PENGEMBANGAN MODEL PENGUKURAN CELAH ANTARPANEL PINTU MOBIL MENGGUNAKAN JARINGAN SARAF TIRUAN DAN PENDETEKSIAN KONTUR

• المؤلف الرئيسي: Rizkyputra, Shergio
• التنسيق: Final Project
• اللغة: Indonesia
• الوصول للمادة أونلاين: https://digilib.itb.ac.id/gdl/view/74389
• المؤسسة: Institut Teknologi Bandung

PT X is an automotive manufacturing company with 5 production plants located in Indonesia, whose vehicles are sold domestically and exported internationally. There exists a final inspection station where each vehicle unit are examined for body-fitting defects within their production line. Currently, the final inspection is performed visually by 4 human operators. The addition of human operators is also expected to reduce the body-fitting inspection time from 96 seconds to 86 seconds. As the production rate increases, this condition is risky in terms of accuracy, inspection speed, and operator workload. Operators must maintain the quality of their inspection process. These factors motivate the development of a side-door gap size automated visual inspection method, specifically with Artificial Neural Network models. This study develops automation method on body-fitting inspection that utilizes two object detection models based on Convolutional Neural Network (CNN), which are Canny Edge Detection (CED) and You-Only-Look-Once (YOLOv7), to detect panels’ gap defects through image processing. The models are trained to detect the size of gaps in white, red, and black colored panels using artificial defects samples as training dataset and real defects as testing dataset. Based on our evaluation, CED is better at predicting gap size detection on white, red, and black panels with average precision-recall scores for each threshold value of 34.97%, 33.68%, and 26.69%, respectively. This score is obtained through photometric transformation of the samples in the form of an optimal combination of a brightness level of 40% and a contrast level of -40% for each sample.