The work consisted in the collection of the user requirements in order to have an accurate definition of the system specifications (Task 1). The next step was the development of the defect database for the definition of the neural network training set in parallel with the design of the vision hardware (Task 2). Based on the specifications defined in task 2, the Vision Equipment has been developed in parallel with the feature extraction software (Task 3). Finally a mechanical system was developed in order to integrate the different sub-systems in a basic prototype (Task 4) to have an evaluation of the system performances (Task 5).
The following results have been achieved:
-Collection of the training-set of fabric defects;
-Development of the Integrated Vision Camera;
-Development of the Feature Extraction software installed on the Integrated Vision Camera;
-Integration of the vision hardware and lighting system for the development of a Scanning Head prototype;
-Development of the mechanical hardware and interfaces to the Scanning Head Driver;
-Development of the control software for the Scanning Head Driver;
-Development of the Defect Detection and Classification software.
The objective of the project is to develop an automatic system (FAST) for the realtime detection and sorting of faults during the weaving of a fabric to provide the textile SMEs with an efficient, cost effective tool for monitoring and improving the production. Currently there are no systems available on the market to automatically detect the faults directly on the loom. The inspection of fabrics is presently performed by human operators. This solution is expensive and not reliable. Attempts have been made in recent years to introduce automatic systems but the high costs, the frequent need for assistance, and the low versatility did not allow their application in the textile industry, in particular for SMEs. The FAST system is based on Near Sensor Image Processing (NSIP). This technology, patented by one of the RTD performers of this project, is a combination of optical sensors and digital processors on the same smart chip. This architecture offers the opportunity to perform image operations (including digitalisation, filtering, edge detection, thresholding, thinning, etc.) directly on the chip, allowing an effective time saving in comparison with conventional image data processing. The smart chip will be further developed and included in an Integrated Vision Camera with the Hierarchical Neural Network recently developed within the ONNI ADC Esprit project. The detection in real time will allow to stop the loom and repair about 70% of the faults with direct improvement of the quality as well as time and cost reduction in the mending phase, in particular as far as silk and jacquard fabrics are concerned. Furthermore the FAST system will be able to monitor the functioning of each loom in the weaving room thus allowing to immediately repair malfunctioning looms. The target price of the FAST system is 10% of the average cost of a loom. The effect on the core group of SMEs will be to significantly increase their markets in the supply of advanced looms or control equipment and high quality fabrics, depending on their position in the supply chain. This is confirmed by the large number of SMEs (10) which joined the consortium. The project will also have a positive social impact because it will reduce the need for distressing work of human operators in charge of the quality control, and create better job opportunities in the textile filiere. The proposed project complies with the scientific and technological objectives of Industrial & Material Technologies Programme, and in particular with Research Tasks 1.1.1.S (adaptation and application of new technologies), 1.1.4.S (application of automation), 1.1.2.S (implementation of computer assisted technologies), 2.3.2.S (development of sensors and equipment).
Funding SchemeCRS - Cooperative research contracts
NE11 0LF Gateshead
4760-034 Vila Nova De Famalicao
19002 Nissiza Karella Koropi-peania
22066 Mariano Comense
583 30 Linköping
4762 V.n. Famalicão