This paper proposes an adaptive method of detecting objects on the image of an optoelectronic device. The method is based on reconstructing a reference signal-image and forming a statistic in the form of the maximum eigenvalue of the selective correlation matrix for making a decision concerning the detection of an object, using the Neyman-Pearson criterion. The information contained in the images recorded is used when there are no a priori data concerning the background-target situation. A block diagram of the algorithm is given, along with the results of estimating the efficiency index for detecting objects under various conditions.