Method for multi-target task instance segmentation

• Main Authors: ZHANG YE, FAN YICHAO, CHEN WEIHUI
• Format: Patent
• Language: Chinese; English
• Published: 24.11.2020
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS

The invention discloses a method for multi-target task instance segmentation. The method comprises the following steps of: 1, performing histogram equalization processing on a sample image; 2, constructing a multi-target feature extraction network; 3, designing a candidate bounding box, and extracting and classifying target objects on the feature map; 4, removing redundant bounding boxes by usinga non-maximum suppression algorithm; and 5, performing binary regression on the candidate bounding box to obtain an instance segmentation result. According to the method, the target object can be segmented from the image by using a binary regression function while the identification precision is improved and the identification efficiency is improved, so that the problem of rough segmentation is solved. 一种用于多目标任务实例分割的方法方法，包括：步骤一，样本图像直方图均衡化处理；步骤二，构建多目标特征提取网络；步骤三，设计候选边界框，并对特征图上目标物体进行提取和分类；步骤四，利用非极大值抑制算法对多余的边界框进行剔除；步骤五，对候选边界框进行二值回归，以获得实例分割结果。本发明能够提高识别精度、加快识别效率的同时使用二值回归函数将目标物体从图像中分割出来，从而解决分割粗糙的问题。