• Optics and Precision Engineering
  • Vol. 31, Issue 9, 1366 (2023)
Daxiang LI, Zhongheng SU*, and Ying LIU
Author Affiliations
  • College of Communication and Information Engineering, Xi'an University of Posts and Telecommunication, Xi'an710121, China
  • show less
    DOI: 10.37188/OPE.20233109.1366 Cite this Article
    Daxiang LI, Zhongheng SU, Ying LIU. Road traffic sign recognition algorithm based on improved YOLOv4[J]. Optics and Precision Engineering, 2023, 31(9): 1366 Copy Citation Text show less

    Abstract

    To address the low recognition accuracy resulting from multiple scale changes in the traffic signs of complex scenes, an improved YOLOv4 algorithm is proposed. First, an attention-driven scale-aware feature extraction module is designed, and the range of receptive fields in each layer is widened to obtain more fine-grained multi-scale features by constructing a hierarchical connection mode similar to the residual structure; this is followed by the generation of a pair of attention maps with directional-aware and position-sensitive characteristics under the attention drive so that the network can focus on key areas with more discrimination. Following this, a feature-aligned pyramid convolution feature fusion module is constructed, and the feature offset between adjacent scale feature maps is obtained via convolution for feature alignment. Finally, the network adaptively learns the optimal feature fusion mode through pyramid convolution and constructs a feature pyramid to identify traffic signs with different scales. Experimental results indicate that the recognition accuracy for the TT100K dataset is improved by 5.4% compared with that of the original YOLOv4 algorithm, which is superior to other recognition algorithms, and the FPS reaches 33.17. Thus, the proposed algorithm satisfies the requirements of accuracy and real-time performance for road traffic sign recognition.
    bx=σ(tx)+cxby=σ(ty)+cybw=pwetwbh=pheth(1)

    View in Article

    Yi=Di(Xi)              i=1Di(Xi+Yi-1)     1<in(2)

    View in Article

    zch(h)=1W0i<Wxc(h,i)(3)

    View in Article

    zcw(w)=1H0j<Hxc(j,w)(4)

    View in Article

    f=δ(F1([zh,zw]))(5)

    View in Article

    gh=σ(Fh(fh))(6)

    View in Article

    gw=σ(Fw(fw))(7)

    View in Article

    yc(i,j)=xc(i,j)×gch(i)×gcw(j)(8)

    View in Article

    P4=W0A3λ43+W1A4λ44+W2A5λ45(9)

    View in Article

    λ43(i,j)=exp(δ43(i,j))n=35exp(δ4n(i,j))(10)

    View in Article

    λ43(i,j)+λ44(i,j)+λ45(i,j)=1(11)

    View in Article

    λ43,λ44,λ45[0,1](12)

    View in Article

    Pl=W1A3λll+W2A4λll+1,                      l=3W0A3λll-1+W1A4λll+W2A5λll+1,l=4W0A4λll-1+W1A5λll,                     l=5.(13)

    View in Article

    Pn=UF^n+1,ΔF^n+1+UF^n,ΔF^n(14)

    View in Article

    Uhw=h'=1Hw'=1WF^h'w'max0,1-h+Δ1hw-h'max0,1-w+Δ2hw-w',               (15)(15)

    View in Article

    RCIoU=ρ2(b,bgt)c2+αν(16)

    View in Article

    ν=4π2arctanwgthgt-arctanwh2(17)

    View in Article

    LCIoU=1-IoU+RCIoU(18)

    View in Article

    α=ν(1-IoU)+ν(19)

    View in Article

    LCEp,y=-log(p),if  y=1-log(1-p),otherwise(20)

    View in Article

    LFL(pt)=-αt(1-pt)γlog(pt)(21)

    View in Article

    pt=p,ify=11-p,otherwise(22)

    View in Article

    Lloss(object)=i=0K×Kj=0MIijobjLCIoU-i=0K×Kj=0MIijobjLCONF+λnoobji=0K×Kj=0MIijnoobjLCONF-                            i=0K×KIijobjcclassesLCLASS,                      (23)(23)

    View in Article

    LCONF=C^ilog(Ci)+(1-C^i)log(1-Ci)(24)

    View in Article

    LCLASS=(1-αt)p^i(c)γlog(1-p^i(c))+                  αt(1-p^i(c)γlog(p^i(c)γ),                  (25)(25)

    View in Article

    P=TPTP+FP(26)

    View in Article

    R=TPTP+FN(27)

    View in Article

    F1=2×P×RP+R(28)

    View in Article

    Daxiang LI, Zhongheng SU, Ying LIU. Road traffic sign recognition algorithm based on improved YOLOv4[J]. Optics and Precision Engineering, 2023, 31(9): 1366
    Download Citation