Classification of Swallow Nest Quality Based on Cleanliness Level Using Computer Vision

Abstract

This research implements computer vision to classify the cleanliness level of swallow nests. The main contribution of this research is how to separate the holes in the swallow nest from being detected as dirt. Experiments were conducted using images of swallow nests with a Fujifilm XT30 camera in JPG format. A total of 1571 images were used, 80% of which were used as training data. By using the white balance feature, white objects in the image will be clearly visible, and other colors look natural, giving a range of dark brown to black colors in the HSV feature extraction. This allows classification to be carried out with 2 two algorithms, namely Random Forest with 97.5% and Support Vector Machine (SVM) with 84.1%.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint