Industrial Object Recognition Research Articles

Industrial object and defect recognition utilizing multilevel feature extraction from industrial scenes with Deep Learning approach

Modern industry requires modern solutions for monitoring the automatic production of goods. Smart monitoring of the functionality of the mechanical parts of technology systems or machines is mandatory for a fully automatic production process. Although Deep Learning has been advancing, allowing for real-time object detection and other tasks, little has been investigated about the effectiveness of specially designed Convolutional Neural Networks for defect detection and industrial object recognition. In the particular study, we employed six publically available industrial-related datasets containing defect materials and industrial tools or engine parts, aiming to develop a specialized model for pattern recognition. Motivated by the recent success of the Virtual Geometry Group (VGG) network, we propose a modified version of it, called Multipath VGG19, which allows for more local and global feature extraction, while the extra features are fused via concatenation. The experiments verified the effectiveness of MVGG19 over the traditional VGG19. Specifically, top classification performance was achieved in five of the six image datasets, while the average classification improvement was 6.95%.

3D Convolutional Neural Networks Initialized from Pretrained 2D Convolutional Neural Networks for Classification of Industrial Parts.

Deep learning methods have been successfully applied to image processing, mainly using 2D vision sensors. Recently, the rise of depth cameras and other similar 3D sensors has opened the field for new perception techniques. Nevertheless, 3D convolutional neural networks perform slightly worse than other 3D deep learning methods, and even worse than their 2D version. In this paper, we propose to improve 3D deep learning results by transferring the pretrained weights learned in 2D networks to their corresponding 3D version. Using an industrial object recognition context, we have analyzed different combinations of 3D convolutional networks (VGG16, ResNet, Inception ResNet, and EfficientNet), comparing the recognition accuracy. The highest accuracy is obtained with EfficientNetB0 using extrusion with an accuracy of 0.9217, which gives comparable results to state-of-the art methods. We also observed that the transfer approach enabled to improve the accuracy of the Inception ResNet 3D version up to 18% with respect to the score of the 3D approach alone.

CdF2:Ga crystal application for dynamic joint transform correlator

Dynamical holographic joint transform correlator that uses cadmium fluoride crystal with gallium bistable impurity centers to generate real-time matched filters is proposed. Optical scheme features and the correlator performance are presented. Output correlation signals of object discrimination both experimentally obtained and theoretically calculated are demonstrated. Industrial object recognition illustrates possible practical application of the correlator proposed.

Affine invariant model-based object recognition

New techniques are described for model-based recognition of the objects in 3-D space. The recognition is performed from single gray-scale images taken from unknown viewpoints. The objects in the scene may be overlapping and partially occluded. An efficient matching algorithm, which assumes affine approximation to the prospective viewing transformation, is proposed. The algorithm has an offline model preprocessing (shape representation) phase which is independent of the scene information and a recognition phase based on efficient indexing. It has a straightforward parallel implementation. The algorithm was successfully tested in recognition of industrial objects appearing in composite occluded scenes. >

Neural Network Based Object Recognition System ForManufacturing Applications

The manufacturing environment these days is characterised by rapid change, posing new challenges and problems to the production and operations manager in the industry. Under this circumstances, process flexibility is fast becoming a major priority for many organisations as they attempt to deal with these changes. In response to the need for process flexibility, increasing attention is being given to integration of computing technologies with the manufacturing systems leading to the development of flexible manufacturing systems aided with high performance vision capabilities. The flexible manufacturing between various batches. Ideally, FMS should be able to accommodate the batch size of one component. This is possible if the programs for machining of different batches can be loaded instantly without operator intervention. An automatic component (object) recognition system, capable of identifying the component loaded on any given machine is required for meeting this objective. Existence of such a system provides an additional latitude in the scheduling of FMS to reduce the down time and improve the utilisation. This requires development of high performance industrial vision systems, which is quite a challenging task, as vision is a complex and computational intensive task, and the typical manufacturing environment is characterised by several different parts with simple or complex geometric descriptions, partially occluded parts, or parts with surface anomalies, blemishes and varying shapes. In this paper the recognition performance of the neural network based classifiers is reported for an industrial object recognition system being developed at the author’s laboratory. The intelligent object recognition system being developed has a modular structure with each module dedicated to different levels of vision in the vision hierarchy. A comparison of two different cases where the proposed neural network based classifiers perform much better than the general statistical classifiers is reported. The first case considered is that of the one degree of freedom tools where the backpropagation neural net has the better recognition performance and the second case considered is that of partially occluded objects where the Hopfield neural net yields better performance.