基于NCNN的特種設(shè)備原始記錄便捷化采集方法的設(shè)計(jì)與應(yīng)用

打開文本圖片集

摘要：文中通過(guò)分析特種設(shè)備原始記錄采集過(guò)程中遇到的問(wèn)題，提出一種基于CNN卷積神經(jīng)網(wǎng)絡(luò)和NCNN高性能神經(jīng)網(wǎng)絡(luò)前向計(jì)算框架的圖像文字識(shí)別方法的應(yīng)用，通過(guò)對(duì)識(shí)別的結(jié)果進(jìn)行近義詞分析、數(shù)據(jù)分類及數(shù)值規(guī)約保證了數(shù)據(jù)的準(zhǔn)確性及識(shí)別結(jié)果的高可用性，從而有效提高檢驗(yàn)人員現(xiàn)場(chǎng)采集原始記錄的效率。

關(guān)鍵詞：卷積神經(jīng)網(wǎng)絡(luò)；高性能神經(jīng)網(wǎng)絡(luò)前向計(jì)算框架；近義詞分析；原始記錄

Design and Application of an NCNN-Based Efficient Collection Method for Original Records of Special Equipment

ZOU Shanqing

（Fujian Special Equipment Inspection and Research Institute， Fuzhou 350008， Fujian， China）

Abstract： By analyzing the problems existing in the collection process of original records of special equipment， it proposes the application of an image text recognition method based on CNN convolutional neural network and NCNN high-performance neural network forward computation framework. This approach ensures data accuracy and recognition result availability through the analysis of semantic proximity， data classification， and numerical statute validation. The method effectively improves the efficiency of inspectors' on-site collection of original records， enhancing overall operational efficiency.

Key Words： Convolutional neural network; High-performance neural network forward computation; Synonyms; Original record

0 引言

近年來(lái)，隨著我國(guó)經(jīng)濟(jì)持續(xù)發(fā)展，特種設(shè)備的數(shù)量與種類也在日益增長(zhǎng)，人機(jī)比矛盾日益突出，因此如何在傳統(tǒng)檢驗(yàn)過(guò)程中應(yīng)用新技術(shù)來(lái)提高檢驗(yàn)檢測(cè)的效率、提升檢驗(yàn)檢測(cè)的質(zhì)量，一直是各檢驗(yàn)機(jī)構(gòu)努力的方向。（剩余5050字）