Research On Multi-feature Target Detection Algorithm Based On Saliency Analysis

The reason why human beings can process a large number of images with different characteristics in a very short time is duo to the selective attention mechanism of the human eye.This mechanism enables people to extract interested objects from a variety of complex scenes,which can effectively reduce the scope of visual processing,thereby greatly saving computing resources.Therefore,it is of great significance to study the human visual attention mechanism and apply it to the fields of computer vision and image processing.However,due to the different color structures and different target characteristics of different images,problems such as large amount of calculation,high background noise,and low accuracy may occur when extracting significant targets of multiple features.In this paper,we focus on the problem of extracting saliency maps with various feature targets and aim to develop an effective method for extracting saliency maps with a variety of feature target efficiently and accurately.The main work are as follows:(1)It mainly describes the research background and significance of the saliency detection of multi-feature targets,also introduces and briefly summarizes the detection models proposed by domestic and foreign scholars at the present stage,as well as analyzes some problems existing in the saliency detection algorithms of the emerging stage.(2)This paper proposes a novel saliency detection algorithm based on hierarchical PCA method.Method Firstly,according to prominently distinct detail,the original image is divided into eight layers.Each layer contains correlated target information of the image layer shared the same feature.Then,the method of grayscale image colorizing is used to transplant color characteristics from source image to hierarchical grayscale image,which purpose is to make the layered image not only reflect the pattern characteristics but also retain the original color feature.After that,PCA technology is used to detect layered images to obtain distinct object's pattern distinctness and color distinctness in the principal component direction.Next,two features are integrated to get the saliency map with high robustness,and to further refine our results,the known priors is incorporated on image organization,which can place the subject of the photograph near the center of the image.Finally,entropy calculation is used to determine the optimal image from the layered saliency map,the optimal map has the least background information and the most prominently salient target.The layered PCA technology can obviously reduce the interference of redundant information,and effectively separate the significant object from the background.At the same time,it accelerates the calculation speed and improves the detection accuracy.(3)A multi-objective detection algorithm combining local quaternion transform domain with global features is proposed.First,the entropy of local information is introduced into the quaternion representation to make it a real part of a quaternion,and then the three imaginary parts of the quaternion are represented by the RG feature,the BY feature,and the brightness feature of the color image.Thereby form a quaternion representation with four feature channels.Then through a Fourier transform to get a rough saliency map.Then,the image is divided into image blocks with the same size,the probability of occurrence of each image block in the entire image is calculated,and the image block with a smaller probability of occurrence is retained as a significant area display to obtain a saliency map considering global contrast.Normalize and multiply the two saliency maps to get the final accurate saliency map.The experimental results show that using the combination of local and global features,the complete target contour can be extracted,and the target detection result with multiple similar features in the graph is more accurate.Therefore,the proposed algorithm can effectively improve the detection accuracy of saliency map,reduce the missed detection rate,and reduce the computational complexity of the algorithm.