Eagle-eye-imitating adaptive mechanism-based unmanned aerial vehicle sea surface small target identification method

Abstract

The invention discloses an eagle-eye-imitating adaptive mechanism-based unmanned aerial vehicle sea surface small target identification method. The method comprises the following steps: 1, modeling aneagle-eye-imitating light and shade adaptive mechanism; step 2, modeling an eagle-eye-imitating color adaptation mechanism; 3, modeling an eagle-eye-imitating foreground and background adaptation mechanism; 4, performing amplitude phase spectrum reconstruction on the image after eagle-eye-imitating adaptation mechanism adjustment; 5, calculating a multi-scale gray difference; step 6, calculatingsignificant information and normalizing the significant information; and step 7, outputting a saliency map of the unmanned aerial vehicle of the eagle-eye-imitating adaptive mechanism for identifyingthe sea surface small target. The invention has the advantages that: 1) an eagle eye mechanism is introduced into a small target detection process, so that the capability of capturing preys in variousenvironments in thousands of meters of high altitude of an eagle can be better mapped to the task requirement of an unmanned aerial vehicle for long-distance identification and tracking of the unmanned ship; 2) the unmanned aerial vehicle can greatly adapt to various sea surface environments, such as sea surface scale light, sunny days, cloudy days, dusk and the like, by using an eagle eye self-adaptive mechanism;.

Description

technical field [0001] The invention relates to a research method for long-distance small target recognition on the sea surface of an unmanned aerial vehicle based on biological vision, in particular to a small target recognition method on the sea surface of an unmanned aerial vehicle imitating an eagle-eye adaptive mechanism, which belongs to the field of computer vision. Background technique [0002] Islands are an important guarantee to ease the development of marine economy and coastal areas, as well as the frontier for safeguarding maritime rights and national security. The relatively complex topography and landforms of these islands make it difficult to obtain basic geographic information. The traditional manual measurement method is difficult to complete the collection task of complex basic terrain data, and there are many problems such as heavy workload, serious loss, and personnel safety. Therefore, an intelligent platform is needed to fill the geographic data gap ...

AI Technical Summary

Problems solved by technology

Although the deep learning algorithm has good performance in small target detection, it needs to manually calibrate a large number of samples, and the training process is quite time-consuming

With the gradual prominence of the advantages of biological vision, visual attention methods such as Itti method based on biological vision principle, visual saliency algorithm based on graph and spectral residual (Spectrum residual, SR) method have been applied to target detection, but most of the methods are not sensitive to noise. The signal is more sensitive, resulting in a higher false alarm rate for small targets

Images