Research On Image Enhancement Algorithm Based On Robot Vision In Complicated Environment

In recent years,machine vision is attracted attention by scholars at home and abroad with the development of robot technology and intelligence.Machine vision is widely used in visual navigation,robot localization,pattern recognition and so on.The image is the important source of machine vision to get outside information and image quality is extremely important.In the process of image acquisition,it will lead to the generation of low illumination image and fuzzy image caused by lack of light,fog haze weather and other environmental factors.In order to improve the visual effect and quality of image,the research of image enhancement algorithm under the condition of low illumination environment and haze weather is necessary.Based on NAO robot experiment platform as the research object,in view of the low illumination environment and haze weather conditions,the image enhancement algorithm is discussed.In this paper,the algorithm is verified through simulation and experimental research and the specific research contents are as follows:Firstly,it is to identify characters under normal light conditions.Pre-processing the acquisition of image,using the method of block and projection to segment character in view of interference and tilt sensitivity issues,the characters are identified based on convolution neural network algorithm.We use the proposed algorithm on the experiment of NAO robot to identify characters on cards,and the experiment shows that the recognition effect is influenced by environmental light and the night character recognition rate is obviously lower than that of the day.Secondly,in view of the dark and low contrast image under low light conditions,the Retinex algorithm based on weighted guide filter is proposed to enhance image,which is combined with the Sigmoid function and the improved Gamma correction function.It can not only improve the brightness and contrast of the image,but also keep the details and restore the color of the image.In order to testify the performance of the proposed algorithm,simulation and experiment are conducted.Finally,for the problem of low quality image whose details are covered under the haze weather,the improved atmospheric scattering model based on the physical model of degraded image is putted forward.The combination of light channel and dark channel algorithm is used to estimate the atmospheric light and transmission,recovering the image without fog.It can solve the color distortion of white sky when recovering the image without fog,and restore the color and details of the image so that the visual effect of image is improved.Then the proposed algorithm is verified through simulation,the quality of the image is evaluated by some parameters.