Abstract:  Now, with the rapid progress of remote sensing technology, remote sensing target recognition has a very important practical significance and wide range of applications whether in civilian or military fields. However, due to high-resolution remote sensing images and other objective conditions, we can’t achieve real-time and accurate target recognition in the process of remote sensing image target recognition. Extreme learning machine (ELM) is a single-hidden layer feedforward neural network (SLFN) with at most N hidden nodes and with almost any nonlinear activation function can exactly learn N distinct observations. It should be noted that the input weights (linking the input layer to the first hidden layer) and hidden layer biases need to be adjusted in all these previous theoretical research works as well as in almost all practical learning algorithms of feedforward neural networks. ELM has the advantage of fast learning speed and is a one-time learning method, which has been widely used in small-sample learning problems. Aircraft and ships as typical artificial objects, having regular shape features, can be regarded as ideal geometric primitives, so aircraft and ship target can be seen as different category shapes. Firstly, to extract remote sensing images feature, and then ELM is applied to remote sensing target recognition, which is an effective way to solve the problem. However, ELM has the disadvantage of profuse hidden nodes (usually the same with number of learning samples) and huge model structure. Profuse hidden nodes inevitably lead to decrease the computation speed of the model. Therefore, the paper firstly puts forward a learning method based on ELM in which the number of hidden layer neurons is determined automatically---adaptive extreme learning machine (adaptive-ELM) which overcomes the disadvantage of profuse hidden nodes in ELM network, and then introduces remote sensing image feature extraction methods. The adaptive-ELM method firstly adaptively determines the number of hidden layer neurons through affinity propagation (AP) clustering sample set, and then make use of the ELM model to classify. AP is a clustering algorithm based on the concept of "message passing" between data points and the objective of AP is to find the optimal set of class representatives. Finally, simulation results are presented to demonstrate the accuracy and practicability of remote sensing target recognition based on adaptive-ELM.