A Survey of Nitrogen Level Estimation in Plants using Image Processing
Abstract
Nutrients in plants are commonly associated with the Nitrogen Content present in the given plant at the given time. The evaluation of Nitrogen content is thus a pretty accurate measure of the health of the Plant in question. Traditional methods of Nitrogen Content estimation relied on either destructive and time-consuming methods which were extremely inefficient or on methods that required human eyesight to compare using a Colour Chart which had a really high probability of being incorrect as the colour on the chart itself may or may not be accurately printed. Thus, the use of Computer based Nitrogen Level estimation Techniques is sought.
This paper tells us about a few techniques used for Nitrogen Estimation and details some of the differences in the techniques whilst giving us a brief idea of the biggest drawbacks of each of the mentioned techniques.
Keywords
Download Options
Introduction
With the turn to digital technologies for almost every aspect of Electronics-based Surveying it was time to make Agricultural Analysis be digitally enabled. The theories for such a change were planted long before the actual technology itself could catch up. Thus, it is only in the last decade that we can see easier ways of creating programs that utilize those theoretical topics. All of the techniques are implemented using the software „MATLAB‟ by MathWorks. With a host of useful Add-Ons, built in applications and an exhaustive library of Functions, MATLAB is near perfect Software for a variety of Engineering and Research tasks.
Almost all of the mentioned techniques use a single reference plant which is documented over a long period of time or has a database prepared around various plants of the same species. Furthermore, all of these papers use slight variation of Artificial Intelligence and/or Machine Learning for better accuracy and efficiency over time. The Colour Recognition in all the papers surveyed below deals with the RGB and HSV (HIS) Colour Scale. We need both the Colour Scales for higher accuracy because just a single scale does not accurately describe the colour we are observing. The other important similarity between all of the papers is the need to sample down the image to a required size. This is performed for achieving a better balance between accuracy and efficiency. A down sampled image uses lesser temporary data whilst still providing us with relatively detailed matrix of the colours present in the actual/original Image.
Conclusion
Thus, we can conclude that the idea of Nitrogen Level Computation using only Image Processing, a Non-Traditional and Non-Destructive method is uniquely suitable for further research. A Common disadvantage in all of the Papers proposed here is the fact that all of them only work for a single type of plant which makes it difficult to use them as a general-purpose system. If a system combining the numerous positives of these systems are coupled to a relatively larger types of data, we can get a Swiss Army Knife tool which could also integrate Water Level Sensors, Soil Analyzer, etc.