ANALISIS KORELASI RESTORASI VISUAL DAN PERFORMA KLASIFIKASI  METODE IMAGE SHARPENING PADA DAUN TOMAT

Muhamad Agil Fadila; Bambang Irawan; Ahmad Faqih; Raditnya Danar Dana; Adi Rizki Renaldi

Authors

Muhamad Agil Fadila STMIK IKMI CIREBON, Indonesia
Bambang Irawan
Ahmad Faqih
Raditnya Danar Dana
Adi Rizki Renaldi

Keywords:

Penajaman Citra, PSNR, SSIM, Performa Klasifikasi, Korelasi Negatif, sampah plastik, MobileNetV2, fine-tuning, klasifikasi citra, deep learning., Deteksi Penyakit Tanaman

Abstract

Kualitas citra yang rendah sering menjadi hambatan signifikan dalam deteksi otomatis penyakit tanaman berbasis Deep Learning. Penelitian ini bertujuan untuk menganalisis korelasi antara restorasi visual objektif dan performa klasifikasi model CNN melalui penerapan metode image sharpening pada citra daun tomat. Tiga metode diuji: Basic Sharpening Kernel (BSK), Unsharp Masking (USM), dan High Boost Filtering (HBF) pada arsitektur MobileNetV2 dengan pendekatan transfer learning. Evaluasi dilakukan menggunakan dua perspektif: metrik kualitas citra (Peak Signal-to-Noise Ratio/PSNR dan Structural Similarity Index Measure/SSIM) serta metrik performa model (Akurasi dan Loss). Hasil penelitian mengungkap temuan anomali berupa korelasi negatif antara kualitas visual dan akurasi model. Metode HBF mencatat restorasi visual terbaik (PSNR 35.36) namun menghasilkan akurasi terendah (97.73%). Sebaliknya, USM dengan restorasi visual terendah justru mencapai akurasi tertinggi sebesar 98.95%, diikuti oleh BSK sebesar 98.78% yang menawarkan stabilitas model terbaik dengan nilai loss terendah (0.0348). Temuan ini membuktikan bahwa metrik visual standar (PSNR/SSIM) tidak dapat dijadikan prediktor tunggal untuk performa klasifikasi CNN, di mana artefak penajaman yang dihasilkan USM justru memperjelas fitur penyakit bagi model.

References

Bhattarai, S., Ny, J., Wang, L., Bhattarai, S., Ny, J., & Wang, L. (2022). The impact of input data quality on the performance of deep learning models for plant disease detection: A review. Computers and Electronics in Agriculture, 198, 107044. https://doi.org/10.1016/j.compag.2022.10704

Pei, X., Zhao, Y. hong, Chen, L., Guo, Q., Duan, Z., Pan, Y., ..., Pei, X., Zhao, Y. hong, Chen, L., Guo, Q., Duan, Z., & Pan, Y. (2023). Robustness of machine learning to color, size change, normalization, and image enhancement on micrograph datasets with large sample differences. Materials & Design. https://www.sciencedirect.com/science/article/pii/S0264127523005014

Wang, J., Wang, Y., Chen, Q., Wang, J., Wang, Y., & Chen, Q. (2022). Image enhancement techniques in plant disease detection: A survey and comparative analysis. Computers and Electronics in Agriculture, 198, 107086. https://doi.org/10.1016/j.compag.2022.

Chowdhury, A., Rahman, M. I., Kaiser, M. S., Chowdhury, A., Rahman, M. I., & Kaiser, M. S. (2022). Improving Low-Resolution Plant Disease Image Classification Using Super-Resolution and Deep Learning. Expert Systems with Applications, 200, 116909. https://doi.org/10.1016/j.eswa.2022.116909

Mehta, S., Saxena, A., Jain, S., Mehta, S., Saxena, A., & Jain, S. (2021). Improving plant disease classification by data augmentation using generative adversarial networks. Materials Today: Proceedings, 47, 4560–4565. https://doi.org/10.1016/j.matpr.2021.05.226

Diao, W., Zhang, L., Chen, X., Yu, Q., Diao, W., Zhang, L., Chen, X., & Yu, Q. (2021). Enhancing low-resolution images for plant disease detection using deep learning techniques. IEEE Access, 9, 117907–117917. https://doi.org/10.1109/ACCESS.2021.3105193

Islam, M. M., Murshed, M., Hou, J., Islam, M. M., Murshed, M., & Hou, J. (2021). Exploring deep learning architectures for plant disease detection: A comprehensive review. Multimedia Tools and Applications, 80(22), 33605–33642. https://doi.org/10.1007/s11042-021-10841-3

Liu, J., Chen, J., Huang, L., Liu, J., Chen, J., & Huang, L. (2022). Transfer learning with MobileNetV2 for real-time plant disease classification. Sensors, 22(1), 293. https://doi.org/10.3390/s22010293

Pradhan, A., Mishra, S., Sahu, M., Pradhan, A., Mishra, S., & Sahu, M. (2023). High boost filtering for medical and agricultural image sharpening: An empirical evaluation. Journal of Digital Imaging, 36(1), 120–130. https://doi.org/10.1007/s10278-022-00644-0

Zhou, W., Li, Q., Sun, Y., Zhou, W., Li, Q., & Sun, Y. (2021). Edge enhancement methods based on sharpening kernels for low-resolution images. Journal of Imaging, 7(9), 153. https://doi.org/10.3390/jimaging7090153

Patil, R., Narote, S. P., Patil, R., & Narote, S. P. (2020). Enhancement of plant leaf disease images using unsharp masking and its effect on classification performance. International Journal of Computer Applications, 176(30), 7–12.