A Position Extraction Method Combining Visual Saliency and Line Segment Intensity

Abstract

To solve the problems of difficult selection of position samples and high manual dependence of position recognition in most existing researches, this paper proposes a position extraction method combining visual saliency and line segment intensity. In order to verify the effectiveness of the algorithm proposed in this study, the two images were tested, and evaluated the accuracy through quantitative indicators. It can be found that the IoU (Intersection of Union) of the two positions are 0.6658 and 0.5319, respectively, which are all greater than 0.5, indicating the effectiveness of the unsupervised extraction method proposed in this research. The recall rate of the position was all greater than 0.83, indicating that the omission rate of the extracted positions by this method was relatively low, all within 17%. In this study, an unsupervised position extraction method is proposed, which can effectively extract target without training samples, and provides a reliable technical means for rapid unsupervised target recognition.