Pest Identification using Image Processing Techniques in Detecting Image Pattern through Neural Network

In rice production, pest invasion is considered as the most challenging task for crop technicians and farmers. Pest invasion can cause serious losses and affect the income of farmers. It is then important to assess their density for pests forecasting decision making. Existing identification techniques of these species comprise of using different traps to detect their presence. However, these traditional methods are labor-intensive and sometimes experts on this field are not available. Another problem is that multiple site and frequent monitoring of rice pests is time consuming and tedious for a crop technician. This can lead to low accuracy and delays in obtaining accurate count of these species. In this study, an identification system was developed to automatically identify the insect pests in the paddy field. Sticky trap was used to capture the insect which continuously monitored by a wireless camera to record the video. Different image processing techniques was utilized to detect and extract the captured insect from the image, and Kohonen Self Organizing Maps neural network was used to identify the extracted insect pests. The results indicate that the proposed automated identification system is capable of accurately identifying the insect pests even though there are lots of lighting variations. The new automated identification system developed in this study provides a reliable identification and was found to be faster than a human expert in identifying the insect pests caught