Research On Night Railway Images Enhancement Algorithm

With the development of city scale and the completed construction of railway,the train technology has made great progress,while the demand for high security of train operation was raised.It is obvious that objects entering railway track illegally will lead to the decrease of transport efficiency and great threat to human life safety.The existing methods include detecting the object by railway view and image recognition.However,the collected images have low contrast and poor visual effect under low lighting environment,and thus the details of images cannot be clearly presented.As a result,feature extraction,analysis and recognition of images are limited.Therefore,it is of great importance to enhance the images of railway at night.This paper analyzes low-illumination enhanced algorithm based on physical model and non-physical model.The basic process of low light intensity image enhancement algorithms is presented and its principles are analyzed.Compared with traditional image enhancement algorithm,the "low lighting image enhancement algorithm with dark channel prior" has the superiority of higher fidelity and retains more details in the railway environment at night.Some problems such as high exposure rate and failure of strong light source region occurred when applying the exiting "low lighting image enhancement algorithm with dark channel prior" in railway image,which are described in this paper based on the characteristics of railway image and experiment.The main works of this paper are as follows:First,failure cause of application of "low lighting image enhancement algorithm with dark channel prior" in the area of strong light source is analyzed.In addition,with image segmentation based on the transmittance threshold,the area of strong light source is used to enhance the images by local variance.Second,the dark channel prior used in the area of rail is verified in this paper,and the dark channel prior of the railway image is proposed.Based on the brightness and slope features of the railway image,the area of rail is obtained by using of Hough transform and edge extraction algorithm,and the area of rail at night is enhanced by low lighting image enhancement algorithm using dark channel prior of the railway image.Experimental results show that with the proposed algorithm,the contrast of image at night is improved and the details of the image are highlighted.Thus a clearer and natural high-quality image is obtained.