Crack monitoring in concrete structures is very important for the maintenance of healthy civil infrastructure. Traditional computer vision-based methods have challenges with various textures and patterns in the background that can interfere with the accurate classification and segmentation of cracks. To overcome this problem, This research modified the region-based method from the OTSU algorithm to segment the crack shapes accurately. The OTSU can calculate the maximum and minimum value of the pixel threshold and differentiate foreground and background objects using the intensity pixel in the image. The masking process with inverted OTSU thresholding in the feature extraction process is used to remove the crack background. First, the mean filter and illumination correction are used to clean up small pixels that look like cracks. Second, OTSU thresholding is used to divide the background and foreground of the image. Then, inverse the thresholding to change the background into the foreground. After that, for the background of the crack, the dilation of the pixel for 1 mm was used to add a region of segmentation for the crack. Afterward, mask the background with the scene to remove the cracked background. Third, train the feature cracks that have been merged using SVM with OTSU for the segmentation of the cracks. Lastly, the width and length of the crack are calculated using Euclidean distance. By combining these methods, the accuracy of the classifier increases from 79% to 87%.