Research And Application Of Automatic Detection Technology For Bridge Bottom Defects Based On Machine Vision

In the process of bridge construction,the service life of the bridge is reduced by the existence of bad operation such as uneven pouring,improper mix of concrete,and the diseases such as dew,honeycomb hemp and cracks which have been exposed to the changes of climatic conditions and natural environment,different loads and so on for a long time after being put into use.The identification and location of diseases is the key technology to realize the automatic detection of bridges.Aiming at the difficulty of disease identification and feature extraction caused by the interference factors such as uneven illumination,surface shedding and stains in the image of bridge bottom disease,this paper studied the image segmentation and feature extraction algorithms of dew,honeycomb hemp and crack under complex background,and mainly completed the following work:(1)Aiming at the problem that the surface breakage or seepage phenomenon affected the image threshold segmentation in the dew-bar image,an automatic threshold segmentation algorithm based on projection combined maximum entropy method and improved genetic algorithm(IGA)was proposed,which first removed the background area by projection method,and then discussed the image segmentation based on the maximum entropy method combined with IGA.Finally,the feature information was obtained by morphological denoising,which improved the accuracy and efficiency of feature extraction.(2)Aiming at the problem of increasing the difficulty of image segmentation with the phenomenon of uneven illumination and multi-background in cellular hemp images,an image segmentation algorithm based on the combination of HSI color space and grayscale fluctuation was proposed,and the threshold value of the best image was found by the S component grayscale fluctuation curve combined with the standard deviation,and the image segmentation was completed.By means of area filtering,the noiseacquisition features were obtained,which effectively overcame the influence of light illumination and multi-background working conditions.(3)Aiming at the problems such as the easy loss of detail and the difficulty of extracting crack image segmentation,a bridge crack extraction algorithm based on Hessian matrix combined segmentation and multi-angle projection filtering method was proposed,and on the basis of Hessian matrix joint segmentation,in order to eliminate its accompanying noise,multi-feature filter was first used to remove the fat type noise.Secondly,the fracture crack was stitched together by the similarity principle,and some noise was filtered out by using the longest crack as the benchmark contrast feature.Finally,through the multi-angle projection filtering method to filter out the isolated noise to obtain cracks,through the experiment to verify that the effect of denoising was better than the traditional denoising method,effectively solve the problem of crack extraction difficulties.(4)The visual detection system of bridge bottom defect was developed based on MATLAB,the automatic recognition classification of bridge disease Image was realized,and the requirement of real-time automatic storage and image visualization was satisfied,which provided important information for subsequent graphics processing.