一种基于多视图投影和深度学习的三维兴趣点提取方法

• Format: Patent
• Language: Chinese
• Published: 29.07.2022
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

本发明公开了一种基于多视图投影和深度学习的三维兴趣点提取方法，包括将标记的3D物体投影到多个2D视图中采集训练数据，构建兴趣点训练概率分布，通过2D图像数据和兴趣点训练概率分布训练神经网络，根据训练好的神经网络和改进的密度峰聚类算法获取概率分布Q，提取3D物体三维兴趣点。通过少量的数据实现3D物体兴趣点的自动检测，既不依赖人为设定的特征描述符，也不依赖大量昂贵的3D训练数据，即能获得令人满意的结果。 The invention discloses a three-dimensional interest point extraction method based on multi-view projection and deep learning, and the method comprises the steps: projecting a marked 3D object to a plurality of 2D views, collecting training data, constructing interest point training probability distribution, training a neural network through the 2D image data and the interest point training probability distribution, obtaining probability distribution Q according to the trained neural network and an improved density peak clustering algorithm, and extracting the three-dimensional interest points of the 3D object. The automatic detection of the interest points of the 3D object is realized through a small amount of data, and a satisfactory result can be obtained without depending on manually set feature descriptors