Infrared dim small target detection method based on first-order partial derivatives in multiple directions

Abstract

The invention relates to an infrared dim small target detection method based on first-order partial derivatives in multiple directions. The method comprises three steps: first, using a facet model to construct a binary cubic function of image grayscale in a small range, solving the first-order partial derivative in each direction in a center region, and designing a convolution template of required coefficients, wherein the operation coefficients can be obtained directly from a convolved image; then, designing an enhanced convolution template through the principle of maximum inner product according to the characteristics of the first-order partial derivative of a small target in each direction to enhance the target in each direction; and finally, fusing the results in all directions by means of multiplication, and suppressing the background as much as possible while further enhancing the target, thus getting a final result. The method can be widely used in dim small target detection of infrared images, and has a broad market prospect and application value.

Description

[0001] (1) Technical field [0002] The invention relates to an infrared weak and small target detection method using multi-directional first-order partial derivatives, belongs to the field of digital image processing, and mainly relates to facet models and target detection technologies. It has broad application prospects in various image-based application systems. [0003] (2) Background technology [0004] The detection of weak and small targets plays a key role in the infrared early warning system. Due to the long imaging distance, targets such as airplanes in the background of the sky and ships in the background of the sea usually become a smaller object. Accurate detection of targets in advance can Take appropriate measures for prevention or deployment. However, due to the large amount of sensor noise and non-uniform interference in the infrared scene, coupled with the strong edges of various clutter in the detection scene itself, the false alarm rate will be too high. Ho...

AI Technical Summary

Problems solved by technology

However, due to the large amount of sensor noise and non-uniform interference in the infrared scene, coupled with the strong edges of various clutter in the detection scene itself, the false alarm rate will be too high. How to effectively detect the target while suppressing the false alarm , has been an important issue

Methods based on maximum-median filtering and maximum-mean filtering (see literature: Dishpandman et al., Maximum-Median and Maximum-Mean Filters for Small Object Detection, International Society for Optical Engineering, Optical Science, Engineering and Instrumentation International Symposium Proceedings, 1999:74-83.(Deshpande S D,Meng H E,Venkateswarlu R,et al.Max-mean and max-median filters for detection of small targets[C] / / SPIE's International Symposium on Optical Science,Engineering,and Instrumentation.International Society for Optics and Photonics,1999:74-83.)) is a classic small target detection method, which replaces the currently operated pixel by selecting the median or mean value of some specific direction positions under the current pixel neighborhood, This is used to filter out small targets, but it is more sensitive to Gaussian white noise, which is easy to cause false alarms. The filtering method represented by it also has the same problem

Some classic small target detection methods based on morphology (see literature: Bai Xiangzhi et al., New Top Hat Transform and its Analysis and Research in the Application of Infrared Weak Small Target Detection, Pattern Recognition, 2010: 43(6): 2145-2156.( Bai X,Zhou F.Analysis of new top-hat transformation and the application for infrared dim small target detection[J].Pattern Recognition,2010,43(6):2145-2156.)) using top-hat transformation to enhance target suppression background, Morphological operations are simple and fast, but when dealing with complex scenes, they are prone to interference such as strong edges, resulting in high false alarms. At the same time, the size of morphological operators is mostly fixed and cannot be adaptively adjusted according to the scene at the same time.

In recent years, some weak target detection methods based on sparse representation have achieved some results (see literature: Gao Chenqiang et al., Infrared block image model for detecting small targets in a single image, American Institute of Electrical and Electronics Engineers Image Processing Transactions , 2013,22(12):4996-5009.(Gao C,Meng D,Yang Y,et al.Infrared patch-image model for smalltarget detection in a single image[J].IEEE Transactions on Image Processing,2013,22 (12):4996-5009.)), but due to the uncertainty of the small target distribution, it is necessary to consider a variety of situations when designing the objective function, resulting in too complicated calculations, and the model lacks interpretability and the effect is limited

Due to the prominence of the gray distribution of weak and small targets, some methods using this difference (refer to the literature: Deng He et al., Infrared small target detection method based on local weight difference measurement, Geography and Remote Sensing Transactions of the American Institute of Electrical and Electronics Engineers, 2016,54(7):4204-4214.(Deng H,Sun X,Liu M,et al.Small Infrared Target Detection Basedon Weighted Local Difference Measure[J].IEEE Transactions on Geoscience and Remote Sensing,2016,54(7) :4204-4214.)) By designing a metric, the metric of the target area can be significantly different from the background area, so as to achieve detection, but due to the complexity and variety of infrared scenes, some simple index metrics are difficult to distinguish the target from the background in detail, so limited effect

Images