Upgrade to Pro — share decks privately, control downloads, hide ads and more …

[DICOMO2019] 路側設置ステレオマイクを用いた車両種別推定手法の初期的評価 / Initial Evaluation of Vehicle Type Estimation using Sidewalk Stereo Microphones

[DICOMO2019] 路側設置ステレオマイクを用いた車両種別推定手法の初期的評価 / Initial Evaluation of Vehicle Type Estimation using Sidewalk Stereo Microphones

Presented in DICOMO 2019, Fukushima, Japan

石田 繁巳, 内野 雅人, 小池 大地, 田頭 茂明, 福田 晃
路側設置ステレオマイクを用いた車両種別推定手法の初期的評価
情報処理学会マルチメディア, 分散, 協調とモバイルシンポジウム(DICOMO2019), pp.1682-1687, Jul 2019
paper: http://id.nii.ac.jp/1001/00202424/
pdf: https://pman0214.netlify.app/static/9a26933151d0e007c5964f0628e614bd/ishida19-dicomo.pdf

Shigemi ISHIDA

July 05, 2019
Tweet

More Decks by Shigemi ISHIDA

Other Decks in Research

Transcript




  1. *1, *1, *1,
    *2, *1
    *1
    *2

    View full-size slide

  2. H>
    n ITS (Intelligent Transport System)
    • 1-;?% .@6$(&)B:,
    K)*I)+

    n A
    • 6$• MG#J
    • (3)!'59/
    ⇒D7/8/=E
    2F
    3

    View full-size slide


  3. n %*
    • #-+ (
    • "$1)%* !./'
    0%,&

    ं྆૸ߦԻ
    ૸ߦํ޲
    ଎౓
    ं྆छผ
    ૸ߦंઢ

    4

    View full-size slide

  4. .;
    n '$8?7)SAVeD [Ishida+18]
    • OR0+19QF83%
    n !*2 [Ishida+19]

    FG77%
    n ICB>6,
    8?• #F8?7)

    ⇒MD7)8?J=5=
    [Ishida+18] ”SAVeD: Acoustic vehicle detector with speed estimation capable of sequential vehicle detection”, IEEE
    ITSC
    [Ishida+19] “Design of acoustic vehicle detector with stead—noise suppression”, IEEE ITSC (will appear)
    [Kubo+19] “KP"' @N-3&8?7)%/4 L:”, IPSJ ITSEH(
    5

    View full-size slide

  5. *-BE
    n JF%6>G:H)
    • KL3=YC9M/P
    [Aljaafreh+10] [Changjun+09]
    [Munich+04]
    • !#"D?
    [Yang+07]
    ⇒T36>JS
    8U
    n NJF%6>G:H)
    • ETC ⇒ &R4VO
    • % !$;W7I
    [Hongliang+04]
    • 6>';W7I
    [Avery+04]
    ⇒X@Z[(Q,A+1<5<0.2
    6
    [Aljaafreh+10] “An evaluation of feature extraction methods for vehicle classification based on acoustic signals”, IEEE ISNSC
    [Changjun+09] “The research of vehicle classification using SVM and KNN in a ramp”, IFCSTA
    [Munich+04] “Bayesian subspace methods for acoustic signature recognition of vehicles”, EUSIPCO
    [Yang+07] “Vehicle identification using wireless sensor networks”, IEEE SoutheastCon
    [Hongliang+04] “A hybrid license plate extraction method baed on edge statistics and morphology”, ICPR
    [Avery+04] “Length-based vehicle classification using images from uncalibrated video cameras”, IEEE ITSC

    View full-size slide

  6. !
    n %

    + ",
    • #
    %

    '*
    ⇒ %

    !+

    -)(
    &
    $
    8
    Vehicle Sound
    t
    t
    t
    t
    Left
    Right
    Vehicle sound arrives at slight
    different time depending on
    vehicle location
    Compensate the time
    difference
    Emphasized
    Sound
    ∆t ∆t

    View full-size slide

  7. (
    9
    1. $&
    $&
    2.

    • '"!Δ"
    3.

    Δ"
    $&

    4. #%

    #
    %
    ڧௐ߹੒
    ಛ௃ྔ
    நग़
    ं྆ݕग़
    ं྆छผ
    ਪఆ
    Ϧϯά
    όοϑΝ
    ग़ྗ
    ӈ

    ड৴࣌ؒࠩ t
    AAAB8XicZVBNS8NAEJ3Urxq/qh69LJZCTyURQY8FPXisYD+gDWWz3bZLN5uwOxFK6J/wJnoQr/4dL/4bt20Q2z4YeLw3w8y8MJHCoOf9OIWt7Z3dveK+e3B4dHxSOj1rmTjVjDdZLGPdCanhUijeRIGSdxLNaRRK3g4nd3O//cy1EbF6wmnCg4iOlBgKRtFKnd49l0gJ9ktlr+YtQDaJn5My5Gj0S9+9QczSiCtkkhrT9b0Eg4xqFEzymVvppYYnlE3oiHctVTTiJsgWB89IxSoDMoy1LYVkobr/JjIaGTONQtsZURybdW8u/nmrq3B4G2RCJSlyxZabhqkkGJP5+2QgNGcop5ZQpoW9lrAx1ZShDcm1MfjrT2+S1lXNt/zxulyv5oEU4QIuoQo+3EAdHqABTWAg4QXe4N0xzqvz4XwuWwtOPnMOK3C+fgFwao/g
    AAAB8XicZVBNS8NAEJ3Urxq/qh69LJZCTyURQY8FPXisYD+gDWWz3bZLN5uwOxFK6J/wJnoQr/4dL/4bt20Q2z4YeLw3w8y8MJHCoOf9OIWt7Z3dveK+e3B4dHxSOj1rmTjVjDdZLGPdCanhUijeRIGSdxLNaRRK3g4nd3O//cy1EbF6wmnCg4iOlBgKRtFKnd49l0gJ9ktlr+YtQDaJn5My5Gj0S9+9QczSiCtkkhrT9b0Eg4xqFEzymVvppYYnlE3oiHctVTTiJsgWB89IxSoDMoy1LYVkobr/JjIaGTONQtsZURybdW8u/nmrq3B4G2RCJSlyxZabhqkkGJP5+2QgNGcop5ZQpoW9lrAx1ZShDcm1MfjrT2+S1lXNt/zxulyv5oEU4QIuoQo+3EAdHqABTWAg4QXe4N0xzqvz4XwuWwtOPnMOK3C+fgFwao/g
    AAAB8XicZVBNS8NAEJ3Urxq/qh69LJZCTyURQY8FPXisYD+gDWWz3bZLN5uwOxFK6J/wJnoQr/4dL/4bt20Q2z4YeLw3w8y8MJHCoOf9OIWt7Z3dveK+e3B4dHxSOj1rmTjVjDdZLGPdCanhUijeRIGSdxLNaRRK3g4nd3O//cy1EbF6wmnCg4iOlBgKRtFKnd49l0gJ9ktlr+YtQDaJn5My5Gj0S9+9QczSiCtkkhrT9b0Eg4xqFEzymVvppYYnlE3oiHctVTTiJsgWB89IxSoDMoy1LYVkobr/JjIaGTONQtsZURybdW8u/nmrq3B4G2RCJSlyxZabhqkkGJP5+2QgNGcop5ZQpoW9lrAx1ZShDcm1MfjrT2+S1lXNt/zxulyv5oEU4QIuoQo+3EAdHqABTWAg4QXe4N0xzqvz4XwuWwtOPnMOK3C+fgFwao/g
    AAAB8XicZVBNS8NAEJ3Urxq/qh69LJZCTyURQY8FPXisYD+gDWWz3bZLN5uwOxFK6J/wJnoQr/4dL/4bt20Q2z4YeLw3w8y8MJHCoOf9OIWt7Z3dveK+e3B4dHxSOj1rmTjVjDdZLGPdCanhUijeRIGSdxLNaRRK3g4nd3O//cy1EbF6wmnCg4iOlBgKRtFKnd49l0gJ9ktlr+YtQDaJn5My5Gj0S9+9QczSiCtkkhrT9b0Eg4xqFEzymVvppYYnlE3oiHctVTTiJsgWB89IxSoDMoy1LYVkobr/JjIaGTONQtsZURybdW8u/nmrq3B4G2RCJSlyxZabhqkkGJP5+2QgNGcop5ZQpoW9lrAx1ZShDcm1MfjrT2+S1lXNt/zxulyv5oEU4QIuoQo+3EAdHqABTWAg4QXe4N0xzqvz4XwuWwtOPnMOK3C+fgFwao/g
    sR(t)
    (null)
    (null)
    (null)
    (null)
    sL(t)
    (null)
    (null)
    (null)
    (null)
    semph(t) = sR(t) + sL(t + t)
    AAACGHicZZBNS8NAEIY3ftb6VfXoZbEIEbEkKuhFKOjBg4cqVgVTwmY7tYu7SdidCCX0V3jzn3gTPYhX8eK/cdMW8eOFhYd3ZpidN0qlMOh5n87Y+MTk1HRppjw7N7+wWFlavjBJpjk0eSITfRUxA1LE0ESBEq5SDUxFEi6j28OifnkH2ogkPsdeCi3FbmLREZyhtcLKlgkDxbCrVQ4q7fZd3KAH1IRnBWxaOHFxMzgCiYziRlipejVvIPof/BFUyUiNsPIRtBOeKYiRS2bMte+l2MqZRsEl9MvrQWYgZfyW3cC1xZgpMK18cFefrlunTTuJti9GOnDLPyZypozpqch2FjeYv7XC/K79XoWd/VYu4jRDiPlwUyeTFBNapETbQgNH2bPAuBb2t5R3mWYcbZZlG4P/9+j/cLFd83dq26e71bo7CqREVskacYlP9kidHJMGaRJO7skjeSYvzoPz5Lw6b8PWMWc0s0J+yXn/AuIXnYQ=
    ૸ߦԻऔಘ

    View full-size slide

  8. :$;*/-4,#
    10
    n SAVeD+) [Ishida+18]
    • ! :$;=7%&6>
    " >2
    • RANSAC)
    S8"
    • ".05%Δ!(!)1#
    [Ishida+18] ”SAVeD: Acoustic vehicle detector with speed estimation capable of sequential vehicle detection”, IEEE
    ITSC
    D/2
    M1
    M2
    x
    O
    d2
    d1
    D/2
    L
    39 40 41 42
    Time [s]
    −1
    .
    5
    −1
    .
    0
    −0
    .
    5
    0
    .
    0
    0
    .
    5
    1
    .
    0
    1
    .
    5
    Sound Delay [ms]
    • '3%• 9(&

    View full-size slide


  9. 11
    1. '"

    2. %'"
    !* +$#Δ"

    (),
    3. &'"
    t
    Left
    Right
    t
    f
    f
    Prepare for windowed data
    f
    Shift in time and
    add to right-
    channel data
    FFT
    FFT

    View full-size slide

  10. n 67% SVM
    • 4 ,#0

    n ,:2
    • 10kHz&67%$;
    • 67%$.!1"'9LPF
    )-3*4
    12



    5+!
    (
    8
    (SVM)
    )-/

    View full-size slide

  11. <*75 (1)
    n > =$ %6
    • ?.1(4&2(4
    • AZDEN SGM-990 SONY HDR-MV1
    • A! = 50cm
    • %6 @;
    # = 3m or 6m
    • 20@48kHz, 16bit
    n +9"1 #)
    n 100,8'-100!
    • 3(2:/

    14
    !"#$%&'%()*
    +,-.()/0%.1
    0)#%$1)$

    View full-size slide

  12. 9.64 (2)
    15
    n +,2$1
    :->708#
    n ,270(: 3
    • =$,46/78/18/$1
    n 5%);: 320ms
    n FFT : 4096' (≅85.3ms)
    • 25%
    n ,2$1": "#$%%
    = 2.0s
    • ,2$1<
    ! *"#$%%
    &9.
    ⇒3FFT ,2708#

    View full-size slide


  13. 16
    ⇒ 4.71%

    normal bike bus
    Estimated Type
    normal
    bike
    bus
    Actual Type
    89.06 7.44 3.49
    8.54 87.33 4.13
    2.92 2.58 94.50
    normal bike bus
    Estimated Type
    normal
    bike
    bus
    Actual Type
    94.24 3.74 2.02
    4.38 93.49 2.13
    1.59 1.11 97.30

    : 90.30% (): 95.01%

    View full-size slide

  14. -6
    17
    n )35198 &7
    • :)5.18%6.16%2.79%
    ⇒,*")32

    /4
    • 2' /4#0+
    n )3512.%

    (.% 46 78 18 142
    !$2.% 17
    (36.96%)
    43
    (55.13%)
    5
    (27.78%)
    65
    (45.77%)

    View full-size slide


  15. n ITSDK(.&.%
    • DK
    6J BI
    n ')-"+!$ GDK(.&.%
    • Z=DLDKC08;A
    • Z=-.DKW 1TDK O
    Q
    W E[ DKRI S4
    n X/V?#,.*'5@MUF8N
    • Y795.01%DKRI S4
    • 239<6JDK>:PH


    19

    View full-size slide

  16. © 2019 Shigemi ISHIDA licensed under CC BY-NC 4.0

    View full-size slide