Research On Recognition Method Of Asphalt-Aggregate Ratio For Asphalt Mixture Based On Digital Image Processing

The testing of asphalt-aggregate ratio for asphalt mixture is the important measure of construction quality control and performance evaluation of asphalt pavement. Therefore, aiming at the problems of the commonly used testing method of asphalt-aggregate ratio, such as be time-consuming, strenuosity, low accuracy, pollution and harmful to human health etc. Based on the image processing technology, the image recognition method for asphalt-aggregate ratio and volume of air voids of asphalt mixtures has been researched in this thesis.(1) This thesis selected five, kinds;o:f asphalt mixture with different aggregates, binders and gradations:top surface layer of 1# SBS modified AC-13C, mid surface layer of 2# SBS modified AC-20C and 3# ordinary AC-20C, lower surface layer of 4# and 5# ordinary AC-25C. The Marshall test method has been put to use in the mix proportion design of asphalt mixtures, and the five determined results of optimum asphalt-aggregate ratios are 5.1%,4.4%,4.3%,3.6% and 3.6%, respectively. The rationality of mix proportion results of asphalt mixtures has been verified through experiments such as the wheel tracking, the freeze-thaw splitting, the immersion Marshall testing, the low temperature bending and the texture depth testing etc.(2) The standard Marshall specimens of different asphalt mixtures are shaped by compaction at different asphalt-aggregate ratios. The horizontal cutting has been done in these specimens to get more slices, and digital photos of sections for different asphalt mixtures were photographed with a digital camera. The image segmentation has been carried out with the thresholding method. The rectangle section asphalt mixture images sized at 900x900 which centered at the center of the specimens have been extracted after mage denoising and extraction etc.(3) The gray frequency distribution curves in the section images were obtained after gray processing to extracted images. The results show that there is a typical double-peak distribution feature for gray frequency distribution curves, and this "double-peak and single valley" feature is irrelevant to the mixture types. The semi-branch curve of left peak changes from sharp to full, the range of gray value it covered becomes large and the low step phenomenon disappears gradually with the increase of asphalt-aggregate ratio of asphalt mixture. So, the semi-branch curve of left peak can reflect the aggregate ratios, and the low step phenomenon relative to volume of air voids.(4) After fitting analysis for the left half branch curves of gray frequency distribution, the reasonable ranges of gray value for asphalt and air void have been determined. The recognition values of many images under different asphalt-aggregate ratio of five kinds of asphalt mixture have been calculated, respectively. The results show that the semi-branch curve of left peak of gray frequency distribution curve can be seen as the superposition of weighted double normal distribution probability density function curves, and the multiple correlation coefficient reaches 0.99. There is a significant linear correlation between the obtained recognition asphalt area ratios and the practical data, so does the void area ratio, and all the correlation coefficients are more than 0.9. However, the linear relations are different in different asphalt mixtures. Accordingly, the results of image recognition were verified by the correlations, and the absolute errors between the recognition values and the practical values are generally within 0.2%. It shows that the results of image recognition have higher precision to meet the demands of engineering application.In conclusion, the application of image processing technology is feasible to recognize asphalt-aggregate ratio and volume of air voids of different asphalt mixtures. The research can provide a quick and simple test method for the engineering application.