Three-Dimensional Plane Target Based On Neural Network Recongnition

Stage of now,whether modern warfare or in the national defense,air target detection and early warning of the attack command is the key to the war situation the decision.Three-dimensional object recognition key technology has become the world technology frontier by them applied,and caused widespread concern of scientists and research.Of which three-dimensional plane as a key target recognition technology can be widely used for military target reconnaissance, missile guidance, warning systems, defense systems, imaging radar, etc.It can also be used for airport security and management, and optical imaging equipment,and on in military civilian will have very extensive development in the future.The subject of three-dimensional object recognition technology research carried out,in which the object of study is three-dimensional image information of objects.I built an image acquisition system to collect the object model aircraft as an experimental three-dimensional images.CCD image acquisition through a reasonable target under test airplane model library,and made a more effective image preprocessing.Which uses a filter, edge detection, thresholding methods such as the image is converted to data, to overcome the difficulties to identify the next step.The main problem is to use different operators for edge detection, and compare, select the Prewitt operator for processing.From the experimental data, the choice of Prewitt operator is the best treatment in the pictures.In binary image processing used in Iterative Image optimal segmentation threshold,be different with other programs.I calculated the image of the moment invariant features, the establishment of the original aircraft model database as input to identify.Finally, the knowledge of neural networks in many network systems, the design of self-organizing feature map neural network, compared to other programs faster, more accurate identification results obtained.Basically reached the three-dimensional image based on the input of the aircraft models to achieve the purpose of identification.