Martins E. Irhebhude, None Adeola O Kolawole, None Michael Chinonye Izuegbu
{"title":"尼日利亚道路上的车辆制造和模型识别的手工制作和转移学习特征技术","authors":"Martins E. Irhebhude, None Adeola O Kolawole, None Michael Chinonye Izuegbu","doi":"10.37231/myjas.2023.8.2.379","DOIUrl":null,"url":null,"abstract":"The vehicle makes and model recognition (VMMR) is a challenging task due to the wide range of vehicle categories and similarities between different classes. Studies have shown that works have recognized vehicles of different countries' make and models. Popular vehicles on Nigerian roads may include products like; Toyota, Honda, Peugeot, Benz, Innoson Vehicle Manufacturing (IVM), etc. The VMMR is important in the intelligent transport system hence, this paper presents a handcrafted and transfer learning model to detect stationary vehicles and classify them based on brand, make, and model. A new dataset was introduced consisting of selected images of popular brands of vehicles driven on Nigerian roads. Framework for a vehicle make and model recognition was developed by extracting features using EfficientNet and HOG models and evaluated on the locally gathered datasets. For classification, a linear Support Machine Vector (SVM) was used. Experimental results showed 94.5% on HOG, 97% with EfficientNet, and 98.1% accuracy when HOG and EfficientNet features were concatenation. The proposed concatenated model outperformed HOG and EfficientNet extracted features by providing higher accuracy and confusion matrix with the highest number of classified images. The study shows the advantages of the proposed model in terms of its accuracy in terms of identifying the vehicle make and model.","PeriodicalId":18149,"journal":{"name":"Malaysian Journal of Fundamental and Applied Sciences","volume":"6 1","pages":"0"},"PeriodicalIF":0.8000,"publicationDate":"2023-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Handcrafted and Transfer Learned Feature Techniques for Vehicle Make and Model Recognition on Nigerian Road\",\"authors\":\"Martins E. Irhebhude, None Adeola O Kolawole, None Michael Chinonye Izuegbu\",\"doi\":\"10.37231/myjas.2023.8.2.379\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"The vehicle makes and model recognition (VMMR) is a challenging task due to the wide range of vehicle categories and similarities between different classes. Studies have shown that works have recognized vehicles of different countries' make and models. Popular vehicles on Nigerian roads may include products like; Toyota, Honda, Peugeot, Benz, Innoson Vehicle Manufacturing (IVM), etc. The VMMR is important in the intelligent transport system hence, this paper presents a handcrafted and transfer learning model to detect stationary vehicles and classify them based on brand, make, and model. A new dataset was introduced consisting of selected images of popular brands of vehicles driven on Nigerian roads. Framework for a vehicle make and model recognition was developed by extracting features using EfficientNet and HOG models and evaluated on the locally gathered datasets. For classification, a linear Support Machine Vector (SVM) was used. Experimental results showed 94.5% on HOG, 97% with EfficientNet, and 98.1% accuracy when HOG and EfficientNet features were concatenation. The proposed concatenated model outperformed HOG and EfficientNet extracted features by providing higher accuracy and confusion matrix with the highest number of classified images. The study shows the advantages of the proposed model in terms of its accuracy in terms of identifying the vehicle make and model.\",\"PeriodicalId\":18149,\"journal\":{\"name\":\"Malaysian Journal of Fundamental and Applied Sciences\",\"volume\":\"6 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.8000,\"publicationDate\":\"2023-10-31\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Malaysian Journal of Fundamental and Applied Sciences\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.37231/myjas.2023.8.2.379\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q3\",\"JCRName\":\"MULTIDISCIPLINARY SCIENCES\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Malaysian Journal of Fundamental and Applied Sciences","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.37231/myjas.2023.8.2.379","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"MULTIDISCIPLINARY SCIENCES","Score":null,"Total":0}
Handcrafted and Transfer Learned Feature Techniques for Vehicle Make and Model Recognition on Nigerian Road
The vehicle makes and model recognition (VMMR) is a challenging task due to the wide range of vehicle categories and similarities between different classes. Studies have shown that works have recognized vehicles of different countries' make and models. Popular vehicles on Nigerian roads may include products like; Toyota, Honda, Peugeot, Benz, Innoson Vehicle Manufacturing (IVM), etc. The VMMR is important in the intelligent transport system hence, this paper presents a handcrafted and transfer learning model to detect stationary vehicles and classify them based on brand, make, and model. A new dataset was introduced consisting of selected images of popular brands of vehicles driven on Nigerian roads. Framework for a vehicle make and model recognition was developed by extracting features using EfficientNet and HOG models and evaluated on the locally gathered datasets. For classification, a linear Support Machine Vector (SVM) was used. Experimental results showed 94.5% on HOG, 97% with EfficientNet, and 98.1% accuracy when HOG and EfficientNet features were concatenation. The proposed concatenated model outperformed HOG and EfficientNet extracted features by providing higher accuracy and confusion matrix with the highest number of classified images. The study shows the advantages of the proposed model in terms of its accuracy in terms of identifying the vehicle make and model.