Research About Potato-shaped Detecting And Algorithm Based On The Hyper-spectral Image Technology

Potato is one of the main products as well as an important economic crop in Ningxia. The quality detecting of potato is the significant link in its process, thus the shaped detecting of potato is the necessary step in order to improve the economic benefits of potatoes. Based on the Hyper-spectral Image Technology, this paper accesses to the image of potatoes and starts the research from the potato-shaped detecting and its identification. The main contents and the conclusion are as follows:①The paper classifies the potato into two types------optimal shaped potato and abnormal potato. Output the gray image of potato with the ENVI software. Then proceed the preliminary image processing operation of the potato, such as filtering process, threshold segmenting process and edge detecting process.②Based on the potato-shaped detecting method of the wave of relative moments, the paper uses the wave to detect the edge of the image and obtains the edge image of potato. Then with the formula of relative moments, it calculates its10relative moments’ results which formed into feature space. After normalized the feature space, the paper classifies the potato with the Euclidean distance. The classification accuracy of the optimal shaped potato and abnormal potato are85%and90%respectively.③Based on the potato-shaped detection of classifying method of polar Fourier descriptor, the paper gets the closed edge information of potato according to the edge detecting and calculates the polar function by using the image barycenter as its limit. Then with a random point in the image, it uses the3x3neighborhood distance to take the sample point by point and calculates the periodic function by putting the circumference as a cycle, thus extracting the registration information after the transformation of the Fourier. Later, it classifies the image according to the Euclidean distance result of the registration information of each two images.④After comparing the potato-shaped detecting method based on the wave of relative moments with that method of based on the polar Fourier descriptor, the paper gets the result that the potato-shaped detecting method based on the wave of relative moments has the more stability and instantaneity, which providing the foundation of the accurate classification for the potato.