|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Wavelet transform and curvelet transform for wafer classification and recognition in solar cell manufacturing
ABSTRACT
This research proposes machine vision schemes for classification and recognition of polycrystalline solar wafers in solar cell manufacturing. Solar power is getting popular as an alternative of electricity energy. Polycrystalline solar cells dominate the current market share because of lower material and manufacturing costs. In solar cell manufacturing, a silicon ingot is sliced into thin wafers and then the wafers are further processed into solar cells. Conventional automatic identification systems such as bar codes, magnetic strips, OCR and RFID need a contacted identity on the object surface, They are not possible to implement for solar wafer tracking and data collection due to the thin, fragile silicon surface of a solar wafer.
The surface of polycrystalline solar wafer shows crystal grains of random sizes and shapes and thus forms a unique multi-grain pattern of the surface. In this study, encoding methods based on wavelet transform, wavelet packets and curvelet transform in the spectral domain and histogram matching of gray-level and gradient angles in the spatial domain are proposed to identify solar wafers to their corresponding ingot lots. In the spectral domain methods, wavelet decomposition, wavelet packets and curvelet transform are used to extract the feature vector of a solar wafer. The similarity between two compared wafers is then caculated by the Euclidean distance of their corresponding feature vectors. Because spectral methods are computationally expensive, the histogram matching of gray-level and gradient angles in the original images is also proposed for fast classification.
The experiments show that wavelet packets and curvelet transform have the best discrimination power to encode and identify solar wafers. The wavelet packets and curvelet transform can reach a 100% recognition rate for ingot lot identification. The gradient angle histogram matching arrives at 97% in correct classification. In the recognition result, the curvelet transform can recognize the slolar wafer within its two neighbors in the cutting sequence of an ingot. The computation times for a solar wafer of size 2600×2600 pixels are 0.78 seconds for the gradient angle histogram matching, 2.23 seconds for wavelet packets and 17.63 seconds for curvelet transform.
Keywords: Machine vision; Automatic identification; Product tracking; Wavelet transform; Wavelet packets; Curvelet transform; Histogram intersection |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
《Demonstration》
【Statistics of classification and processing time for 204 wafers from 10 ingots.】 |
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
