LED-based adaptive multispectral camouflage via environmental spectral reproduction in the UV–Vis–NIR band

Ground-based military equipment faces growing exposure to multispectral reconnaissance, and conventional camouflage measures such as fixed coatings and nets cannot fully follow the spectral evolution of natural backgrounds. This paper presents an adaptive spectral reproduction framework for multispectral camouflage built around a 62-channel light-emitting diode (LED) array that reproduces the background spectral signature within the ultraviolet–visible–near-infrared (UV–Vis–NIR, 340–1020 nm) band. A radiometric model is formulated first; it relates the actively emitted spectrum to the apparent radiance received by a remote multispectral sensor and removes the ambiguity between reflectance matching and emission matching. An iterative reconstruction algorithm is designed that combines batch-wise non-negative least-squares (NNLS) fitting, cosine-similarity-based LED selection, per-batch tiny-current removal, and residual refinement. The algorithm is evaluated on 17 environmental spectra drawn from nine categories (diurnal variations, vegetation, soil, water, grassland, desert, camouflage coating, camouflage net, and other targets). With MAX_EPOCHS = 50, SCALE_FACTOR = 1.0, and MAX_LED_PER_BATCH = 8, the mean Pearson correlation coefficient (PCC) is 0.9999, the mean coefficient of determination (R 2 ) is 0.9999, the mean root-mean-square error (RMSE) is 0.0031, the mean spectral angle mapper (SAM) is 0.37°, and the mean fit ratio is 0.9990, exceeding global NNLS, bounded-variable least-squares (BVLS), L1-sparse fitting, differential evolution, and a neural-network-based optimizer under identical objectives and metrics. The mean RMSE remains near 0.020 at a signal-to-noise ratio (SNR) of 5 dB. A field validation within the spectral bands of an airborne multispectral sensor is established to quantify target–background apparent-radiance similarity, contrast, classification accuracy, and detection probability in real scenes. Because the LED array covers only the UV–Vis–NIR band, the framework acts as a spectral-mismatch reduction layer for this band.