Detection of oil rape seed losses before harvest by Image analysis within Fog computing

Abstract

Losses in the production of oilseed rape can occur before harvest, caused by the spontaneous opening of mature pods and the fall of seeds on the ground. Different seed pods, among other things, ripen on the same plant at different times. So, in the cultivation of similar crops, one of the most critical moments is determining the right time for harvesting, because late harvest implies overripeness and opening of the shell, which leads to seed spoilage, losses and mechanical damage during threshing. One way of monitoring seed shedding and thus potential losses is by placing a container between rows of oilseed rape plants and monitor the number of seeds that fall from open shells. The presented model of the system, which consists of sensor devices with associated cameras, positioned above the position, has a function to transmit images of the current state. Central to this paper is an image analysis application that can be performed near sites on computer-aided devices within Fog Computing. In this way, the results of the analysis of images on the number of seeds in the container are obtained almost immediately and can be forwarded to the Cloud platform or directly to the user who will take appropriate action. By obtaining timely information on the number of scattered seeds, it is possible to organize the harvest in an optimal way in order to avoid losses and prevent over-ripeness of oilseed rape.