Research On Extraction Of Potholes Features Based On Digital Image Recognition

The technology is primarily used in the field of road maintenance, which can extract potholes features effectively based on digital image recognition. The study is based on image processing technology. The mobile devices are utilized to acquire potholes images. Then the images feature extraction methods are researched. The main contents in the paper are consisted of three distinct parts. The first one is measuring the actual size of the potholes images. The second is extracting the potholes image edge. And the last one is analyzing and designing the mobile applications, which can acquire potholes images.At present, the methods of measuring the actual size of objects in the images are mainstream, which are based on reference materials or acquisition system. The method based on reference material uses a known size of the object in the images to measure the object size. The method is convenient and flexible, but the reference material should be set every time when measuring the size of objects, what’s worse the reusability is low. The other method based on acquisition system measures the size of the image object by calibrating the acquisition system. Though it has high reusability and accuracy, because of the particular nature of the acquisition system, the measuring system has a lower flexibility, and cannot be carried and moved easily. Therefore, it is necessary to research a method without the reference material, which is utilized to measure the size of the image object with high flexibility, easily. In addition, as for image feature extraction, the image edge information cannot be ignored in all kinds of image features. Therefore, the study of edge feature extraction algorithm for potholes images is also crucial in the field of road maintenance.Firstly, a measurement method for potholes without reference materials is proposed according to research the measurement methods for potholes size. The pothole images are photographed by using the mobile devices, and then the direction and gravity parameters of it are utilized to pre-process and correcting pothole images. Then mathematical models to measure potholes are established and the formulas are deduced. Through the analysis of measuring errors and the verification of contrastive experiments, the results show that the method can achieve high accuracy of measurement and the error is acceptable. Compared with the traditional measurement methods, the problem for relying on reference materials is solved, and it has low dependence on the environment. The more important is that the convenience of information collection is improved obviously.Secondly, through studying the potholes image edge extraction algorithms, a method of image edge extraction with improved PCNN and morphological, named IM-PCNN, is proposed. Optimized the basic PCNN model, simplified the original model parameters, and improved the linear input and the calculation method of the output pulse. The effects on potholes edge from the surrounding environment can eliminate, in a certain extent, after enhancing the image. Then improved PCNN models and dilation & erosion characteristics of morphological are used to extract the potholes edge. Through the verification of contrastive experiments and the analysis of edge evaluation, the results show that the method is more effective than the traditional edge detection method on the road potholes image edge extraction. This method can effectively restrain the influence that the road surrounding environment to the potholes edge, with strong anti-interference ability, and the extracted edges are clearer, available.Finally, the mobile application is analyzed and designed, which used to acquire potholes images. So the proposed method can be implemented in practical engineering applications effectively. The mobile application analysis, design, detailed design and other aspects are expounded in the paper, which will lay the groundwork for developing program in the future.