Research On Image Preprocessing Method For Seawater Pearl Recognition

Seawater pearls are often produced in the tropical and subtropical shallow waters, the unique growing environment created their unique physiological characteristics. As China’s only tropical island province, Hainan is an important seawater pearls breeding bases, but the seawater pearls are usually identified and classified by visual inspection in our province, this method has many shortages such as low separation efficiency, poor accuracy. It is difficult for customers to judge the quality of the seawater pearls only by their experiences. Therefore, designing a seawater pearl automatic identification system has an important practical meaning.The identification system usually contains five modules:image acquisition, image preprocessing, feature extraction and fusion, the establishment of the database and indexing, the retrieval and comparison. The image preprocessing includes image de-noising, enhancement, segmentation and so on, which is the basis for the realization of automatic recognition. Image preprocessing is directly related to the quality of seawater pearl image feature extraction, therefore, investigating the suitable method and technology for seawater pearl image preprocessing is very important for the construction of automatic recognition system.Aiming at the issues in seawater pearl image preprocessing, the following questions are discussed theoretically and proved by experiments. The first is about seawater pearl image de-noising, by analyzing the noise types caused by image acquisition and the process of the mobile communication, discusses the mean filter, median filter and Gaussian filter, the experiment results show that the median filtering de-noising method is better for seawater pearl image de-noising. The second is about pearl image edge detection, by analyzing and comparing the Roberts, Sobel, Prewitt, Canny and Log edge detection method, proves that the Canny operator based on edge feature extraction of seawater pearl is more effective. The last is pearl image segmentation, on the basis of analysis and testing, explored the application of the region growing method in seawater pearl image segmentation. This method contains pixel gray mean and standard deviation parameters. A region growing algorithm based on local gray adaption is presented, in this algorithm, the threshold of region growing criteria can be selected adaptively according to the local gray distribution features and segment the target area from the complex image. The experiments show that this method is effective.