<![CDATA[Watermelon in Latin from Citrullus Lanatus Tunb is one of the vines that is developing very rapidly in Indonesia, besides that Indonesia is also the largest producer of watermelons in the world. It was found that the interest of many people is fond of watermelon. It can be seen that watermelon yield data in 2014 was 653,974. The type of watermelon that is often in demand is watermelon which has no seeds, which has a soft texture, does not contain too much water, and has a paler flesh than the seeded watermelon. This research carried out feature extraction on the texture of watermelon rind using the Gray Level Co-Occurrence Matrix (GLCM) method, then using the Support Vector Machine (SVM) to classify the sweetness or unsweetness of watermelon fruit without seeds with a distance of 10-11 cm. This research requires a camera on the system to be able to take images for detection and class classification can be done. If the system can detect the sweetness of watermelon without seeds, it will be displayed on 16x2 LCD. The GLCM features used in this study are correlation, contrast, homogeneity, energy, and dissimilarity. Tests conducted using the SVM kernel, namely linear, RBF, and polynomial kernels. To get the best d and θ values ​​in this study, several trials were carried out with values ​​of d=1,2 and θ=0â°, 45â°, 90â°, 135â°. The best d and θ values ​​after testing the detection of the sweetness of watermelon without seeds were d=2 and θ=45â° with an accuracy of 95%. Hardware integration testing is carried out from various sides of the image taking, namely from the front, back, and top. Obtained the best accuracy in testing the integration of hardware using the best d and θ values, which is 80% on the back side using a polynomial kernel. In testing the computation time of the system with a value of d=2 and θ=45â°, the best kernel was obtained, namely the polynomial kernel to detect the sweetness of watermelon without seeds, which took 13.13 seconds.]]>
Copyrights © 2021
