Automating the assessment of sensor quality in the production of thin-film nitrate sensors can yield significant advantages. Currently, the inspection process is extremely time and labor intensive, requiring technicians to manually examine sensors from each batch to determine their performance. Not only is manually examining sensors costly, it also takes days to conclude the results. It is possible to utilize image based learning approach to entirely automate the quality assessment process by accurately predicting the performance of every sensor; this allows for instant performance analysis and rapid changes to the fabrication parameters.The fabrication parameters will directly control the thickness of the ion-selective membrane (ISM) of the nitrate sensor. The thickness of the ISM directly affects the texture on the sensor’s surface. Because of the reliable correlation between sensor performance and sensor surface texture, it allows us to use learning methods to predict sensor performance through images instead of direct measurements.We propose a method to predict sensor quality using noncontact sensor images through a series of image processing techniques followed by machine and deep learning.
To enhance images, one often has to apply a filtering operation (denoising). However, there are several issues within the denoising. One of them is that sometimes denoising can be not efficient. Another issue regards a selection of an appropriate filter and setting of its parameters. As a particular case, we consider a 2D DCTbased filter with 8x8 pixel fully overlapping blocks where one of the parameters is a proportionality factor (PF) used in the threshold setup. We show that a performance of the considered filter in the sense of standard PSNR and visual quality metric PSNR-HVS-M can be predicted before applying image filtering procedure. This prediction is sufficiently faster than the denoising itself and accurate enough. We demonstrate that, having DCT statistics in a limited number of image blocks, such a prediction can be done for several values of PF. This allows deciding is it worth applying filtering to an image at hand. If the denoising is desired, it is also possible to select the PF optimal value for the considered image and noise intensity. Such a procedure, in some cases, can result in improvement of output PSNR or PSNR-HVS-M by up to 1 dB in comparison to the default parameters setup.
Synthetic aperture radar (SAR) images are corrupted by a specific noise-like phenomenon called speckle that prevents efficient processing of remote sensing data. There are many denoising methods already proposed including well known (local statistic) Lee filter. Its performance in terms of different criteria depends on several factors including image complexity where it sometimes occurs useless to process complex structure images (containing texture regions). We show that performance of the Lee filter can be predicted before starting image filtering and which can be done faster than the filtering itself. For this purpose, we propose to apply a trained neural network that employs analysis of image statistics and spectral features in a limited number of scanning windows. We show that many metrics including visual quality metrics can be predicted for SAR images acquired by Sentinel-1 sensor recently put into operation.