IDENTIFICATION of POSITIVE GRAM BIOPESTICIDE BACTERIA USING FUZZY CLUSTERING LEVEL SET and RANDOM FOREST

The quality of agricultural land is an important factor for corn farmers in Madura. Control of maize plants is affected by the use of pesticides when viewed from a positive perspective can help humans in terms of eradicating pests that damage agricultural land. But on the other hand, pesticides also have a negative impact on humans and the surrounding environment, namely the breakdown of the food chain. To overcome this problem, biopesticides can be used in the form of bacteria that can kill plant pest organisms (OPT). Bacillus thuringiensis is one of the bacteria that can produce protein crystals that are insect killers (insecticides) when undergoing a sporulation process. By studying this identification, it is expected that farmers can analyZe the use of pesticides and replace them with bio-pesticides that are more environmentally friendly. In this study, an image processing approach was used to identify the presence of biopesticide bacteria. At the pre-processing stage, the stages of a culture of bacterial colonies were carried out on Blood Agar Plate media, followed by repairs to the results of image morphology. Fuzzy Clustering level set is one of the methods used in the image segmentation process. The results of form extraction are then used in the training process to determine the type of bacteria in the sample. This method makes identification of agricultural land easier, faster and costs less. The result is that in processing 100 training data and 25 data testing using Bacilli bacteria with 9 bacterial morphological attributes and 2 identification classes the accuracy value of the Random Forest decision tree was 91%.