Enhanced pose adjustment system for wing-box assembly in large aircraft manufacturing

Strict quality requirements in aircraft manufacturing demand high accuracy concerning pose alignments of aircraft structures. However, even though a pose adjustment system does pass the accuracy verification, the pose of the large complex structure has difficulty in smoothly and efficiently converging on the desired pose in large aircraft assembly. To solve this problem, we developed a pose adjustment system enhanced by integrating physical simulation for the wing-box assembly of a large aircraft. First, the development of the pose adjustment system, which is the base of the digital pose alignment of a large aircraft’s outer wing panel, is demonstrated. Then, the pose alignment principles of duplex and multiple assembly objects based on the best-fit strategy are successively explored. After that, the contributor analysis is conducted for nonideal pose alignment, in which the influences of thermal and gravity deformations on the pose alignment are discussed. Finally, a physical simulation-assisted pose alignment method considering multisource errors, which uses the Finite Element Analysis (FEA) to integrate temperature fluctuation and gravity field effects, is developed. Compared with a conventional digital pose adjustment system driven by the classical best-fit, the deviations of the Key Characteristic Points (KCPs) significantly decreased despite the impacts of thermal and gravity deformations. The developed pose alignment system has been applied to large aircraft wing-box assembly in Aviation Industry Corporation of China, Ltd. It provides an improved understanding of the pose alignment of large-scale complex structures.

Convolutional Neural Networks for Real-Time Eye Tracking in Interactive Applications

Face alignment and eye tracking for interactive applications should be performed with very low latency or users will notice the delay. In this chapter, a face alignment method for real-time applications is introduced featuring a convolutional neural network architecture for face and pose alignment. The performance of the novel method is compared to a face alignment algorithm included in the freely available OpenFace toolkit, which also focuses on real-time applications. The approach exceeds OpenFace's performance on both our own and the 300W test sets in terms of accuracy and robustness but requires significant parallel processing power, currently provided by the GPU. For the eye tracking application, stereo cameras are used as input to determine the position of a user's eyes in three-dimensional space. It does not require synchronized recordings, which may contain redundant information, and instead prefers staggered recordings, which maximize the number of possible model updates.

Face Pose Alignment with Event Cameras

Event camera (EC) emerges as a bio-inspired sensor which can be an alternative or complementary vision modality with the benefits of energy efficiency, high dynamic range, and high temporal resolution coupled with activity dependent sparse sensing. In this study we investigate with ECs the problem of face pose alignment, which is an essential pre-processing stage for facial processing pipelines. EC-based alignment can unlock all these benefits in facial applications, especially where motion and dynamics carry the most relevant information due to the temporal change event sensing. We specifically aim at efficient processing by developing a coarse alignment method to handle large pose variations in facial applications. For this purpose, we have prepared by multiple human annotations a dataset of extreme head rotations with varying motion intensity. We propose a motion detection based alignment approach in order to generate activity dependent pose-events that prevents unnecessary computations in the absence of pose change. The alignment is realized by cascaded regression of extremely randomized trees. Since EC sensors perform temporal differentiation, we characterize the performance of the alignment in terms of different levels of head movement speeds and face localization uncertainty ranges as well as face resolution and predictor complexity. Our method obtained 2.7% alignment failure on average, whereas annotator disagreement was 1%. The promising coarse alignment performance on EC sensor data together with a comprehensive analysis demonstrate the potential of ECs in facial applications.

Correção de Pose Facial a Partir de Modelos 3D

Facial recognition systems have to deal with a variety of problemsfor better accuracy results, such as lighting, obstruction, and posevariation, which occur when comparing an image to be detectedwith a previously identified image. In this context, this work aimsto use a pose alignment technique developed by Gang Pan. Togetherwith the Iterative Closest Ooint (ICP) and Average FaceModel (AFM) techniques, in order to perform a pose correction,a beginning of 3D facial models, in fully faces (90◦) or separatelyrotated, and test the result of this facial alignment with the PrincipalComponent Analysis (PCA), Linear Discriminant Analysis(LDA), and Support Vector Machine (SVM) recognition and classificationalgorithms related to the Local Binary Pattern (LBP), DiscreteCosine Transform (DCT), and Gaussian Filter preprocessing techniques.The classification algorithms will be tested in parallel andindependently counted, where one result will not interfere in anyother case, with the use of identifying which algorithm has the bestaccuracy. To perform the tests, a facial, text, infrared and visiblelight database was created with frontal images on the left, right,top, bottom and random face pose, resulting in a population of 90subjects and approximately 1600 colors.