实例分割和边缘优化算法的研究与实现

摘 要：近年来，实例分割技术正受到越来越多的关注。Mask R-CNN 实例分割方法是实 例分割领域中的重要方法，但是用 Mask R-CNN 方法得到的结果中，每个分割出的实例的边缘 往往不够理想，无法与真正的边缘完全吻合。针对此问题，提出了一种用显著性目标提取方法 得到的结果与 Mask R-CNN 实例分割结果相结合的方法，从而得到更好的实例分割边缘。首先， 利用 Mask R-CNN 对图片进行识别，得到实例分割的结果。然后用 PoolNet 对待检测图片进行 处理，得到图片中的显著物体信息。最后用 PoolNet 的结果对实例分割的掩码图边缘进行优化， 从而得到边缘更好的实例分割结果。经过测试，该方法可以对绝大多数待检测目标较为显著的 图片在一些重要指标上得到比 Mask R-CNN 更好的分割结果。

Research and implementation of instance segmentation and edge optimization algorithm

Abstract: In recent years, the instance segmentation technology has received more attention. Although the Mask R-CNN instance segmentation method is important in the field of instance segmentation, the resultant edge of each instance cannot entirely match the real edge. In order to solve this problem, a method was proposed that combined the result of the salient object extraction with that of the mask R-CNN instance segmentation, so as to produce a better edge of instance segmentation. First, the image was recognized by Mask R-CNN, with the segmentation result obtained. Then PoolNet was utilized to process the detected image, resulting in the salient object information in the image. At last, the edge of the mask image was optimized by the result of PoolNet, attaining a better result of the edge segmentation. After testing, this method can yield better segmentation results than Mask R-CNN for most of images with salient targets in some important indexes.