Research On The Algorithms Of Gravels Coverage Measurement Of The Synchronous Pavement Surface Dressing

Synchronous Pavement Surface Dressing(SPSD) is a find of advanced road constructing technology, and the gravel covered rate is an important parameter of SPSD. The traditional methods to measure the gravel covered rate usually depend on observation or counting by human naked eyes. These kinds of method are rough, inefficient, affected by human subjection factor and hard to be involved in an automatic control system of construct machine. Meanwhile, the image processing based technology of industrial inspection or control is growing rapidly. Studying and researching on the image processing algorithms, for realizing the measurement of the gravel covered rate, is the main purpose of this paper.Firstly, the image pro-processing algorithms which is necessary for the gravel covered rate measurement system is studied, including geometrical distortion restoration and image enhancement. Then, the research and experiment about the implementation of morphological processing on image of gravels. Eventually, a new method based on watershed and morphology features is presented. This segmenting and merging method can be used for granular objects counting.The main work can be summarized as the following three aspects:1. For restoring the geometrical distortion caused by the lens, people always interpolate new pixel value according to the position getting from the geometrical model of the distortion. The interpolation will make different influence on different position of the image, because these is different intensity of distortion in the image(slight distortion in the middle and strong ones far away from the middle). In this work, we use different enhance parameter in different position in the image, according to the frequency analyse related to the distortion.2. The morphological image processing techniques are studied in the second part on this work, including Region Growing, mark based watershed algorithm and watershed algorithm based on distance transform. We compare the performance of these methods on gravels image by experiment. And we point out that a set of morphological erosion process can be identical with the watershed algorithm based on distance transform, in terms of counting task.3. For the task of segmentation and counting of granular objects with low quality images, a new method which based on watershed and morphology is presented. First, an oversegmented result is gotten by conducting the watershed algorithm to the smoothed image. Then the region merge stage, which is directed by using the morphological features defined in this article, as well as weighted Mahalanobis distance and region adjacency graph(RAG), takes place. Meanwhile, property of weighted Mahalanobis distance and mechanisms to reduce computational complexity without deterioration of the performance is discussed. Experiments with real images demonstrate the validity of the method, that it is invariant under translation of gray scale, contrast and noise level. And it performs better than any present method, in terms of precision-recall criterion.At the end of the article, a summary of my work are given and the future work are discussed.