Research On Visual Feature Extraction Method For Small And Medium-sized Objects In Image

With the rapid development of modern science and technology,image target detection and recognition has gradually become an important and popular research direction in the field of image processing.Feature extraction and matching of the target is an essential stage in the process of image target detection and recognition.However,due to the limitation of imaging distance and observation equipment,the target in the image is often far away from the photographing equipment,resulting in even the attention of The size of the target itself is large enough in reality,but the target area occupies very few pixels in the image,which makes the edge of the target fuzzy and the feature extraction of the target more difficult.In this paper,the visual feature extraction method of the small target image is studied.According to the characteristics of the small target image,the edge feature extraction algorithm is improved.The experimental results show that it can meet the needs and improve the accuracy of target matching.The main research and work of this paper are as follows:1.Firstly,the existing visual feature extraction algorithm is analyzed and studied on the small target image,and the performance of the edge extraction algorithm is evaluated.It is found that when statistical feature extraction and matching are performed on the small target image,the target type cannot be distinguished accurately.When edge feature extraction is performed,the fuzzy edge in the small target image cannot be extracted,so that the image matching is accurate The rate is reduced.2.The edge extraction algorithm based on fuzzy system is studied.Aiming at the deficiency that the conventional edge extraction algorithm can not extract the fuzzy details of the target,the non-maximum suppression of gradient image and the interference removal of input image are proposed.The experiment shows that the improved edge extraction algorithm of fuzzy system in this paper can effectively extract the edge of the small target.3.According to the characteristics of small target image interference,the main line extraction process is designed to solve the problem that there are many interference lines in the edge image and the main line of the target is not obvious.4.The algorithm of shape feature extraction is studied.The main edge of small target obtained by conventional edge extraction algorithm and the main edge of small target obtained by improved edge extraction algorithm are extracted and matched.This paper analyzes the characteristics of small target image,improves the matching effect of shape context algorithm when the shape is deformed and noisy,uses the shape center of gravity as the reference point to describe the shape and uses the distance between vectors as the matching cost.The experimental results show that the matching accuracy,precision and recall rate are improved by using the small target edge obtained from the algorithm flow.