A DCNN-based arbitrarily-oriented object detector with application to quality control and inspection
Kai Yao, Alberto Ortiz, Francisco Bonnin‐Pascual
Abstract
Following the successful contribution of machine vision systems to automated inspection and quality control, in this paper, we propose a new bounding boxes-based regression solution aiming at recognizing generic targets for which detecting their orientation may be beneficial. Our solution consists in a two-stage arbitrarily-oriented object detection method making use of indirect regression of oriented bounding boxes parameters. Besides, in order to be able to recognize targets of different sizes and shapes, the solution adopts a multi-scale approach. The resulting performance is evaluated against datasets from two different industry-related case studies: while one involves the detection of a number of object classes in the context of a quality control application, the other stems from the visual inspection domain and deals with the localization of image areas corresponding to scene points affected by a specific sort of defect. The detection results that are reported for both tasks show that competitive performance levels can be achieved in both cases despite the differences among them and their specific challenges.