Background Categorization for Automatic Animal Detection in Aerial Videos Using Neural Networks
This paper addresses the problem of animal detection in natural environment from aerial videos. Since the natural environment is usually composed of several fundamental elements such as trees, grass, streams, etc., it is proposed to distinguish the animal
- PDF / 8,690,735 Bytes
- 13 Pages / 439.37 x 666.142 pts Page_size
- 47 Downloads / 176 Views
2
1 Department of Electrical Engineering, French South African Institute of Technology, Tshwane University of Technology, Staatsartillerie Road, Pretoria 0001, South Africa [email protected] Department of Mechanical Engineering, Mechatronics and Industrial Design, Tshwane University of Technology, Staatsartillerie Road, Pretoria 0001, South Africa [email protected]
Abstract. This paper addresses the problem of animal detection in natural environment from aerial videos. Since the natural environment is usually composed of several fundamental elements such as trees, grass, streams, etc., it is proposed to distinguish the animal by categorizing the background into several classes. From the manually labeled samples, texture as well as brightness features are extracted to train a feedforward Neural Network. Then the classifier is applied to filter the test frame to locate potential animal regions. Four texture measures calculated from Grey Level Co-occurrence Matrix (GLCM) are used for texture feature description. Instead of obtaining these texture measures from grey level images, it is proposed to carry out calculation for every channel of the RGB image. The implemented results illustrate that this feature extraction method works well and the texture feature is a decisive factor in background categorizing. Keywords: Image segmentation Background categorization analysis Animal detection Neural network
Texture
1 Introduction 1.1
Animal Detection in a Natural Environment
Applications of Unmanned Aerial Vehicles (UAVs) are of great benefit to the field of nature conservation. Compared to conventional ways of wildlife surveys, it is far more economical to collect information from the region by flying a UAV (mounted with a proper camera) across it. It is especially efficient in some real-time tasks, such as monitoring and anti-poaching. The possibility of applying computer vision techniques to automatically analyse the aerial videos are increasingly being investigated [1, 2]. In general, tasks that wildlife conservation may concern include: (1) Counting the number of animals to monitor the distribution and abundance of animal species. © Springer International Publishing AG 2016 F. Schwenker et al. (Eds.): ANNPR 2016, LNAI 9896, pp. 220–232, 2016. DOI: 10.1007/978-3-319-46182-3_19
Background Categorization for Automatic Animal Detection
221
(2) Identification and investigation of a particular animal. (3) Tracking and monitoring of a herd and risk estimation. Animal detection in the video is to decide whether or not an animal of specific species is present in the scene and where it is located. Most object detection algorithms are based on machine learning mechanisms. Different views of an object were learned by a set of classifiers using positive and negative examples. By dividing the image into standard-sized sub-windows (patch) and passing them through trained filters, it can be then determined the existence and location of the object. For example, the neural networks [3], support vector machines [4], Bayesian net
Data Loading...
