cover Q. In case of Q=0.80 and 10773 MPCA components greedy hill climbing selects 10 and 240 features and genetic search selects 1746 and 1903 components. It is so because of the complexity of considered selection problem. The input set is very high dimensional and discriminative properties are scattered across MPCA components. It is extremely difficult to improve the classification accuracy by adding a single attribute, because only specified subsets of features have distinctive abilities. It is a reason why greedy hill climbing with wrapper subset evaluator terminates at first local extreme and it not able to discover efficiently individual gait features. Fig. 9. Compression rates However much more important assessment of selected MPCA components gives classification accuracy presented in Fig. 10. It shows percentage of correctly classified gaits of a test set in respect to MPCA variation cover Q and applied selection strategy. Globally best precision is 95%, obtained by 449 attributes selected from the input set containing 29400 MPCA components for Q=0.87, which means 0.03% of compression rate, by hill climbing search and CFS evaluator. Genetic search with CFS evaluator is very similar. It has 92.5% of classification precision and it has very similar dependency on a number of features of an input space. Much worse results gives wrapper evaluator, especially if combined with hill climbing search, which is caused by the same reason as explained in compression rates analysis.
Selection of Individual Gait Features Extracted by MPCA 269 Fig. 10. Selection rates 8 Conclusions The general conclusion are consistent with those presented in [6]. Multliinear principal component analysis is an effective method of feature extraction of a motion data, which allows to identify humans precisely. However individual gait features are scattered across numerous MPCA components. Thus attribute selection simplifies gait description, discovers most discriminative features and allows for more efficient classification. The identification performed on the basis of reduced MPCA space, as presented in Fig. 10 is improved significantly in comparing to raw MPCA space as presented in Fig. 6. What is more, although strong dimensionality reduction, selected attributes preserve most of individual gait features. The obtained features subsets strongly depends on utilized subset evaluation strategy. Wrapper evaluation approach which in general is considered to be more precise, in our problem gives significantly worse results. It can be explained by scattering of individual gait features across MPCA components, so correlation based measures as for instance CFS evaluator, are more efficient in the input space exploration. Acknowledgment. The work has been supported by The Polish National Science Centre as project number UMO-2011/01/B/ST6/06988. References [1] Goldberg, D.E.: Genetic algorithms in search, optimization and machine learning. Addison-Wesley (1989) [2] Hall, M.A.: Correlation-based Feature Subset Selection for Machine Learning. Hamilton, New Zealand (1998)
Page 1 and 2:
Studies in Computational Intelligen
Page 3 and 4:
Aleksander Nawrat · Zygmunt Kuś E
Page 5 and 6:
“In God We Trust, All others we o
Page 7 and 8:
VIII Preface valuable information i
Page 9 and 10:
Contents Part I: Design of Object D
Page 11 and 12:
Contents XIII Technology Developmen
Page 13 and 14:
XVI List of Contributors Adam Czorn
Page 15 and 16:
XVIII List of Contributors Henryk M
Page 17 and 18:
XX List of Contributors Józef Wron
Page 19 and 20:
2 Design of Object Detection, Recog
Page 21 and 22:
4 A. Babiarz et al. 1 Introduction
Page 23 and 24:
6 A. Babiarz et al. The description
Page 25 and 26:
8 A. Babiarz et al. a) b) Fig. 5. T
Page 27 and 28:
10 A. Babiarz et al. o o o “micvo
Page 29 and 30:
12 A. Babiarz et al. a) b) Fig. 7.
Page 31 and 32:
14 A. Babiarz et al. • The camera
Page 33 and 34:
16 A. Babiarz et al. the camera ser
Page 35 and 36:
18 A. Babiarz et al. a) b) Fig. 11.
Page 37 and 38:
20 A. Babiarz et al. patrol point i
Page 39 and 40:
22 A. Babiarz et al. particular ang
Page 41 and 42:
24 A. Babiarz et al. The sample cod
Page 43 and 44:
Recognition and Location of Objects
Page 45 and 46:
Page 47 and 48:
Page 49 and 50:
Page 51 and 52:
Page 53 and 54:
Page 55 and 56:
Page 57 and 58:
Page 59 and 60:
Page 61 and 62:
Page 63 and 64:
48 P. Demski et al. Fig. 1. Automat
Page 65 and 66:
50 P. Demski et al. E-Grip firing m
Page 67 and 68:
52 P. Demski et al. 4 Automatic Tar
Page 69 and 70:
54 P. Demski et al. implementing su
Page 71 and 72:
Object Tracking for Rapid Camera Mo
Page 73 and 74:
Page 75 and 76:
Page 77 and 78:
Page 79 and 80:
Page 81 and 82:
Page 83 and 84:
Page 85 and 86:
Page 87 and 88:
Page 89 and 90:
Page 91 and 92:
Object Tracking in a Picture during
Page 93 and 94:
Page 95 and 96:
Page 97 and 98:
Page 99 and 100:
Page 101 and 102:
Page 103 and 104:
Page 105 and 106:
Page 107 and 108:
94 Construction of Image Acquisitio
Page 109 and 110:
96 G. Bieszczad et al. effectivenes
Page 111 and 112:
98 G. Bieszczad et al. such a way t
Page 113 and 114:
100 G. Bieszczad et al. Table 1. No
Page 115 and 116:
102 G. Bieszczad et al. Fig. 7. Pri
Page 117 and 118:
104 G. Bieszczad et al. Fig. 10. Re
Page 119 and 120:
106 G. Bieszczad et al. Fig. 13. Ti
Page 121 and 122:
108 G. Bieszczad et al. Fig. 16. Op
Page 123 and 124:
110 G. Bieszczad et al. has to make
Page 125 and 126:
112 G. Bieszczad et al. MTF functio
Page 127 and 128:
114 G. Bieszczad et al. [11] Krupi
Page 129 and 130:
116 D. Bereska et al. Presented in
Page 131 and 132:
118 D. Bereska et al. Fig. 2. Image
Page 133 and 134:
120 D. Bereska et al. Fig. 4. Ortho
Page 135 and 136:
Omnidirectional Video Acquisition D
Page 137 and 138:
Page 139 and 140:
Page 141 and 142:
Page 143 and 144:
Page 145 and 146:
Page 147 and 148:
Page 149 and 150:
Part III Design of Vision Based Con
Page 151 and 152:
Vision System for Group of Mobile R
Page 153 and 154:
Page 155 and 156:
Page 157 and 158:
Page 159 and 160:
Page 161 and 162:
Page 163 and 164:
Page 165 and 166:
Page 167 and 168:
Page 169 and 170:
A Distributed Control Group of Mobi
Page 171 and 172:
Page 173 and 174:
Page 175 and 176:
Page 177 and 178:
Page 179 and 180:
Page 181 and 182:
Page 183 and 184:
Page 185 and 186:
Page 187 and 188:
Page 189 and 190:
178 D. Bereska et al. The aim of th
Page 191 and 192:
180 D. Bereska et al. 3 Control Alg
Page 193 and 194:
182 D. Bereska et al. Fig. 5. Schem
Page 195 and 196:
184 D. Bereska et al. 4.1 Mechanica
Page 197 and 198:
186 D. Bereska et al. 4.1.3 Open-Lo
Page 199 and 200:
188 D. Bereska et al. Fig. 16. Comp
Page 201 and 202:
Probabilistic Approach to Planning
Page 203 and 204:
Page 205 and 206:
Page 207 and 208:
Page 209 and 210:
Page 211 and 212:
Page 213 and 214:
Page 215 and 216:
206 Practical Applications of Class
Page 217 and 218:
208 M. Mellado and K. Skrzypczyk an
Page 219 and 220:
210 M. Mellado and K. Skrzypczyk a
Page 221 and 222:
212 M. Mellado and K. Skrzypczyk wh
Page 223 and 224: 214 M. Mellado and K. Skrzypczyk Fi
Page 225 and 226: 216 M. Mellado and K. Skrzypczyk 6
Page 227 and 228: Prototyping the Autonomous Flight A
Page 241 and 242: Feature Extraction and HMM-Based Cl
Page 255 and 256: 248 K. Daniec et al. In this articl
Page 257 and 258: 250 K. Daniec et al. Fig. 2. Applic
Page 259 and 260: 252 K. Daniec et al. by the ability
Page 261 and 262: 254 K. Daniec et al. The idea of us
Page 263 and 264: Selection of Individual Gait Featur
Page 273: Selection of Individual Gait Featur
Page 279 and 280: 274 A. Nawrat et al. alia weather c
Page 281 and 282: 276 A. Nawrat et al. have an embedd
Page 283 and 284: 278 A. Nawrat et al. particular cha
Page 285 and 286: 280 A. Nawrat et al. • 1-phase el
Page 287 and 288: 282 A. Nawrat et al. The phase cond
Page 289 and 290: 284 A. Nawrat et al. Table 1. Compa
Page 291 and 292: 286 A. Nawrat et al. P3 P3 P3 71101
Page 293 and 294: 288 A. Nawrat et al. S, VA 4,510 4,
Page 295 and 296: 290 A. Nawrat et al. Comparison dif
Page 297 and 298: Technology Development of Military
Page 315 and 316: 312 A. Czornik, A. Nawrat, and M. N
Page 325 and 326:
322 A. Czornik, A. Nawrat, and M. N
Page 327 and 328:
Page 329 and 330:
Page 331 and 332:
328 M. Koźlak, A. Kurzeja, and A.
Page 333 and 334:
Page 335 and 336:
Page 337 and 338:
Page 339 and 340:
Page 341 and 342:
Page 343 and 344:
Page 345 and 346:
Page 347 and 348:
344 Conclusions Control algorithms
show all

[Studies in Computational Intelligence 481] Artur Babiarz, Robert Bieda, Karol Jędrasiak, Aleksander Nawrat (auth.), Aleksander Nawrat, Zygmunt Kuś (eds.) - Vision Based Systemsfor UAV Applications (2013, Sprin

Create successful ePaper yourself

Delete template?

Save as template?