Multi-band composite detection and recognition of aerial infrared point targets

Because of the low detection probability and recognition efficiency of the single-band signal of an infrared point target, the multi-band composite detection probability is analyzed and a multi-band recognition method is proposed. By using the binary hypothesis method, we can conveniently deduce the multi-band theoretical detection threshold and obtain the composite detection probability and false alarm probability. The numerical simulation results show that multi-band composite detection ability is superior to that of single-band. For the difficult classification between background and target spectral signals with serious interference noise in reality, this paper proposes a highly effective spectral signal recognition method based on the optimized AdaBoost algorithm. By virtue of the high classification accuracy of the AdaBoost algorithm, with the Pruning classifier as a basic classifier to be learnt, the proposed method abandons any abnormal data that seriously deviate from the mean value center of the correction classification data, the end result of which reduces the number of iterations and enhances the classification effect of the AdaBoost algorithm. Compared with other algorithms from test results, the optimized AdaBoost algorithm has higher recognition accuracy and quicker computation, and it reduces the sensitivity to noise to prevent the phenomenon of over-fitting.