Fig. 1. Different types of features in heterogeneous modes. (a1)--(a4) Linear features; (b1)--(b4) T-type features; (c1)(c2) cross and diagonal features

Fig. 2. Processing flow of TDFF algorithm

Fig. 3. Extraction results of different features. (a1)(a2) Traditional LBP features; (b1)(b2) T-MFLBP features

Fig. 4. Comparison of detection rates of different methods in different datasets

Fig. 5. Comparison of false alarm rates of different methods in different datasets

Fig. 6. Detection rate comparison curves of different methods in different dimensions

Fig. 7. Comparison curves of false alarm rates in different dimensions by different methods

Fig. 8. Comparison curves of detection rates of different feature fusion methods under different number of iterations

Fig. 9. Comparison curves of false alarm rates of different feature fusion methods under different number of iterations