An Image Sensing Method To Capture Soybean Growth State For Smart Agriculture Using Single Shot Multibox Detector

Recently, smart agriculture, a new approach to farming using ICT, has been received great attention. To control cultivate condition precisely, it is important to capture the growth state of plants as well as environmental factors such as temperature, moisture, solar radiation, etc. In this paper, we propose an image sensing method to detect soy flowers and seedpods as growth factors using a state-of-the-art deep learning architecture called Single Shot MultiBox Detector (SSD). Images of soybeans were taken at Hokkaido Agricultural Research Center from Year 2015 to 2017 and we carry out the performance test for our system using a dataset of soybean images. The detection accuracy for seedpods and flowers are 0.586 and 0.646 in F-measure, respectively.