Image processing method of multi-spectral camera

Abstract

The invention discloses an image processing method of a multi-spectral camera. The image processing method includes the steps: firstly, obtaining and caching all channel images of a multi-spectral-band camera, extracting one camera as a reference spectral band, carrying out rapid detection processing on the reference spectral band, and finding the position of a target in the image; calculating possible positions and areas of the target in the other two spectral bands, and extracting neighborhood slices of suspicious target positions in all spectral band images in sequence according to the positions; carrying out fine detection on all spectral segment neighborhood slices, and extracting features of each spectral segment target; and fusing the features to construct a target feature vector, performing target association and identification among a plurality of frames by using the target feature vector, and taking the associated target feature vector and identification result as the outputof image processing. According to the image processing method, the target of the reference spectrum is rapidly processed and positioned, so that the processing complexity and resource requirements ofthe multi-spectrum camera image are reduced, and the target detection efficiency and speed are remarkably improved.

Description

Technical field: [0001] The invention belongs to the technical field of image processing, and its main content is aimed at a small target detection algorithm in low-to-signal-to-noise ratio images, and is especially suitable for fast and high-precision detection of air targets under complex backgrounds in multi-spectral remote sensing images. Background technique: [0002] The current small target detection methods are basically proposed for images obtained from single-band (medium-wave, long-wave, or visible) detection, but affected by the optical diffraction limit, under the same detection distance, the detection methods of different wavebands are different. The resolution of the image is different, the longer the wavelength, the lower the resolution. Therefore, the characteristics of the target and the background in the single-spectrum infrared image are relatively small, so it often results in high false alarm rate and low detection rate. Features such as outline, color...

AI Technical Summary

Problems solved by technology

[0004] However, no matter what kind of fusion strategy and structure is adopted, processing the full image area of ​​each spectral segment image separately will lead to a multiplication of calculations, especially in the process of processing visible light images, the number of pixels imaged in the same field of view is often 4 to 10 times that of infrared images. At the same time, it can be seen that the number of target features processed by the detection algorithm is more and the dimension is high, and the proportion of the target area is low, which will cause the detection algorithm to be time-consuming and redundant. Although the detection effect can be guaranteed, the real-time performance cannot be satisfied.

Images