target identification by remote control system at borders

Border surveillance poses significant challenges, especially in vast and rugged terrains. despite substantial research in target identification, reliable identification remains difficult at borders due to factors such as camouflage and background similarity. this paper presents an intelligent target identification algorithm for remotely controlled environments, leveraging statistical and sparse dictionary-based features. the input image is segmented into super-pixels, from which discriminative features are extracted. multiple background sparse dictionaries are then generated, and an update class is assigned to each based on the sparsity of the super-pixel representations. these classes guide an iterative refinement of the dictionaries. final target identification is performed using a joint criterion based on sparse coding errors and representation coefficients. simulation results on a custom dataset show the proposed method achieves a mean accuracy of 96.8% and outperforms several existed methods in border-specific scenarios