CVPR #****

CVPR 2008 Submission ****. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. CVPR #****

1

100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149

150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199

Abstract

This project is going to implement a object recognition module for a on field environmental observing system. Specifically, tracking bee population and activities by on field camera data. I have taken CSE166 in Fall 2009 as the qualification for CSE 190a project.

1. Introduction This project is going to join the on going research of a

ecological image database system in California Institute for Telecommunications and Information Technology(Calit2) and Taiwan Forestry Research Institution(TFRI). This project is going to use a computer vision approach to count the number of bees in the image database which were taken from the Shanping Workstation in Taiwan. Furthermore, this project is attempt to make the program as portable as possible so that a on-field camera is able to determine whether or not sending more image or video to the workstation.

2. Milestones

2.1. Week 1 Installation necessary tools (find a possible way of connecting the interfaces such as Java and Matlab) compare the test result of different libaries.

2.2. Week 2 Build a minimum algorithm for bee recognition. Build a practice set of images, identify possible problems for the image set.

2.3. Week 3-4 Build a basic program to distinguish background and the bees. The program should be able to handle different ligtings aross different time of a day as well.

2.4. Week 5 Add wasp recognition to the prorgam in order to triger the wasp attack event.

2.5. Week 6-8 Add aglinment across different images, so that the change of camera viewpoint. The program should be able to handle external events (such as birds or other objects in the scene) and fatal errors (completely dark image or camera outrage)

2.6. Week 8-10 Testing and integration phrase. It is expected to run this

program on an embeded computer with MIPS processor. CPU and memory usage should be optimized.

3. Problems to be solved

3.1. How to distinguish background and the objects of interest

By observing the images, the good side is backgrounds do not change frequently, but the down side is backgrounds containt a lot of distractions (black dots which look very much alike to a bee)

3.2. How to handle the cases when bees group together and stack on top of each other

As illustrated in Figure 1. when bees are closed to each other. When a wasp attacks the bees, bees group together around the wasp and vibrate in order to defence their nest. If a wasp can be recognize the program should be able to triger the camera and perform a video capture [3].

Proposal for CSE 190a Projects in Vision & Learning: Implementation of Object Recognition for a Environmental Observing System

Haili Wang (haw007@ucsd.edu)

Paper ID ****< replace **** here and in header with paperID>

CVPR #****

CVPR 2008 Submission ****. CONFIDENTIAL REVIEW COPY. DO NOT DISTRIBUTE. CVPR #****

2

200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249

250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299

3.3. Camera alignment and lighting problems. It would be much easier to perform the counting task if the program is able to take the background away. However, camera position changes over time because of wind, gravity or any other external force. For anther thing, lighting of the scene changes over time. In order to ignore the background, a baseline should be drawn.

3.4. On field camera integration problems The final report will adress difficulties and problems of integration with a realtime on field camera.

4. Softwares to build upon This project is going to be developed in Matlab and Java, and interfaced with the open source middle ware Data Turbine[2], specifically, Ring Buffered Network Bus (RBNB) and/or Realtime Data Viewer (RDV) [4].

5. Datasets This project will use the images from archiving gallery of Shanping workstation of TFRI [1]. Image resolution is 1600*1200 since September 2008. White painted background since December 2008.

References [1] Sensor Archiving Gallery of Shanping Workstation

http://srb2.tfri.gov.tw/wsn/ [2] Open Source DataTurbine Initiative: Streaming Data

Middleware and Applications, Tony Fountain, Sameer Tilak, Jane Rutherford and delegation from Canada, (2008)

[3] Ecological image databases: From the webcam to the researcher, John Porter, Chau-Chin Lin, David E. Smith, Sheng-Shan Lu (2009)

[4] A visualization environment for scientific and engineering data http://code.google.com/p/rdv/

Figure1: Sample image from TFRI Upper: A sample image from TFRI website taken at Feb 04, 2009 7:02 AM [1] original

size 1600*1200 pixels Lower: Cropped image from the upper image