Research On The Key Technologies Of Multi-spectral IFF System

Identifing the attributes of targets by IFF system is an effective way to avoid fratricide. Traditional cooperative IFF system which identify attributes of targets by interrogator and transponder, is vulnerable to be interfered, deceived, and can not recognize the neutral targets. Non-cooperative IFF system based on multispectral imaging technology which regards targets as the external environment identifies the attributes of targets by analyzing their features and solved the problems of cooperative IFF system.This paper researchs the key technologies of multispectral IFF system deeply. First, image fusion, global and local feature extraction, and other key technologies of multispectral IFF system are studied theoretically and verified experimentally.(1), proposes a multispectral image fusion method based on non-subsampled contourlet transform by analyzing the characteristics of the target image in different spectral. For low frequency coefficients, when the correlation of low frequency coefficients of different spectral images is high, an averaging method is used for fusion, otherwise the weighted gradient phase congruency is used for fusion when the correlation is low, which preserved the structural information of the images; a Gaussian-weighted SML is used for high- frequency coefficients fusion to retain the image details. In this paper, the visible image and infrared image fusion is performed for an example, experiments verify the effectiveness of the proposed algorithm.(2), in terms of global feature extraction, Charlier moments is constructed by orthogonal C harlier polynomials and its affine invariant is derived in this paper, and a feature extraction method integrated a variety of discrete orthogonal moments invariants is proposed, which overcame the information redundancy of non-orthogonal moments invariants and discretization error of continuous orthogonal moments. Classification experiments show that the proposed method with respect to a single feature has higher classification accuracy.(3), in terms of local feature extraction, a corner detection algorithm based on background suppression Gabor energy filter is proposed, and extract the characteristic corners which succinctly reflect the nature of the target shape features on the constraint of target skeleton, then the redundant corners is removed; for the lack of affine invariance of multi-scale Gaussian kernel integral invariant, image normalization method is adopted to obtain affine invariance in this paper, and select scales according to the rate of change of the shape feature, which reduced the information redundancy between the adjacent scale integration invariant. The local feature vector based on bag-of-words model is built for target classification. Example shows that the corner descriptors and multi-scale Gaussian kernel integral invariants proposed this paper are superior with respect to other corner descriptors and contour descriptors.Secondly, based on analyzing of the computational complexity of the key algorithms, a method of extracting the minimum bound ing rectangle area of the target is proposed, which decreased the amount of data to be processed in the algorithms; as for the convolution involved in image fusion and feature extraction, blocking convolution is used to reduce the computation time and improve the efficiency of online identification.Finally, a hardware platform based on multi-processor parallel processing architecture is designed according to the technical indicators of multispectral IFF system, semi physical commissioning for the IFF system is completed. The real-time of the system is verified through fusing visible image and infrared image, detecting and describing features fo r real targets and recognition of object attribute verified the effectiveness of the proposed algorithms.