Infrared target identification method based on unchanged rotary morphology neural net

Abstract

The invention relates to an infrared target identification method based on rotating constant morphology neural network comprising, accomplishing target localized region dividing, normalizing the divided target, extracting target shape information positioning the location of the target in the mode library according to the nearest method, using the output result as the input for sorted network, finally realizing the correct categorization of the target. The invention realizes a longer image-forming distance.

Description

Technical field: [0001] The invention relates to an infrared target recognition method based on a rotation invariant morphology neural network, which can realize the recognition of objects in the air far from an imaging sensor, incomplete target shape information and posture changes. It is an infrared detection and recognition system, A core technology of infrared early warning system and large field of view target monitoring system can be widely used in various military and civilian systems. Background technique: [0002] Infrared imaging technology is a non-contact testing technology, which can easily detect the invisible heat radiation emitted by the target and convert it into a visible image. A key field involved in information acquisition is the research of infrared detection technology and methods, and its important position has become increasingly prominent. Infrared imaging technology is favored because of its good concealment, wide detection range, high positioning accur...

AI Technical Summary

Problems solved by technology

[0005] 1. The target is far away from the imaging sensor. Due to the accuracy of the imaging sensor and other reasons, the target shape information is incomplete, and the traditional target recognition method cannot be applied;

[0006] 2. The vibration of the imaging platform itself and the movement of the target often show that the attitude of the target changes, which increases the difficulty of recognition;

[0007] 3. The amount of data is large and difficult to process in real time

[0008] Researchers at home and abroad have proposed some methods for target recognition, such as template matching, statistical pattern recognition, syntactic or structural pattern recognition, and neural network recognition methods, but when the target information is incomplete, or the target posture changes, it always shows The problem of large amount of calculation makes it difficult to process video image sequences in real time

Images