2 research outputs found
Texture-based of Naja Kaouthia snake recognition using K-Nearest Neighbour (KNN) / Nurul Hafeeza Aswani Aziz
No full textSnake is one of the dangers animal and afraid by people. Conventionally, the method to recognize snake's species is done manually by collecting the data from the patients itself. However it is very hard to use these data as a reference as the information collected are uncertain due to incorrect impression of the snake's species. Thus this study proposed a prototype of recognition that specifically to recognise Naja Kaouthia species. There are three phases involved in this study which are data collection, processing (i.e extraction and recognition) and post-processing. A total of 20 images have been captured at Taman Rama dan Reptilia, Malacca and each images produced 10 data of extraction. In the processing phase, mean variance moving window was used for the extraction process. The part of snake that has been used for this study is the internasal. Therefore the region of interest method will only be focussed at this part of snake where the texture of the internasal is mean, standard deviation and magnitude. As for the recognition, The K-Nearest Neighbour had been used. The Naja Kaouthia using K-Nearest Neighbour algorithm is identified as a promising method in snake recognition which produced 100% accuracy rate for training data and 100% for testing data
Statistical Texture Mean-Windowing Feature of Snake Identification
Full text linkSnake identification has been explored in various domains such as the image processing domain. In Malaysia, many of the snake species are non-venomous but still dangerous to the human. Conventionally, snake identification is evaluated by collecting the information from the patient. However, it is very hard and difficult to recognize the venomous and non-venomous snake types. Also, doctors need to inject the anti-venom into the patient which produced the side effect. Therefore, this paper classified the venomous snake of Naja Kaouthia and other venomous snake species. All the image datasets have been captured at Malacca Butterfly & Reptile Sanctuary, Melaka. The statistical vectors are extracted by using the normalized mean-moving windows. The taxonomical statistical texture vectors of snake region features are classified using Tree, K-Nearest Neighbor, Support Vector Machine, and Naïve Bayes classifiers. Results showed that most of the classifiers produced an accuracy rate of 100%
