Abstract: In this second part of the research, the definition of form proposed in the first part is used for vehicle classification. The background information for this task solution is the images generated by optical-electronic systems of aircrafts. Obviously, the choice of a particular trait in particular the shape, starts with the study of its information content. In this study, under the informative form its ability to provide the correct classification of objects is meant. Therefore, the value of the information content is presented by the probability of correct classification. The presented results of computer experiments concerning the assessment of the information content for the mixture consisting of selected objects showed a high information content of a form sign. The set of objects is meant by the scene. Each object is a fragment of a homogeneous random field with its mean value and correlation function. It is assumed that an object correlation function tends to zero as the distance increases between random variables. It allows to evaluate the unknown arithmetic mean value of an object, calculated according to its image. From these assumptions, it follows that a scene from a formal point of view, is a locally homogeneous random field. The projection of an object, developed as the result of segmentation according to scene presentation usually differs from the previously prepared projection from the base of projections. This stems from the fact that some projection pixels may be erroneously classified to the background and some background pixels may also by erroneously classified to the object. The magnitude of discrepancy depends on the properties of the scene, determining its complexity. To assess the complexity, it is proposed to use the probability of correct projection form recognition, developed according to the scene image. The research provides, the results of computer experiments to assess the complexity of a scene, depending on the magnitude of a correlation radius and the signal/noise ratio. They show that despite the high potential of informative features of the form the classification result depends largely on the quality of the image that reflects the actual properties of the scene.