Research On Commodity Image Retrieval Technology Based On Improved Fourier Descriptoron

At present,the key research direction in the field of image retrieval is to perform accurate image retrieval and positioning from a large number of digital images.Due to cost,time-consuming and subjective issues,the traditional text-based image retrieval technology(TBIR)is not prominent in the face of massive image database resources.In contrast,content-based image retrieval technology(CBIR)can better adapt to this situation,and the retrieval effect is better when faced with a variety of image database resources.CBIR improves the defects of image?based retrieval technology in image retrieval.In image retrieval,the image content is automatically extracted according to the underlying features such as color,texture and shape of the image,so that the whole image retrieval process can be effectively completed.The shape of the product image is considered to be a more advanced feature than the color and texture of the image,because a certain semantic information can be extracted therefrom,so people habitually associate with the target according to the shape of the product image.In the content-based image retrieval process,the shape of the image should satisfy the characteristics of translation,rotation and scale.The traditional image feature retrieval algorithm based on shape features has limitations in the selection of feature extraction methods,and the accuracy of its retrieval results cannot be ensured.In order to improve the performance of commodity image retrieval,it is important to describe the shape characteristics of the product image information.In response to the problem described above,this paper proposes a commodity image retrieval method based on the improved Fourier descriptor.Firstly,pre-process the image and extract the edge of the image;secondly,pass the minimum inertia axis and determine the starting position of the contour,then perform Fourier transform to obtain the feature vector;finally,the Euclidean distance and correlation coefficient are used to calculate the to-be-retrieved The feature similarity measure of the feature vector of the single product image and the feature image of the product image in the entire product image database,and the search results of the product image are evaluated by the average precision ratio(ARP)and the average recall ratio(ARR).The experimental results show that the proposed algorithm is better than the traditional shape feature retrieval algorithm in the retrieval of commodity image retrieval,and the retrieval effect is good.