Construction method of natural scene data set suitable for machine vision

• Main Authors: ZHANG JIE, CHEN LIQIONG, WEI WENLAN, FAN CI'EN, WANG JIALE, ZOU LIAN, CHENG MOFAN
• Format: Patent
• Language: Chinese; English
• Published: 05.11.2019
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention discloses a construction method of a natural scene data set suitable for machine vision. The construction method comprises the following steps: determining a classification category of the data set based on an ecosystem type; downloading an original image from the Internet by using the keyword; carrying out preliminary matching judgment on the downloaded images, and carrying out secondary classification on the fuzzy category images to form an image data set; and verifying the data set by using a convolutional neural network, and combining the ambiguous categories according to theconfusion matrix. The problem that natural scene classification is difficult to perform accurate and objective definition like object classification is solved, reliable data support is provided for training a deep convolutional neural network, and the method can be used for a natural scene recognition system. 本发明公开了一种适用于机器视觉的自然场景数据集的构建方法，包括如下步骤：基于生态系统类型确定数据集的分类类别；利用关键词从互联网上下载原始图像；对下载的图像进行初步匹配判断，对模糊类别图片进行再次分类，形成图像数据集；利用卷积