Akademik Veri Yönetim Sistemi
17th INTERNATIONAL CONFERENCE ON ENGINEERING & NATURAL SCIENCES, Praha, Çek Cumhuriyeti, 3 - 07 Mayıs 2025, ss.216-220, (Tam Metin Bildiri)
Detecting strawberry flowers plays a crucial role in yield estimation and resource allocation in agriculture. Traditional manual methods are being replaced by computer vision and object detection techniques, which enhance accuracy and significantly reduce processing time. In this study, a Haar Cascade-based model was developed to automate strawberry flower detection. The model was trained using cropped positive and negative samples extracted from 600×600 pixels original images at various resolutions (24×24, 50×50, 100×100, 224×224, and 256×256 pixels). To optimize performance, several image processing techniques were applied using the OpenCV library, including resizing, grayscale conversion, dilation, Gaussian blurring, and histogram equalization. Experimental results demonstrated that the highest F1 score was achieved with 224×224 pixels input images, indicating the most balanced performance between Precision and Recall. While the 256×256 resolution resulted in the highest Precision (%97.68), it also led to a decrease in Recall (%84.77), suggesting a trade-off between false positives and false negatives. These findings highlight the significant impact of input image resolution on object detection accuracy. This study demonstrates the feasibility of computer vision-based object detection techniques in agricultural applications and underscores the importance of optimization strategies in enhancing model performance. Keywords: Haar Cascade classifier, Object detection, Strawberry flower