The Research On Fogging Algorithm Of UAV Images

In recent years,with the development of science and technology,the unmanned aerial vehicle(UAV)vision navigation system has been widely used in military reconnaissance,earthquake relief,traffic monitoring and other fields.But when UAV performs outdoor missions,the influence of the fog weather,often leads to a loss of image contrast,image quality degradation,edge and texture information fuzzy,and even serious color distortion,which makes it difficult for the later target recognition and information extraction etc.,and affects the utility of UAV vision navigation system in outdoor play.Therefore,in order to reduce the impact of fog weather on UAV vision navigation system,UAV image defogging has a strong theoretical and practical significance.Based on the atmospheric scattering model,this thesis studies the UAV image fogging algorithm and applies it to UAV vision navigation system to ensure the normal operation of UAV vision navigation system in fog weather.The specific research contents are as follows:(1)Dark channel prior defogging algorithm based on image segmentationAiming at the problem of color distortion and darkness of bright primary colors in the bright region,dark channel prior defogging algorithm based on image segmentation is proposed.Firstly,the image is divided into the sky region and the non-sky region by the method of adaptive segmentation,and the transmittance is estimated respectively.Then the transmittance of the sky region and the non-sky region is combined,and the foggy image is reconstructed by the fog-day degradation model,which improves the problem that the processing results of uniformly estimated transmittance is not ideal in the bright area Atmospheric light value is re-estimated by the quad tree search method to reduce the estimated error;Finally,we enhance the edge of the restored image and the get better visual effects image.(2)Image defogging based on wavelet and atmospheric dissipation functionThe method combines the wavelet transform and the atmospheric dissipation function to perform fast defogging of the image.Firstly,the first-level wavelet decomposition of the fog image is carried out because of that fog mainly affects the low frequency region of the image.The low frequency components are de-fog-treated and the atmospheric light is estimated by weighted filtering.Finally,the new low-frequency components are reconstructed with other high-frequency components with wavelet coefficients,and the final de-fog image is obtained.By observing the experimental results of fog images on different occasions,it can be seen that the algorithm can eliminate the "white edge" phenomenon of the sudden change area of the depth of field better and has a good effect.(3)A haze removal method based on the combination of dark channel prior and Retinex algorithmFrom the angle of image restoration and image enhancement,a new method based on the combination of dark channel prior and Retinex algorithm is proposed.Firstly,we use dark channel prior to removal the effect of fog in essence,which enhances the image details.However,due to the brightness of the target scene is usually less than the atmospheric light,the restored image is usually overall dimmed and has a poor visual effect.In order to further improve the quality of the image,the SSR algorithm is used to enhance the image,adjust the brightness of the image and improve the contrast of the image.Experimental results show that the proposed method improves the detail information of the image,adjusts the brightness of the image,and improves the contrast of the image.The results are clear and natural.Compared with the dark channel prior defogging algorithm based on image segmentation,this method has a better defogging effect when dealing with complex,colorful and textured images.