บทความวิจัยการประชุมวิชาการระดับชาติมหาวิทยาลัยนอร์ทกรุงเทพพ.ศ. 2568

การประยุกต์ใช้โครงข่ายประสาทเทียมแบบคอนโวลูชันสําหรับจําแนกประเภทโรคใบอ้อย

The Application of convolution neural network for sugarcane leaf disease classification

ผู้จัดทำ: ณัฐวุฒิ จุ้ยเบี้ยว และเตชค์รูสิณป์ เพียซ้าย
ระดับการศึกษา: -
สาขาวิชา/กลุ่มสาขา: ด้านสหวิทยาการ วิทยาศาสตร์และเทคโนโลยี
0 เข้าชม 0 ดาวน์โหลด
ดาวน์โหลดเอกสาร

บทคัดย่อ (ภาษาไทย)

โรคใบอ้อยเป็นโรคที่ส่งผลต่อการเพาะปลูกและผลผลิตการแปรรูปของอ้อย เพื่อป้องกันการแพร่ระบาด ของโรคใบอ้อยจึงจําเป็นต้องทําการเข้าพื้นที่เพื่อตรวจสอบคุณลักษณะและป้องกันการแพร่ระบาดของโรคให้ถูกต้อง โรคใบอ้อยที่พบทั่วไปและมีลักษณะอาการใกล้เคียงกัน ได้แก่ โรคใบด่าง โรคเส้นกลางใบแดง โรคราสนิม และโรค ใบเหลือง เพื่อจําแนกประเภทโรคใบอ้อยได้รวดเร็ว บทความนี้นําเสนอการประยุกต์ใช้โครงข่ายประสาทเทียมแบบ คอนโวลูชัน (CNN) สําหรับจําแนกประเภทโรคใบอ้อยจากภาพถ่ายดิจิทัล โดยภาพ 10,000 ภาพ ประกอบด้วยภาพ โรคใบอ้อยแต่ละโรคจํานวนละ 2,000 ภาพ และภาพอาการใบปกติ ภาพทั้งหมดได้ถูกแบ่งสัดส่วนเป็น 80:20 สําหรับ การฝึกสอนและการทดสอบ การประยุกต์นี้ใช้เครื่องมือของโปรแกรม Visual Studio Code โปรแกรมไลบรารีคําสั่ง TensorFlow และโปรแกรมภาษาไพธอนเวอร์ชัน 3.10.16 ผลการทดลองพบว่า การใช้ประยุกต์ CNN สําหรับจําแนก ประเภทโรคใบอ้อยให้ค่าความแม่นยําที่ดีประมาณ 97%

Abstract (English)

Sugarcane leaf disease is a disease that affects sugarcane cultivation and processing. To he leaf disease is a disease that affect prevent the spread of sugarcane leaf diseases, it is necessary to enter the area to examine the characteristics and to prevent the spread of the diseases correctly. Sugarcane leaf diseases that are common and have similar symptoms include Mosaic disease, Red Rot disease, Rust disease, and Yellow Leaf disease. To classify sugarcane leaf diseases quickly, this article presents the application of the Convolutional Neural Network (CNN) to classify sugarcane leaf diseases from digital images. The 10,000 images include 2,000 sugarcane leaf disease images of each disease and the normal leaf symptom images. These images are divided into 80:20 ratio for the training and testing. This application uses the tools of the Visual Studio Code, TensorFlow library, and Python version 3.10.16 programs. The experimental results show that the CNN application to classify sugarcane leaf diseases gives a good accuracy of approximately 97%.