Underwater images are degraded due to the complexity of the underwater environment and medium scattering and absorption. The image degeneration includes low contrast, color distortion, and blur, which makes the task of underwater vision difficult. A perceptually optimized cycle consistency generative adversarial network (CycleGAN-VGG) is proposed to restore distorted underwater images. We adopt the framework of CycleGAN so that our method does not require pairs of distorted underwater images and corresponding ground truth images for training. Inspired by perceptual loss, we use a combination of perceptual loss, cycle-consistent loss, and adversarial loss. The multiterm loss functions guarantee that the output has the same content and structure as the input while color looks like ground truth images. In an objective assessment, our method is significantly higher than other methods in terms of colorfulness metrics and contrast metrics. It confirms that our method effectively restores the color of underwater scenes and enhances some image details and shows a superior performance when compared to other state-of-the-art methods.
A novel method of droplet concentration measurement is presented based on analysis of angular distribution of elastic light scattering patterns, which provide a possible way to get the information of complex refractive index and inverse the liquid concentration for both absorbing and non-absorbing solutions. A mathematical model is established and an inverse algorithm corresponding to this model is also proposed which calculates the complex refractive index accurately. Experiments were carried out on both absorbing dye solutions and non-absorbing ethanol solutions with different concentrations. The results show a good agreement with the theoretical analysis and prove the potential use of this method.
Personal safety in public places has become the primary demand of modern people's social life, and the detection of dangerous liquids is a key technology in the field of security inspection. The spectral drop analysis method is used to study the identification of flammable liquids in this paper. The spectral droplet analysis system is constructed with the fiber-capacitance drop sensor and spectrometer, and through the combination of signal processing circuit and the acquisition software, the experimental platform is designed. Experiments with typical liquid samples are completed and the three-dimensional fingerprints of the samples are obtained. And the characteristic values of liquid samples are extracted after data processing, the methods of rapid identification of flammable liquids based on spectral and drop fingerprint information are researched. With the spectral information, the characteristic wavelength points are selected to extract the characteristic parameters of the sample using the principal component analysis method. And the discriminant prediction models are established by distance discrimination, Bayes discriminant and Fisher discriminant. With the drop fingerprint data, the characteristic parameters are extracted with waveform analysis method, and then the extreme learning machine algorithm is used to build classification and identification model. The experimental results show that it is feasible to identify flammable liquids by spectral droplet analysis method.
Ultraviolet-visible (UV-Vis) spectroscopy technology is used to measure chemical oxygen demand (COD) of water. The standard samples are prepared using potassium hydrogen phthalate. With different pretreatment methods and various modeling methods, the COD prediction models’ performance based on raw spectra are compared, and the sensitive wavelengths are selected on basis of the prediction results. In order to build prediction models with optimal performance, the water quality parameters’ effects on the detection of COD are also researched, and the experiments are carried out to find the relationship between COD and the sample’s temperature, turbidity. Then a combined method based on UV-Vis spectrum and water quality parameters is developed. The samples’ temperature and turbidity data are normalized with Min-Max Normalization method, and then different coefficients are assigned to the two parameters to form a new data, basing on the correlation coefficients of the models established by fusing the spectral information with temperature and turbidity respectively. A prediction COD model with the fusion data of water quality parameters and spectral information is established, using Partial least Squares(PLS) method. The experimental results show optimal performance (Mean ARE=2.46; RMSEP=1.92) for the prediction set. And this COD detection method set the foundation for further implementation of online analysis of water quality.
The droplet analysis technology and the detection principles of water quality parameters are combined to achieve quantitative detection of multi-parameter of water. The detection platform is designed based on fiber and capacitance droplet analysis technology, which is mainly composed of the droplet sensor, dissolved oxygen probe, liquid supply pump, photoelectric conversion elements, and the signal processing circuit. The detection of three quality parameters (refractive index, turbidity and dissolved oxygen) is carried out on this platform through experiments. For the turbidity of the water, the sample’s rainbow-peak value of the fingerprint obtained with the droplet sensor is proved to be highly correlated with turbidity. And the prediction model of turbidity is established by regression analysis method with Formazine standard solution, with he maximum relative error 3.9%. The measurement model of dissolved oxygen is researched by collecting the fluorescence signal excited by the dissolved oxygen probe and the sample’s temperature, and the performance of the BP neural network model and the regression model is compared. And it shows that BP neural network model performs better in the detection of dissolved oxygen. The measurement model of refractive index is determined through regression analysis, and the value of the rainbow-peak is selected as the key factor through the experiments with NaCl solution. The establishment of the three parameters’ detection model shows us a method to realize multi-parameter detection for environmental water quality.
Owing to the difference in physical and chemical properties, the liquid drops' growth states are dissimilar to different
liquids under same conditions. And this drop growth difference to various liquids is embodied in the corresponding
drop's contour feature obviously. Thus the liquid identification method based on CCD imaging system will be
introduced in detail in this paper. Through experiments to different liquids, the region area, boundary girth, drop length,
drop plumpness, drop circularity, and the profile edge of the liquid drop image will be extracted and analyzed. And with
these information the liquid identification can be realized. From sample experiments the region area and the drop
plumpness is more effective than other parameters in liquid discrimination. And the boundary girth and drop length
difference is very small to some liquids, thus they are the realitive weaker character to liquid drops.