Yang et al. Table 3. Test Results of Four Air Infrared Object Recognition Algorithm Launch distance (m) Horizonal entry angle Type of maneuver Launching conditions Time interval between groups(s) 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 7000 10 10 10 10 10 10 10 10 40 40 40 40 40 40 40 40 100 100 100 100 100 100 100 100 160 160 160 160 160 160 160 160 Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump Left Jump 0.4 0.4 0.4 0.4 0.7 0.7 0.7 0.7 0.4 0.4 0.4 0.4 0.7 0.7 0.7 0.7 0.4 0.4 0.4 0.4 0.7 0.7 0.7 0.7 0.4 0.4 0.4 0.4 0.7 0.7 0.7 0.7 Number of decoys per group 2 2 4 4 2 2 4 4 2 2 4 4 2 2 4 4 2 2 4 4 2 2 4 4 2 2 4 4 2 2 4 4 Number of groups 24 24 12 12 24 24 12 12 24 24 12 12 24 24 12 12 24 24 12 12 24 24 12 12 24 24 12 12 24 24 12 12 MultiFeature fusion Recognition rate(%) SVM NB TAN 2DPCANetNB (Ours) 61.38 77.16 82.44 90.73 64.43 82.1 83.17 87.22 68.24 81.47 85.73 83.06 64.78 81.32 84.37 88.18 69.38 78.51 80.34 93.95 59.8 81.9 86.07 90.54 69.17 79.7 82.26 86.72 68.33 80.99 81.55 91.54 64.46 76.58 85.57 89.22 63.34 78.13 85.15 93.23 68.49 83.04 82.02 85.22 65.68 82.59 81.62 93.48 58.98 81.67 85.15 91.92 65.45 82.92 83.12 95.88 66.74 78.24 80.6 86.82 69.21 81.6 85.32 96.13 66.34 81.57 80.58 85.51 59.49 81.15 82.13 95.77 61.24 78.02 85.04 94.65 63 77.62 86.59 95.32 59.27 81.15 85.67 85.92 64.88 79.12 86.92 98.14 66.85 78.6 80.24 97.07 69.81 81.32 85.6 97.75 63.99 78.44 81.16 92.5 63.88 78.17 86.1 91.25 60.5 76.73 85.47 94.75 60.47 80.37 85.32 92 59.7 76.34 82.51 95.04 58.62 82.07 85.9 93.75 58.76 78.26 83.7 97.33 68.31 77.55 84.99 94.57 97.58 93.08 91.82 93.86 96.29 94.45 92.23 93.67 91.27 94.45 93.81 95.13 93.1 94.75 91.14 95.03 92.79 94.33 96.07 94.65 92.94 96.86 97.66 96.14 94.63 93.31 96.11 96.2 97.58 95.21 96.08 95.08 Before testing, we use the pretraining model trained by large-scale datasets to better represent the difference between target and interference. Bayesian network need design network structure and connection form according to feature vector extracted by 2DPCANet, and design conditional probability table. The test conditions set for the FEBRUARY 2023 experiment include the full-range countermeasure ability of point source artificial decoys and continuous release against the interference of typical artificial decoys of " volley + continuous shooting. " The release interval is 0.1 s; the object maneuverability is not less than 6g, and all simulation images are used as the test set for the IEEE A&E SYSTEMS MAGAZINE 13