Research On Snowflake Shape Recognition And Classification Algorithm

Snow quality is an important factor to determine the performance of the Winter Olympic snow games.The shape is the main feature of snow.At present,in domestic and foreign research,the shape of the snowflake is mainly analyzed by manual method.This method has a large workload and poor data accuracy.In response to this problem,image processing and pattern recognition techniques are used to extracted the features of snowflake contour and snowflake texture and automatic classification.The snowflake image is processed by median filtering,binarization and morphological closed operation to remove textures within the snowflake image,and extract the edges,convex hulls,and minimum bounding boxes of the snowflake.Based on these parameters,the profile characteristics of the aspect ratio,the squareness,the area ratio,the perimeter ratio,the shape parameter,the density,and the invariant moments are calculated.The gray level co-occurrence matrix of the snowflake image is calculated.The energy,entropy,contrast,correlation variance and mean of the snowflake image are obtained to calculate the eight snowflake texture feature parameters.These jobs provide reliable data support for snowflake shape classification.After analysis and research,the support vector machine and BP neural network are used to design the classifier.The principles of the two algorithms are analyzed and a classifier model is established.Snowflake outline features and texture features are classified separately.Both classifiers are able to achieve higher recognition rates.The comparative analysis results show that the recognition and classification of snow texture using support vector machine classifier is faster and more accurate.The BP neural network classifier is better for snow contour recognition and classification.The automatic classification of snowflake shape has been completed,which provides a reference value for the study of snow quality.