基于YOLO目標(biāo)檢測(cè)的煙草配送合格審查算法設(shè)計(jì)

打開文本圖片集

中圖分類號(hào)：TP391.41 文獻(xiàn)標(biāo)志碼：A 文章編號(hào)：1003-5168（2025）08-0042-06

DOI：10.19968/j.cnki.hnkj.1003-5168.2025.08.007

Abstract： [Purposes] Based on YOLO object detection framework，a tobacco distribution compliance review algorithm was designed to improve the compliance and accuracy of tobacco distribution process by using image recognition technology.[Methods] The algorithm uses ResNet-18 as the backbone network and combines spatial pyramid pooling （SPP） module to extract and fuse multi-scale image features. Through target detection of elements such as signers，tobacco，shelves and counters in distribution images，and setting upan audit mechanism to automaticalydetermine whether the images are compliant，automated review of distribution behavior is realized.The experimentaldata came from the standardized and non-standard distribution images screened by professionals，and the labeled data included elements such as personnel，tobacco，shelves，and counters.[Findings] The experimental results show that the A P value of the model in the category of "tobacco" is 0.893 1，and the m A P of the whole model in the test set is 0.683 9，showing good recognitionabilityand stability.[Conclusions]The research algorithmcan effctively support the inteligent distribution management of tobacco industry. In the future，the robustness and accuracy can be further improved by expanding the data set and optimizing the model to meet the needs of more complex scenarios.

Keywords： YOLO target detection; tobacco compliance detection; tobacco compliance review; image recognition

0 引言

2022年11月9日，習(xí)近平總書記在給2022年世界互聯(lián)網(wǎng)大會(huì)烏鎮(zhèn)峰會(huì)的賀信中指出，“當(dāng)今時(shí)代，數(shù)字技術(shù)作為世界科技革命和產(chǎn)業(yè)變革的先導(dǎo)力量，日益融人經(jīng)濟(jì)社會(huì)發(fā)展各領(lǐng)域全過(guò)程，深刻改變著生產(chǎn)方式、生活方式和社會(huì)治理方式。（剩余6451字）