Object Detection Experiment in CBIR Works using Gradation Contour Line

Image retrieval continues to be developed using an alternative approach that is not only based on the information contained in the image. Retrieval techniques to achieve good quality also provide low computational value even though they have to process large-dimensional image data. Usually, CBIR searches for images using a query that begins with the feature extraction process (image content). Image feature parameters that can be used to shoot images can be histograms, colors, textures, and shapes, certain types of objects, certain types of events, individual names, locations, and emotions. However, indexing of images with the contour line feature is still quite attractive, CBIR with more distinctive features on an object is still rarely used. The results of this study using contour lines as a feature in object detection work, the results obtained to match the image with an accuracy of 83% - 84%. The test data used 550 object images taken randomly from the Caltech 101 dataset with several contour error experiments. The highest value of image matching results in the object detection task with contour line characteristics was 84.381%. This is obtained with the condition error size = 10, Contour Error = 10, and error X = 10. This means that the best contour line feature is obtained with a contour error of 10.