Angel Dacal-Nieto

A genetic algorithm approach for feature selection in potatoes classification by computer vision

Potato quality control has improved in the last years thanks to automation techniques like machine vision, mainly making the classification task between different quality degrees faster, safer and less subjective. We present a system that classifies potatoes depending on their external defects and diseases. Firstly, some image processing techniques are used to segment and analyze the potatoes. Then, a classifier is used to decide the group the potato belongs to. For the feature selection task, we have designed an ad-hoc genetic algorithm which maximizes the classification percentage. This approach is used to perform an optimization in the search of the better feature combination. The system shows to be effective in real operation simulations (working with unwashed potatoes covered with dust and sand,), what seems to be a good starting point in the development of the system.

Common scab detection on potatoes using an infrared hyperspectral imaging system

The common scab is a skin disease of the potato tubers that decreases the quality of the product and influences significantly the price. We present an objective and non-destructive method to detect the common scab on potato tubers using an experimental hyperspectral imaging system. A supervised pattern recognition experiment has been performed in order to select the best subset of bands and classification algorithm for the problem. Support Vector Machines (SVM) and Random Forest classifiers have been used. We map the amount of common scab in a potato tuber by classifying each pixel in its hyperspectral cube. The result is the percentage of the surface affected by common scab. Our system achieves a 97.1% of accuracy with the SVM classifier.

Non-destructive detection of hollow heart in potatoes using hyperspectral imaging

We present a new method to detect the presence of the hollow heart, an internal disorder of the potato tubers, using hyperspectral imaging technology in the infrared region. A set of 468 hyperspectral cubes of images has been acquired from Agria variety potatoes, that have been cut later to check the presence of a hollow heart. We developed several experiments to recognize hollow heart potatoes using different Artificial Intelligence and Image Processing techniques. The results show that Support Vector Machines (SVM) achieve an accuracy of 89.1% of correct classification. This is an automatic and non-destructive approach, and it could be integrated into other machine vision developments.

A machine vision system for the calibration of digital thermometers

Automation is a key point in many industrial tasks such as calibration and metrology. In this context, machine vision has shown to be a useful tool for automation support, especially when there is no other option available. A system for the calibration of portable measurement devices has been developed. The system uses machine vision to obtain the numerical values shown by displays. A new approach based on human perception of digits, which works in parallel with other more classical classifiers, has been created. The results show the benefits of the system in terms of its usability and robustness, obtaining a success rate higher than 99% in display recognition. The system saves time and effort, and offers the possibility of scheduling calibration tasks without excessive attention by the laboratory technicians.

Entropy of gabor filtering for image quality assessment

A new algorithm for image quality assessment based on entropy of Gabor filtered images is proposed. A bank of Gabor filters is used to extract contours and directional textures. Then, the entropy of the images obtained after the Gabor filtering is calculated. Finally, a metric for the image quality is proposed. It is important to note that the quality of the image is image content-dependent, so our metric must be applied to variations of the same scene, like in image acquisition and image processing tasks. This process makes up an interesting tool to evaluate the quality of image acquisition systems or to adjust them to obtain the best possible images for further processing tasks. An image database has been created to test the algorithm with series of images degraded by four methods that simulate image acquisition usual problems. The presented results show that the proposed method accurately measures image quality, even with slight degradations.

Digital instrumentation calibration using computer vision

This paper describes a computer vision system designed to automatically read the displays of digital instrumentation. The system is used in calibration sessions where many measurements have to be made and where we are interested in getting the whole numerical series downloaded on a host computer. Before our system was running, a human operator had to inspect the instruments at the right times (required by the calibration procedure) and to write down all the results. Note that we are speaking of very simple and sometimes old instruments that usually do not provide a digital interface or a removable memory (and if they do, we do not have a standard interface accepted by all the manufacturers). Results show the benefits of this system, obtaining a success rate higher than 99% in display recognition

Automatic reading of digital instrumentation

This communication describes a computer vision system designed to automatically read the displays of digital instrumentation. The system is used in calibration sessions where many measurements have to be made and where we are interested in getting the whole series downloaded on a host computer. Before our system was running, a human operator had to inspect the instruments at the right times required by the calibration protocol and write down all the results. Note that we are speaking of very simple and sometimes old instruments that usually do not provide a digital interface or a removable memory.

Industrial Wearable Exoskeletons and Exosuits Assessment Process

Industrial wearable exoskeletons and exosuits represent a vibrant technology with revolutionary potentials to enhance the operating conditions, health and safety of the worker. It brings forward the important social and technological goal of helping the workers instead of replacing them. An effective assessment process is a core for the sustainability and deployment of these devices in the industry. We present a process based on the evaluation criteria to validate the Impact on Worker, Appropriation to the Task, Utility to the Task, Usability and Safety. We test this criterion with the help of objective and subjective methods, which depend upon assessment techniques, assessment devices, surveys and subjective scales. In the end, we share our experience of implementing this process, and we point out industrial needs which can help future research and development directions.

Occluded Dark Field: A New Inspection Technique of Convex Mirrored Surfaces

The inspection of shiny or chromed surfaces is usual in many industries, especially in the automotive auxiliary companies. However, the inspection over those objects is a great challenge, since the acquisition procedures require specific designs due to reflection properties of mirrored surfaces, especially if the piece is convex, or specular in many directions. The best known illumination techniques are not effective in these cases: either they overexpose the piece, missing important defects, or they require a progressive scanning using a set of images instead of only one. In this paper we present a new image acquisition technique that uses a light–absorber dome which encases a Dark Field ring illumination. The so named Occluded Dark Field allows the detection of defects on convex mirrored surfaces by acquiring one only image. This improvement saves acquisition and execution time on the inspection, allowing a fastest quality control over all the production.

Rapid infrared multi-spectral systems design using a hyperspectral benchmarking framework

We present a benchmarking framework to design multi-spectral systems working in the NIR range for multiple purposes. This framework is composed of a hyperspectral imaging hardware and an ad-hoc software that performs pattern recognition experiments (image acquisition, segmentation, feature extraction, feature selection, classification and evaluation steps) comparing different algorithms in every step. For each experiment, we obtain a solution using a generic hyperspectral system, but we also obtain enough data to design a specific multi-spectral system in order to decrease the overall execution time. This improvement is based in the feature selection step, that provides the most relevant wavelengths for the problem. The framework has been tested for detecting internal and external features in potatoes, determining the origin of honey, and studying fecundity parameters in hen eggs.