Dehazing Method Based on Fast Pseudosphere Filtering for Visual Navigation of Intelligent Tractor in Haze Weather Scene

In order to make the visual navigation of intelligent tractors adapt to the foggy weather, an image dehazing method using the fast pseudosphere filter is proposed, to get a clear image from a foggy picture with the sky as main background.Firstly, the filter template for the visual images is obtained based on filter parameters by using the reduced rank approximation method.Then, the edge smoothness is guaranteed through pseudosphere filtering to get the exact dark channel and the atmosphere scattering veil.Finally, a clear view of the image is recovered quickly with the correction solution of the atmospheric dissipation model.The experimental results show that this method can improve the objective assessment indicator of the image dehazing by 54.7%, 37.6%, 35.2% and 44% compared with the guided image filtering method, Tarel median filtering method, multi-scale retinex method and the wavelet-based retinex method.Furthermore, the time consumption of this algorithm is about 0.18s, which can meet requirements of the intelligent tractors for real-time visual navigation.