Research On Edge Detection Of Original Oil Tank Bottom Image Based On Infrared Imaging

In order to improve the safety and efficiency of cleaning the large crude oil storage tanks,realizing the AI cleaning,our research group constructs a vision system based on infrared imaging for the complex environment in the sludge cleaning process in crude oil storage tanks.The system realizes the rapid identification and positioning of the oil sludge at the bottom of the tank.The entire sludge recognition and positioning system includes six processing modules including image characteristic analysis,image preprocessing,image edge detection,sludge range judgment,target positioning and parallel acceleration.This paper mainly studies the image preprocessing and edge detection modules and also proposes the improved algorithms to strengthen the quality of infrared images,obtain the clearer and complete contour edges.The paper provides technical support for subsequent research on image registration and target positioning.Firstly,aiming at the overall darkness of the infrared image obtained from the crude oil storage tank,blurring,lack of texture details,etc.,this paper proposes an infrared image based on nonlinear transformation and multi-scale convolution detail enhancement from the aspects of image brightness and contrast.Enhanced algorithm.The algorithm first uses an adaptive nonlinear transformation algorithm to improve the overall brightness of the image;secondly,it uses multi-scale bilateral filtering to decompose the image into a basic layer and multiple detail layers,uses CLAHE to improve the contrast of the basic layer,and uses a nonlinear weighted fusion method.Process multiple detail layers,and fuse the processed two sub-images to obtain a detail image;finally,fuse the brightness image and the detail image to obtain an enhanced image.Simulation experiments show that the algorithm can not only improve the image brightness,but also enhance the image contrast,highlight the texture details,the sense of hierarchy is significantly improved,and the visual effect is better.Secondly,in order to improve the image processing effect of the edge detection link.Aiming at the traditional Canny edge detection operator's sensitivity to noise,the presence of false edges,and lack of adaptability,this paper improves it,using fast guided filtering to preserve the edge and reduce noise of the image,and increase the gradient difference in the diagonal direction.And through the Otsu method based on variance and weight to select the threshold to improve the adaptability of the algorithm.The performance of the improved Canny detection algorithm is significantly improved,but it still cannot improve the influence of noise in the infrared image on the detection effect.Therefore,in order to further improve the noise resistance of the algorithm,this paper introduces mathematical morphology into the improved Canny algorithm,abandons the traditional single-scale processing,and selects three structural elements of different sizes and shapes to perform morphological edge detection on the image,and process it The latter result is merged with the processing result of the improved Canny operator,and an edge image with clear outline and low noise is obtained.Finally,Comparing with the image in noise or not,the algorithm we proposed is proved effective both in qualitative and quantitative.Which is not only good at anti-noise,but also getting the clear,smooth and highlight image edge.