In this paper, a set of improvements made in drill wear recognition algorithm obtained during previous work is presented. Images of the drilled holes made on melamine faced particleboard were used as its input values. During the presented experiments, three classes were recognized: green, yellow and red, which directly correspond to a tool that is in good shape, shape that needs to be confirmed by an operator, and which should be immediately replaced, since its further use in production process can result in losses due to low product quality. During the experiments, and as a direct result of a dialog with a manufacturer it was noted that while overall accuracy is important, it is far more crucial that the used algorithm can properly distinguish red and green classes and make no (or as little as possible) misclassifications between them. The proposed algorithm is based on an ensemble of possibly diverse models, which performed best under the above conditions. The model has relatively high overall accuracy, with close to none misclassifications between indicated classes. Final classification accuracy reached 80.49% for biggest used window, while making only 7 critical errors (misclassifications between red and green classes).

In this article, a Siamese network is applied to the drill wear classification problem. For furniture companies, one of the main problems that occurs during the production process is finding the exact moment when the drill should be replaced. When the drill is not sharp enough, it can result in a poor quality product and therefore generate some financial loss for the company. In various approaches to this problem, usually, three classes are considered: green for a drill that is sharp, red for the opposite, and yellow for a tool that is suspected of being worn out, requiring additional evaluation by a human expert. In the above problem, it is especially important that the green and the red classes not be mistaken, since such errors have the highest probability of generating financial loss for the manufacturer. Most of the solutions analysing this problem are too complex, requiring specialized equipment, high financial investment, or both, without guaranteeing that the obtained results will be satisfactory. In the approach presented in this paper, images of drilled holes are used as the training data for the Siamese network. The presented solution is much simpler in terms of the data collection methodology, does not require a large financial investment for the initial equipment, and can accurately qualify drill wear based on the chosen input. It also takes into consideration additional manufacturer requirements, like no green-red misclassifications, that are usually omitted in existing solutions.

This paper presents a novel approach to the assessment of decision confidence when multi-class recognition is concerned. When many classification problems are considered, while eliminating human interaction with the system might be one goal, it is not the only possible option—lessening the workload of human experts can also bring huge improvement to the production process. The presented approach focuses on providing a tool that will significantly decrease the amount of work that the human expert needs to conduct while evaluating different samples. Instead of hard classification, which assigns a single label to each class, the described solution focuses on evaluating each case in terms of decision confidence—checking how sure the classifier is in the case of the currently processed example, and deciding if the final classification should be performed, or if the sample should instead be manually evaluated by a human expert. The method can be easily adjusted to any number of classes. It can also focus either on the classification accuracy or coverage of the used dataset, depending on user preferences. Different confidence functions are evaluated in that aspect. The results obtained during experiments meet the initial criteria, providing an acceptable quality for the final solution.

This paper presents an improved method for recognizing the drill state on the basis of hole images drilled in a laminated chipboard, using convolutional neural network (CNN) and data augmentation techniques. Three classes were used to describe the drill state: red -- for drill that is worn out and should be replaced, yellow -- for state in which the system should send a warning to the operator, indicating that this element should be checked manually, and green -- denoting the drill that is still in good condition, which allows for further use in the production process. The presented method combines the advantages of transfer learning and data augmentation methods to improve the accuracy of the received evaluations. In contrast to the classical deep learning methods, transfer learning requires much smaller training data sets to achieve acceptable results. At the same time, data augmentation customized for drill wear recognition makes it possible to expand the original dataset and to improve the overall accuracy. The experiments performed have confirmed the suitability of the presented approach to accurate class recognition in the given problem, even while using a small original dataset.

This paper presents a time-efficient approach to the drill wear classification problem that achieves a similar accuracy rate compared to more complex and time-consuming solutions. A total of three classes representing drill state are recognized: red for poor state, yellow for elements requiring additional evaluation, and green for good state. Images of holes drilled in melamine faced chipboard were used as input data, focusing on evaluating differences in image color values to determine the overall drill state. It is especially important that there are as few mistakes as possible between the red and green class, as these generate the highest loss for the manufacturer. In green samples presented in gray-scale, most pixels were either black (representing the hole) or white (representing the chipboard), with very few values in between. The current method was based on the assumption that the number of pixels with intermediate values, instead of extreme ones, would be significantly higher for the red class. The presented initial approach was easy to implement, generated results quickly, and achieved a similar accuracy compared to more complex solutions based on convolutional neural networks.