Heterogeneous Gray-Temperature Fusion-Based Deep Learning Architecture for Far Infrared Small Target Detection

Figure 7

Comparison of detection results according to data fusion method. All the same normalization methods were used with values between -1 and 1. The image (a) shows the case where only the radiometric temperature data is used for three channels, (b) shows the case where only the gray scale data is used, (c) shows the case where the temperature data is fused using only one channel, and (d) is an example in which temperature data is fused to two channels.

(a)

(b)

(c)

(d)