Hasil Pencarian  ::  Simpan CSV :: Kembali

Abstrak :
In this research it is applied a pattern recognition system by using an artificial neural networks to recognize several samples on weaving products, such as plain weave, twill weave, and sateen weave. In order to extract textural characteristics or features from sample images, it is used the Neighboring Grey Level Dependence Matrix (NGLDM)-method as proposed by Sun, which is invariant under rotation and linear grey level transformation. Five textural features i.e. Small Number Emphasis, Large Number Emphasis, Number Non uniformity, Second Moment, and Entropy will be used as the representative features of sample images. Those features are used as input to the neural networks, which have learned by the back propagation method. Baths methods (continuous and periodic) for changing the interconnection weights, and the performances of the two types of neuron transfer functions are also observed and investigated, in order to obtain an optimal network configurations. The results of experiment will be very useful for the next stage of research in designing an integrated vision system for the recognition of weaving product's quality in textile industry.

Abstrak :
In this research, it is designed a simple inspection model for defect detection on woven fabrics at the weaving stage of processing, based on texture analysis. Textural features that extracted by using the NGLDM (Neighboring Greylevel Dependence Matrices) from the several avail-able samples either for normal or defective weaving products, are intelligently recognized by a neural network computational system. The model is useful in textile industry, which provide woven qualities produced by weaving machines, therefore, from the defect's information one can separates those products which have different grades to be processed at the dye finishing stage, and may check previous yarn's treatment, mechanical failures, etc. The inspection system is equipped by a flatbed conveyor, a CCD camera, and a microcomputer IBM-PC/AT 386 with a Computer Eyes image grabber card. The testing results of defect detection on the available samples, indicate more than 80% of recognition level can be achieved. In the future, it is anticipated that the system may be developed, in order to reduce much more human intervention for the defect detection.