Go to DBLP homepageVision-Based Deep Q-Learning Network Models to Predict Particulate Matter Concentration Levels Using Temporal Digital Image Data
Abstract: Particulate matter (PM) has been revealed to have detrimental effects on public health, social economy, agriculture, and so forth. Thus, it became one of the major concerns in terms of a factor that can reduce “quality of life” over East Asia, where the concentration is significantly high. In this regard, it is imperative to develop affordable and efficient prediction models to monitor real-time changes in PM concentration levels using digital images, which are readily available for many individuals (e.g., via mobile phone). Previous studies (i.e., DeepHaze) were limited in scope to priorly collected data and thereby less practical in providing real-time information (i.e., undermined interprediction). This drawback led us to hardly capture drastic changes caused by weather or regions of interests. To address this challenge, we propose a new method called Deep Q-haze, whose inference scheme is built on an online learning-based method in collaboration with reinforcement learning and deep learning (i.e., Deep Q-learning), making it possible to improve testing accuracy and model flexibility in virtue of real-time basis inference. Taking into account various experiment scenarios, the proposed method learns a binary decision rule on the basis of video sequences to predict, in real time, whether the level of PM10 (particles smaller than 10 in aerodynamic diameter) concentration is harmful (<svg xmlns:xlink="http://www.w3.org/1999/xlink" xmlns="http://www.w3.org/2000/svg" style="vertical-align:-0.6370101pt" id="M1" height="7.88973pt" version="1.1" viewBox="-0.0498162 -7.25272 7.75925 7.88973" width="7.75925pt"><g transform="matrix(.013,0,0,-0.013,0,0)"><path id="g117-92" d="M512 230V281L75 514V456L453 256V254L75 55V-3L512 230Z"/></g></svg>80<svg xmlns:xlink="http://www.w3.org/1999/xlink" xmlns="http://www.w3.org/2000/svg" style="vertical-align:-3.42943pt" id="M2" height="9.39034pt" version="1.1" viewBox="-0.0498162 -5.96091 14.2444 9.39034" width="14.2444pt"><g transform="matrix(.013,0,0,-0.013,0,0)"><path id="g113-234" d="M538 96L524 119C492 88 454 63 446 63C439 63 434 70 440 101C463 223 491 341 518 448H508L433 422L401 276C355 192 240 56 188 56C163 56 154 89 163 133C184 233 207 338 235 443L230 448L152 424L58 17C40 -60 23 -143 23 -185C23 -241 42 -261 63 -261C92 -261 117 -241 125 -227L124 -221C109 -209 89 -165 89 -103C89 -55 96 -13 105 26H107C121 -3 134 -12 151 -12C172 -12 194 -5 221 16C279 61 330 124 384 194H386C381 159 377 141 368 105C343 3 363 -12 383 -12C416 -12 486 35 538 96Z"/></g><g transform="matrix(.013,0,0,-0.013,6.694,0)"><path id="g113-104" d="M546 430L539 434C529 434 505 438 495 440C473 444 450 448 430 448C352 448 265 412 213 366C145 306 96 203 96 103C96 22 135 -12 160 -12C190 -12 238 14 262 32C310 68 368 120 411 184H413C403 117 396 75 384 21C353 -118 325 -158 291 -184C270 -200 241 -205 208 -205C133 -205 90 -164 74 -110C70 -98 58 -100 49 -107C34 -119 23 -140 23 -155C23 -190 74 -261 166 -261C219 -261 280 -233 314 -208C383 -157 446 -79 470 81C491 223 529 388 546 430ZM456 386C452 357 433 283 420 252C402 216 366 174 325 129C288 88 239 56 212 56C192 56 182 77 182 120C182 165 199 242 226 292C256 348 281 377 311 389C327 395 353 402 375 402C408 402 436 394 456 386Z"/></g></svg>/<svg xmlns:xlink="http://www.w3.org/1999/xlink" xmlns="http://www.w3.org/2000/svg" style="vertical-align:-0.2063999pt" id="M3" height="11.7978pt" version="1.1" viewBox="-0.0498162 -11.5914 15.3415 11.7978" width="15.3415pt"><g transform="matrix(.013,0,0,-0.013,0,0)"><path id="g113-110" d="M766 88L752 113C719 83 690 64 681 64C674 64 672 74 679 103L724 292C758 436 724 448 701 448C680 448 666 442 639 429C594 407 514 350 441 252H439L447 289C476 423 450 448 419 448C398 448 379 441 355 427C307 400 234 344 162 249H160L180 324C203 409 197 448 170 448C144 448 82 413 23 349L35 321C57 343 96 374 108 374C115 374 117 371 111 341C87 227 57 112 24 -6L32 -12C53 -4 81 4 108 6C119 68 134 128 149 171C177 229 309 383 364 383C387 383 388 355 373 282C354 190 330 92 303 -6L309 -12C332 -4 356 3 386 6C396 63 411 122 424 171C458 236 590 383 642 383C658 383 664 369 652 315L603 91C587 20 593 -12 619 -12C642 -12 708 23 766 88Z"/></g><g transform="matrix(.0091,0,0,-0.0091,10.257,-5.741)"><path id="g50-52" d="M290 377C321 398 342 415 358 430C378 450 389 473 389 502C389 578 329 635 238 635H237C184 635 137 610 109 578L64 515L88 493C112 529 154 573 208 573S303 542 303 482C303 409 233 370 141 341L149 308C165 313 190 319 215 319C272 319 341 283 341 193C342 98 292 43 222 43C163 43 122 72 96 94C88 101 79 100 70 94C61 87 47 73 46 60C44 47 48 37 62 23C76 10 118 -12 165 -12C238 -12 430 62 430 223C430 297 379 359 290 375V377Z"/></g></svg>) or not. The proposed model shows superior accuracy compared to existing algorithms. Deep Q-haze effectively accounts for unexpected environmental changes in essence (e.g., weather) and facilitates monitoring of real-time PM10 concentration levels, showing implications for better understanding of characteristics of airborne particles.
Loading