ANALISIS PENGARUH THRESHOLD PADA METODE CANNY DAN SOBEL DALAM DETEKSI TEPI CITRA CABAI

Sri Rahayu; Ridwansyah; Jajang Jaya Purnama

doi:10.33365/teknoinfo.v19i2.162

Sri Rahayu Universitas Bina Sarana Informatika
Ridwansyah Universitas Bina Sarana Informatika
Jajang Jaya Purnama Universitas Bina Sarana Informatika

DOI: https://doi.org/10.33365/teknoinfo.v19i2.162

Abstrak

Abstract

The utilization of digital image processing technology in agriculture has developed rapidly, particularly for identifying and classifying horticultural commodities such as chili peppers. Chili peppers possess diverse visual characteristics, including color changes based on ripeness levels and irregular shapes. Misidentification can lead to losses during harvest, making accurate automatic detection essential. One of the key approaches in image processing is edge detection, which functions to extract object contours from the image background. However, the results of this process are highly influenced by the threshold parameters used. Inappropriate thresholds can result in the loss of important details or the appearance of disruptive noise, thereby reducing detection accuracy. Therefore, this study was conducted to analyze the effect of threshold parameter variations on the performance of two commonly used edge detection methods: Canny and Sobel. The edge detection process was carried out by applying the Canny and Sobel methods using the Kaggle Notebook platform. Canny edge detection involves Gaussian blur to reduce noise, calculation of intensity gradients, and the use of two threshold values. Meanwhile, Sobel calculates gradients along the X and Y axes to highlight pixel intensity changes.

The data used in this study consists of 11 images of various types of chili peppers, captured using a smartphone camera, then resized and converted to grayscale to simplify color information. The results from both methods were analyzed visually and quantitatively using metrics such as the number of edge pixels, PSNR (Peak Signal-to-Noise Ratio), and SSIM (Structural Similarity Index). The results of the study show that the Canny method is capable of producing clearer edges with less noise compared to Sobel, especially in images with low lighting. However, the Sobel method excels in processing speed and implementation simplicity. These findings highlight the importance of selecting the appropriate threshold values to enhance edge detection accuracy and efficiency. Consequently, the results of this study can serve as a reference in the development of more accurate and reliable automated systems for chili pepper identification and classification based on digital image processing.

Keyword: Canny, Chili, Edge Detection Image Processing, Sobel.

Abstrak

Pemanfaatan teknologi pengolahan citra digital dalam bidang pertanian telah berkembang pesat, khususnya untuk mengidentifikasi dan mengklasifikasi komoditas hortikultura seperti cabai. Cabai memiliki ciri visual yang beragam, seperti warna yang berubah sesuai tingkat kematangan dan bentuk yang tidak seragam, jika salah mengidentifikasi maka akan menyebabkan kerugian pada saat panen sehingga deteksi otomatisnya membutuhkan metode yang tepat. Salah satu pendekatan yang penting dalam pengolahan citra adalah deteksi tepi (edge detection), yang berfungsi untuk mengekstraksi kontur objek dari latar belakang gambarnya. Namun, hasil dari proses deteksi ini sangat dipengaruhi oleh parameter threshold yang digunakan. Threshold yang tidak sesuai dapat menyebabkan hilangnya detail penting atau munculnya noise yang mengganggu, sehingga menurunkan akurasi deteksi. Oleh karena itu, penelitian ini dilakukan untuk menganalisis pengaruh variasi parameter threshold terhadap performa dua metode deteksi tepi yang umum digunakan, yaitu Canny dan Sobel. Proses deteksi tepi dilaksanakan dengan menerapkan metode Canny dan Sobel dalam platform Kaggle Notebook. Canny edge detection melibatkan proses Gaussian blur untuk mengurangi noise, perhitungan gradien intensitas, dan penggunaan dua nilai threshold. Sedangkan Sobel melakukan perhitungan gradien pada sumbu X dan Y untuk menyoroti perubahan intensitas piksel. Data yang digunakan dalam penelitian ini berasal dari 11 gambar cabai berbagai jenis yang diambil menggunakan kamera ponsel, kemudian melalui proses resizing dan konversi ke grayscale guna menyederhanakan informasi warna. Hasil dari kedua metode dianalisis baik secara visual maupun kuantitatif menggunakan metrik seperti jumlah piksel tepi, PSNR (Peak Signal-to-Noise Ratio), dan SSIM (Structural Similarity Index). Hasil penelitian menunjukkan bahwa metode Canny mampu menghasilkan tepi yang lebih jelas dan minim noise dibandingkan Sobel, terutama pada gambar dengan pencahayaan rendah. Namun, metode Sobel memiliki keunggulan dalam kecepatan pemrosesan dan kesederhanaan implementasi. Temuan ini menegaskan pentingnya pemilihan nilai threshold yang tepat dalam meningkatkan akurasi dan efisiensi deteksi tepi. Dengan demikian, hasil penelitian ini dapat dijadikan acuan dalam pengembangan sistem otomatisasi identifikasi dan klasifikasi cabai berbasis pengolahan citra digital yang lebih akurat dan andal.

Kata Kunci: pengolahan citra deteksi tepi, Canny, Sobel, Cabai.

Referensi

DAFTAR PUSTAKA

Sudirman, PENGENALAN DESKRIPSI VARIETAS CABAI MERAH BESAR. PT. Nasya Expanding Management, 2022.

L. Soesanto, Kompendium Penyakit-Penyakit Cabai. Andi, 2024.

N. C. M. K. T. G. P. P. J. C. W. Yimyam, “Sorting Red and Green Chilies by Digital Image Processing,” in International Conference on Cybernetics and Innovations (ICCI), 2023.

D. J. Vanness, J. Lomas, and H. Ahn, “A health opportunity cost threshold for cost-effectiveness analysis in the united states,” Ann. Intern. Med., vol. 174, no. 1, pp. 25–32, 2021.

G. Q. Ai, A. Dmitry, I. Trond, and C. Joseph, “Quantum error correction below the surface code threshold Google Quantum AI and Collaborators,” 2025.

A. B. K. M. R. I. R. A. T. A. M. M. Y. A. D. D. D. Andayani, “Detection of Katokkon Chili Maturity using Convolutional Neural Network with Transfer Learning Model DenseNet169,” in International Workshop on Artificial Intelligence and Image Processing (IWAIIP), 2024.

C. D. C. S. M. C. M. V. Caya, “Identification of Philippine Chili Peppers using Canny Edge Detection and Feed– Forward Artificial Neural Network,” in International Conference on Humanoid, Nanotechnology, Information Technology, Communication and Control, Environment, and Management (HNICEM), 2023.

D. L. C. Sibarani, “Red Chili Leaf Disease Identification via GLCM and SVM Analysis,” in International Conference of Computer Science and Information Technology (ICOSNIKOM), 2023.

P. P. V. S. R. K. V. S. Y. S. K. B. S. S. M. Madupalli, “Deep Learning Model YOLOv5 for Red Chilies Detection from Chilly Crop Images,” in International Conference for Convergence in Technology (I2CT), 2023.

Y. Pei and Yingchao Zhang, “A Study on the Integrated Development of Artificial Intelligence and Tourism from the Perspective of Smart Tourism,” Yingchao Zhang, vol. 1852, no. 3, 2021.

R. Kumari, A. Kumar, H. L. Mandoria, B. K. Pandey, and Subodh Prasad, “A Review on Comparative Study of Different Edge Detection Techniques,” Int. Res. J. Eng. Technol., vol. 7, no. 10, 2020.

O. T. Tan, Sean Huey ; Lam, Chee Kiang ; Kamarudin, Kamarulzaman ; Ismail, Abdul Halim ; Rahim, Norasmadi Abdul ; Azmi, Muhamad Safwan Muhamad ; Yahya, Wan Mohd Nooriman Wan ; Sneah, Goh Kheng ; Seng, Moey Lip ; Hai, Teoh Phaik ; Lye, “Vision-Based Edge Detection System for Fruit Recognition,” J. Phys. Conf. Ser., vol. 2107, no. 1, 2021.

T. S. A. M. K. C. V. E. D. P. K. V. M. A. S. P. E. S. M. J. M. Lorenzo, “Edge Detection Aided Geometrical Shape Analysis of Indian Gooseberry (Phyllanthus emblica) for Freshness Classification,” Food Anal. Methods, vol. 15, pp. 1490–1507, 2022.