Perbandingan Algoritma YOLOv4 dan Scaled YOLOv4 untuk Deteksi Objek pada Citra Termal

AZIZAH AULIA RAHMAN; SISLY DESTRI AGUSTIN; NUR IBRAHIM; NOR CAECAR KUMALASARI

doi:10.26760/mindjournal.v7i1.61-71

Perbandingan Algoritma YOLOv4 dan Scaled YOLOv4 untuk Deteksi Objek pada Citra Termal

AZIZAH AULIA RAHMAN, SISLY DESTRI AGUSTIN, NUR IBRAHIM, NOR CAECAR KUMALASARI

Sari

ABSTRAK

Minimnya visibilitas pejalan kaki dan pengendara pada malam hari karena kurangnya pencahayaan pada lampu jalan menyebabkan kecelakaan rentan terjadi pada rentang waktu tersebut. Sistem penglihatan komputer berbeda dengan manusia, semua objek dengan suhu di atas nol dapat memancarkan radiasi inframerah jika direkam menggunakan kamera termal. Dalam penelitian ini penulis mengidentifikasi citra termal dalam bentuk citra RGB dengan algoritma YOLOv4 dan Scaled YOLOv4 sebagai deteksi objek. Performa sistem diukur berdasarkan nilai presisi, recall, f1-score, dan mAP. Eksperimen dilakukan pada dataset citra termal dengan objek manusia. Skenario yang digunakan adalah mendeteksi objek dengan jarak 5m, 10m, 15m, dan 20m. Hasil deteksi didapatkan algoritma Scaled YOLOv4 CSP lebih unggul dengan nilai pengujian precision 94,3%, recall 83,8%, f1-Score 88,7%, dan mAP 86,9%. Hasil tersebut dipengaruhi oleh ukuran citra dan jumlah dataset dari citra training, citra validation, dan citra uji.

Kata kunci: Citra Termal, YOLO, YOLOv4, Scaled-YOLOv4, Deteksi Objek

ABSTRACT

The lack of visibility of pedestrians and drivers at night due to lack of lighting in street lights makes accidents prone to occur during this time. The computer vision system is different from the humans, any object with a temperature above zero can emit infrared radiation when using a thermal camera. In this study, the authors identify thermal images in RGB using YOLOv4 and Scaled YOLOv4 as object detection algorithms. System performance is measured based on the value of precision, recall, f1-score, and mAP. Experiments were carried out on a thermal image dataset with human objects. The scenario used was to detect objects at a distance of 5m, 10m, 15m, and 20m. The detection results show that Scaled YOLOv4 CSP algorithms is the best, based on the test value of 94.3% precision, 83.8% recall, 88.7% f1-Score, and 86.9% mAP. These results are influenced by the size of the image and the number of datasets from training images, validation images, and test images.

Keywords: Thermal Image, YOLO, YOLOv4, Scaled-YOLOv4, Object Detection

Teks Lengkap:

PDF

Referensi

Arif, M., Alandani, R., Arya, M., & Thermal, A. R. (2017). Thermal Vision pada Manusia dengan Pengaruh Terhadap Warna Pakaian. 5th Indonesian Symposium on Robotic Systems and Control, (pp. 243â€“248).

Batchuluun, G., Kang, J. K., Nguyen, D. T., Pham, T. D., Arsalan, M., & Park, K. R. (2021). Deep Learning-Based Thermal Image Reconstruction and Object Detection. IEEE Access, 9, 5951â€“5971.

Bochkovskiy, A., Wang, C.-Y., & Liao, H.-Y. M. (2020). YOLOv4: Optimal Speed and Accuracy of Object Detection.

Hammam, H., Asyhar, A., Wibowo, S. A., & Budiman, G. (2020). Implementasi Dan Analisis Performansi Metode You Only Look Once (Yolo) Sebagai Sensor Pornografi Pada Video Implementation and Performance Analysis of You Only Look Once (YOLO) Method As Porn Censorship in Video. e-Proceeding of Engineering, 7 (2), 3631â€“3638.

Kirimtat, A., & Krejcar, O. (2018). FLIR vs SEEK in Biomedical Applications of Infrared Thermography. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics): Vol. 10814 LNBI (Nomor September). Springer International Publishing.

Kristo, M., & Ivasic-Kos, M. (2018). An overview of thermal face recognition methods. 2018 41st International Convention on Information and Communication Technology, Electronics and Microelectronics, (pp. 1098â€“1103).

Kristo, M., Ivasic-Kos, M., & Pobar, M. (2020). Thermal Object Detection in Difficult Weather Conditions Using YOLO. IEEE Access, 8, 125459â€“125476.

Nowosielski, A., Malecki, K., Forczmanski, P., Smolinski, A., & Krzywicki, K. (2020). Embedded Night-Vision System for Pedestrian Detection. IEEE Sensors Journal, 20 (16), 9293â€“9304.

Rasyid, M. R., Tahir, Z., & Syafaruddin, N. (2019). Digital Image Processing for Detecting Industrial Machine Work Failure with Quantization Vector Learning Method. Journal Pekommas, 4(2), 131.

Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You Only Look Once: Unified, Real-Time Object Detection. IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

Sonka, M., Hlavac, V., & Boyle, R. (2013). Image processing, analysis and machine vision (4th ed.). USA: Cengage Learning.

Tanaji, M. R. (2021). Detection of Human Targets from Thermal Images. International Journal of Engineering Research & Technology (IJERT), 10(02), 590â€“593.

Tran, N. (2018). GLOBAL STATUS REPORT ON ROAD SAFETY 2018. World Health Organization.

Tumas, P., Nowosielski, A., & Serackis, A. (2020). Pedestrian Detection in Severe Weather Conditions. IEEE Access, 8, 62775â€“62784.

Wang, C. Y., Bochkovskiy, A., & Liao, H. Y. M. (2021). Scaled-yolov4: Scaling cross stage partial network. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, (pp. 13024â€“13033).

Wang, S., Zhao, J., Ta, N., Zhao, X., Xiao, M., & Wei, H. (2021). A real-time deep learning forest fire monitoring algorithm based on an improved Pruned + KD model. Journal of Real-Time Image Processing, 18.

DOI: https://doi.org/10.26760/mindjournal.v7i1.61-71