Detection of pest using Odor substance based on Deep Learning Algorithms

In agriculture, early-stage control of pests is animmense challenge to the farmers. It affects growth and the economy respectively. This study brings out a new way to analyze the pest based on odor substance. Each pest is different from one another and produces different odors. Odor can make the work easier when compared with other detection methodologies because it helps to identify the pest while it is hidden like a deeply placed pest, swirled pest, etc., Hence, the proposed system has taken the odor as a key point and gait towards on it. Here, we’ve implemented five sensors to analyze five different smells such as pungent, misty, sweet smell, and musty, etc., and concatenate the process with Faster R-CNN to proceed with the feature extraction. The pseudocodecan be used for further development to bring out effective processes with perfect time and accuracy. The accuracy of pest detection is relatively increased to 3% over CNN based process. Some samples have been given to check the performance metrics of the proposed progression.