Research on Intelligent Recognition Technology of Aggregate Parameters of Loose Asphalt Mixture Based on Image Processing Technology

Accurate acquisition of aggregate characteristics (shape, size and spatial position) is the basis for in-depth analysis of asphalt mixture grading, uniformity and surface texture. This paper proposes a modified recognition algorithm for loose asphalt mixtures based on digital image processing. The process first converts a True Color RGB image into a binary image during pre-processing. Then a Euclidean distance transform of the binary image is performed, which can be used to get the regional maximum value. In order to avoid over-segmentation caused by the traditional watershed algorithm, a modified watershed segmentation algorithm based on the extended-maxima transform is developed, effectively limiting the number of regional maximums to a reasonable range. Then the watershed ridge lines are superimposed on the original image. Hence, the aggregates are separated correctly, especially touching particles. Seventy images of untreated aggregates consisting of different particle sizes were tested. The results showed that the improved method could effectively segment the particles with accuracy as high as 98%. Finally, the proven methods are programmed and used to identify the particles of loose asphalt mixture. The results showed that the physical information of the loose asphalt mixture aggregates could be adequately recognized, and the accuracy is 96%, which is a solid foundation for the subsequent in-depth research.