研究目的
To address the issue of segmentation of partially overlapping objects with a known shape in machine vision applications, specifically focusing on objects that can be approximated using an ellipse.
研究成果
The proposed method for segmentation of multiple partially overlapping approximately elliptical shape objects in silhouette images using radial symmetry was shown to achieve high detection and segmentation accuracies and outperformed the competing methods in all datasets. Future work could include the generalization of the method with more complex convex objects.
研究不足
The method assumes that the objects to be segmented are clearly distinguishable from the background of the image and their contours form approximately elliptical shapes. The computational time could be significantly improved.
1：Experimental Design and Method Selection
The method utilizes silhouette images and starts with seedpoint extraction using bounded erosion and fast radial symmetry transform. Extracted seedpoints are then utilized to associate edge points to objects to create contour evidence. Finally, contours of the objects are estimated by fitting ellipses to the contour evidence.
2：Sample Selection and Data Sources
The experiments were carried out using one synthetically generated dataset and two datasets from real-world applications, including crystal particles images captured by transmission electron microscopy and a nanoparticles dataset.
3：List of Experimental Equipment and Materials
Not specified in the paper.
4：Experimental Procedures and Operational Workflow
The segmentation process starts with pre-processing to build the image silhouette and the corresponding edge map. The binarization of the image is obtained by background suppression based on the Otsu’s method. The edge map is constructed using the Canny edge detector.
5：Data Analysis Methods
The performance measures used include True Positive Rate (TPR), Positive Predictive Value (PPV), and average distance (AD) from detected seedpoints to the ground truth object center point.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
