Implementasi Sistem Deteksi Pelanggaran Parkir Kendaraan Menggunakan YOLOv8 dan Object Tracking di Kota Bogor

Abstract

Pelanggaran parkir pada ruas jalan perkotaan dapat mengurangi kapasitas jalan dan mengganggu kelancaran lalu lintas. Meskipun CCTV telah tersedia di berbagai titik Kota Bogor, pemanfaatannya masih terbatas pada pemantauan manual. Penelitian ini bertujuan mengimplementasikan sistem deteksi pelanggaran parkir berbasis YOLOv8 dan object tracking serta mengevaluasi kinerja model deteksi kendaraan. Penelitian menggunakan metode Cross-Industry Standard Process for Data Mining (CRISP-DM) dengan data rekaman CCTV Kota Bogor. Dataset terdiri atas tiga kelas kendaraan, yaitu mobil, sepeda motor, dan angkot. Model YOLOv8 dilatih menggunakan 4.603 citra dan diintegrasikan ke sistem berbasis web yang menerapkan object tracking dan analisis durasi berhenti kendaraan untuk mendeteksi pelanggaran parkir. Hasil evaluasi menunjukkan nilai precision 0,730, recall 0,751, dan mAP@0.5 0,793. Sistem mampu mendeteksi pelanggaran parkir secara otomatis serta menyajikan informasi pola pelanggaran. Hasil penelitian menunjukkan bahwa pendekatan yang digunakan dapat mendukung pemantauan dan analisis pelanggaran parkir berbasis CCTV

Parking violations on urban roads can reduce road capacity and disrupt traffic flow. Although CCTV is available at various locations throughout Bogor City, its use is still limited to manual monitoring. This study aims to implement a parking violation detection system based on YOLOv8 and object tracking and evaluate the performance of the vehicle detection model. The study employed the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology using CCTV footage from Bogor City. The dataset consists of three vehicle classes: cars, motorcycles, and public minivans (angkot). The YOLOv8 model was trained using 4,603 images and integrated into a web-based system that applies object tracking and vehicle stopping-duration analysis to detect parking violations. Evaluation results showed that the model achieved a precision of 0.730, a recall of 0.751, and an mAP@0.5 of 0.793. The system can detect parking violations in real-time and provide information on violation patterns. The findings indicate that the proposed approach supports CCTV-based parking violation monitoring and violation pattern analysis