Aiming at the low accuracy and high computational complexity of convolutional neural networks in recognizing and classifying photovoltaic infrared hot spot images
as well as the difficulty in identifying the hot spot features on photovoltaic infrared images
an algorithm based on persistent homology is proposed to extract topological features from grayscale photovoltaic hot spot images. Firstly
the photovoltaic infrared hot spot image is converted to grayscale. Then
persistent homology calculation is performed on the grayscale image to obtain a barcode
from which the topological features are extracted to compose a new image. Finally
a convolutional neural network is employed to recognize and classify the new image. The experimental results demonstrate that the grayscale photovoltaic infrared image is a single-channel image
resulting in lower computational complexity. The extracted topological features of the photovoltaic infrared hot spot image are easier to identify and classify
FROSINI P, LANDI C.Size theory as a topological tool for computer vision[J]. Pattern recognition and image analysis, 1999, 9(4): 596-603.
CARLSSON G.Topology and data[J]. Bulletin of the American mathematical society, 2009, 46(2): 255-308.
SINGH G, MEMOLI F, CARLSSON G.Topological methods for the analysis of high dimensional data sets and 3D object recognition[C]// Eurographics symposium on Point-Based graphics, Prague, Czech, 2007.
CARLSSON G, ISHKHANOV T, DE SILVA V, et al.On the local behavior of spaces of natural images[J]. International journal of computer vision, 2008, 76(1): 1-12.