Multi-feature Flower Recognition System Based On Mobile Terminal

With the rapid development of artificial intelligence,image processing level and pattern recognition technology have made great progress,the use of smart phone photo recognition has been applied in the field of flower identification.The flower classification belongs to fine grained category of image classification.On the one hand,the background of flower image taken in natural state is relatively complex,and whether accurate image segmentation can be achieved has a great impact on the recognition results.On the other hand,the shooting angle of flowers,the change of illumination and the incomplete petals will also have a great impact on the recognition effect.At present,several mature flower recognition APPs have been applied,but there is still a lot of room for optimization in performance.The optimization of recognition rate and recognition speed needs further research.In this paper,a multi-feature flower recognition system based on mobile terminal is implemented.On the basis of extracting the color,shape and texture features of the corolla,the shape features of the stamen region are extracted,and the above features are fused according to the idea of multiple kernel learning to realize the classification and recognition of flower images.This paper mainly carries out research work from the following aspects:Pretreatment.This paper uses GrabCut algorithm based on SLIC to segment the image.In order to reduce the computational complexity of GrabCut algorithm,the image is first divided into super-pixels and then segmented.Feature extraction.In the feature extraction stage,low complexity features are selected to represent the characteristic information of flowers,color histograms are used to represent the color features of flowers,Hu moment invariants and geometric features are extracted to represent the shape features of flower images,and gray level co-occurrence matrix is used to represent the texture features of flower images.Inorder to improve the recognition rate of blooming flower species(including flower stamen regions),the stamen region is introduced.Shape characteristics.Feature fusion.A multi-feature fusion algorithm based on multiple kernel learning is used to describe the features of flower images.The features mentioned above are fused with multiple kernel learning ideas to realize the complementary advantages of multiple features and enrich the expression of image features.The experimental results show that the recognition rate of this algorithm is significantly higher than that of simple splicing fusion,and it can also achieve good classification results for blooming flower species with similar color and shape.On the basis of the above work,a plant flower recognition simulation system based on Android is designed and implemented.The system basically realizes the function of collecting images by mobile camera,preprocessing and recognizing flower images.