熔痕图像分类实验数据集

电气故障熔痕是火灾调查中最为典型的、提取数量最多的、研究最为广泛的痕迹物证，在火灾现场快速、精准判定熔痕的熔化性质，是确定调查方向、构建证据链条、分析起火原因的关键。目前火调人员进行现场勘验时，主要依靠手工清理现场的方式寻找电气线路残骸中的熔痕，依靠个人经验或台式的Cu导线短路熔痕智能识别装置判断熔痕的熔化性质。因此，亟须针对火调人员进行电气线路熔痕专项勘验的实际工作场景，研制穿戴式电气火灾熔痕智能识别装备。本项目基于MobileViT开发熔痕图像分类模型，应用本数据集训练、优化模型，提高模型的电气故障熔痕识别能力。Electrical fault melt marks are the most typical, most commonly recovered, and most widely studied trace evidence in fire investigation. Rapid and accurate determination of the melting nature of melt marks at fire scenes is critical for defining investigation directions, establishing evidence chains, and analyzing fire causes. Currently, when conducting on-site inspections, fire investigators mainly rely on manual site cleaning to locate melt marks in electrical wiring remnants, and judge the melting nature of the melt marks based on personal experience or desktop intelligent recognition devices for short-circuit melt marks of copper (Cu) wires. Therefore, there is an urgent need to develop wearable intelligent recognition equipment for electrical fire melt marks tailored to the actual working scenarios of fire investigators conducting special inspections of electrical circuit melt marks. This project develops a melt mark image classification model based on MobileViT, and uses this dataset to train and optimize the model to enhance its ability to recognize electrical fault melt marks.

搜集汇总

背景与挑战

背景概述

该数据集专为熔痕图像分类实验设计，旨在通过训练和优化基于MobileViT的图像分类模型，提升电气故障熔痕的智能识别能力，以支持火灾调查中穿戴式智能识别装备的研制。数据集包含大量图像文件，格式主要为jpg，数据量约7.35GB，适用于火灾安全与应急救援领域的相关研究。