Fig. 1. Clustering deviation effect of different number of detected targets

Fig. 2. Target classification effect of K-means method. (a)(b) Original images; (c)(d) weak noise condition; (e)(f) strong noise condition

Fig. 3. Positions of target point before and after horizontal projection. (a) Before horizontal projection; (b) after horizontal projection

Fig. 4. Dimension reduction mapping principle

Fig. 5. Effects of dimension reduction mapping. (a) 3D imaging result; (b) direct projection; (c) keyhole imaging projection

Fig. 6. Motion characteristics of three flight targets. (a) Bird; (b) rotary wing UAV; (c) fixed wing UAV

Fig. 7. Optical flow detection results of three targets. (a) Bird; (b) rotary wing UAV; (c) fixed wing UAV

Fig. 8. Optical flow angle information of three flight targets. (a) Bird; (b) rotary wing UAV; (c) fixed wing UAV

Fig. 9. Amplitude and angle distribution of optical flow abnormal frame (taking fixed wing UAV as an example). (a) Maximum amplitude distribution; (b) angle distribution

Fig. 10. Optical flow detection process for air moving target