Detection method and device for small and dim targets in infrared sequence images

Abstract

The invention provides a detection method for small and dim targets in infrared sequence images. The method comprises the steps that low-dimensional small-data image signals are collected according to a generated random projection space, background suppression and small and dim target information preserving processing are carried out on single-frame sub-images in the collected low-dimensional small-data image signals and the original infrared sequence images according to the mahalanobis distance, a segmentation threshold tau is selected according to the infrared sequence images obtained after the processing is carried out, and the small and dim targets in the infrared sequence images are detected according to the segmentation threshold tau. The invention further provides a detection device for the small and dim targets in the infrared sequence images. According to the detection method, the image processing data size is decreased, the detection flow is simple and easy to achieve, the data operation time is shortened, and meanwhile the real-time detection performance of a related detection system can be remarkably improved.

Description

technical field [0001] The invention belongs to the technical field of infrared image processing, and in particular relates to a detection method and a detection device for weak and small targets in infrared sequence images. Background technique [0002] The detection of weak and small targets in infrared sequence images is an important research topic in the field of infrared target automatic recognition. As we all know, in the infrared sequence images containing weak and small targets, the weak and small targets generally appear as dots, occupying only a few to a dozen pixels, and are usually in complex background environments such as clouds, sea surface, sea antenna, and ground, so weak and small targets It is easy to be overwhelmed by background clutter, which brings great difficulties and influences to follow-up target tracking and other work. [0003] At present, for the detection of weak and small targets in infrared sequence images, conventional processing methods ar...

AI Technical Summary

Problems solved by technology

As we all know, in the infrared sequence images containing weak and small targets, the weak and small targets generally appear as dots, occupying only a few to a dozen pixels, and are usually in complex background environments such as clouds, sea surface, sea antenna, and ground, so weak and small targets It is easy to be overwhelmed by background clutter, which brings great difficulties and influences to subsequent target tracking and other work

[0003] At present, for the detection of weak and small targets in infrared sequence images, conventional processing methods are classical filtering methods based on background prediction, such as maximum median filtering, morphological Top-Hat filtering and two-dimensional minimum mean square error filtering, etc. These filtering methods Both can well suppress simple and stable backgrounds and highlight weak and small targets, but for complex and changeable non-stationary backgrounds, the suppression effect of these methods on background clutter is often unsatisfactory, and the false alarm probability of detecting weak and small targets is high

[0004] The existing technology also uses the idea of ​​discriminant analysis to realize the suppression of background clutter and the extraction of weak and small targets, and the multi-scale image sequence formed after multiple anisotropic nonlinear diffusions of single-frame infrared images containing weak and small targets The RX operator discrimination is used to suppress background clutter and detect weak targets, but the amount of scalar cube formed by it is large, the storage space occupancy rate is high, and the operation and processing time is long; the original sequence image uses generalized maximum likelihood After judging by the comparison criterion, we can get the motion trajectory of the weak and small target, and then complete the detection of the weak and small target in the complex background. This method directly operates on the original sequence image, so the amount of data calculation is large and the complexity of the operation is high. The time-consuming is also relatively long, which will bring pressure on data processing and storage to the infrared early warning system, and reduce the real-time detection performance of the system; after performing multi-scale wavelet decomposition on a single frame of infrared images containing weak and small targets, all its sub- Composing a data cube with an image, and then using the RX discriminant operator to suppress background clutter and extract weak and small targets, so as to realize the method of weak and small target detection. This method is mainly proposed for the detection of weak and small targets in single-frame infrared images. Among them, the data cube to be processed not only has a large amount of data and requires a large amount of storage space, but also has relatively high computational complexity when processing these data, which has high requirements for the design and operation of the detection system.

Images