http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
( Chi-hsuan Tseng ),( Ching-lu Hsieh ),( Yan-fu Kuo ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1
Fish body length is considered as an important index for resource management. Many organizations put restrictions on the size of caught fish that can be retained. In conservation ecology, fish body length is also used as an indicator for determining the sexual maturity. Conventionally, fish body length was measured manually using rulers or tape measures. Manual methods are, however, time consuming, labor intensive, imprecise and subjective. This study proposed a method to automatically measure fish body length from images with complex background. In the approach, a convolutional neural network classifier was first developed to detect fish head, fish caudal, and color plate in an image. Pixel to distance ratio was then calculated using the known length (25cm) of the color plate. Next, fish body length was estimated as the distance between the fish head and caudal. The approach reached an accuracy of 95.53% for fish body length estimation.
Quantifying and Clustering Texture Traits in Flowers of Genus Sinningia
( Tzu-ting Hung ),( Hao-chun Hsu ),( Yan-fu Kuo ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1
The flowers of genus Sinningia has a high degree of diversity in stripe and spot patterns. Delimiting these pattern as textural traits usually rely on horticulturalists’ judgment. However, the judgment by intuitive observation is subjective. This study aimed to quantify the stripe and spot pattern of genus Sinningia flowers automatically using machine vision and to cluster the textural traits using machine learning. The image of ventral petal was acquired using flatbed scanners. Two regions of interest (ROI), lobe region and tube region, were identified and were used for the feature quantification. The features of stripe and spot patterns were then quantified from the ROI using Gabor and Laplacian of Gaussian filters, respectively. The k-means clustering algorithm was next applied to the feature of patterns. The clusters were significantly associated with the textural traits.
Automatic Fish Species Identification using Convolutional Neural Networks
( Yi-chin Lu ),( Ching-lu Hsieh ),( Yan-fu Kuo ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1
Fish is a worldwide major food source. In recent years, overfishing has become a serious problem. Overfishing exhausts fish resources, endangers some fish species, and also threatens the entire marine food chain. Hence, organizations put regulations to prevent overfishing. Typically, the species of the fish caught are recorded and reported by ocean observers. However, the manual reporting method is laborious and time-consuming. This study proposed to recognize fish species from images automatically using deep convolutional neural networks (CNN). A first deep CNN was used to identify fish types (e.g., tuna, marlin, shark, and other). A second deep CNN was used to distinguish species of tuna fish, including Thunnus alalunga (Albacore), Thunnus obesus (Bigeye tuna), Thunnus albacares (Yellowfin tuna). A third deep CNN was used to determine the species of marlin fish, including Makaira nigricans (Atlantic blue marlin), Istiophorus platypterus (Indo-Pacific sailfish), Xiphias gladius (Swordfish). Each deep CNN was a fine-tuned VGG-16 model. The experimental results showed that the proposed method reached an average accuracy of 97.9%.
Classifying Endemic Fagaceae Species in Taiwan using Leaf Images
( Hao-chun Hsu ),( Cheng-hao Lee ),( Chih-kai Yang ),( Fang-hua Chu ),( Ming-jer Tsai ),( Yan-fu Kuo ) 한국농업기계학회 2018 한국농업기계학회 학술발표논문집 Vol.23 No.1
Fagaceae is one of the plant family which dominate the broad-leaved forests in Taiwan and have considerable value in economy and ecology. Traditionally, plant species identification based on leaf morphologies and is conducted using naked-eye observation. This study is proposed to distinguish the Fagaceae species using image processing and machine learning. In this study, leaf images of 10 Fagaceae species were collected. A serial of traits relevant to leaf morphologies, such as morphological, color, shape, and venation traits, were quantified from the leaf images. A support vector machine classifier was then developed to identify the species using the quantified traits. The proposed approach reached an identification accuracy of 95.8%.