Bird-related outages have become the third failure cause of overhead transmission lines. Different bird species may cause diverse line faults. In order to improve the pertinence of bird-related outage prevention, it is necessary for inspectors to identify different bird species. This paper proposes a bird recognition method based on image processing, feature extraction and machine learning. The Grabcut algorithm is used for foreground extraction of bird images. Based on fine-grained classification ideas, the bird head is selected as the discriminative part for feature extraction. The color, texture, and shape features of the bird head are respectively characterized by color moments, gray-level co-occurrence matrix (GLCM), and geometric descriptions. Altogether 25 features are extracted to characterize the bird species. Taking these features as input parameters and the corresponding bird species as outputs, an intelligent classification model is established by multi-class support vector machine (SVM). This model is used to classify five types of birds which threaten the safe operation of transmission lines. The kernel principal component analysis (KPCA) is applied for feature dimension reduction. After feature selection and model training, the recognition accuracy reaches 88%. This study provides a reference to construct a bird identification system for transmission line inspectors to recognize harmful bird species.
