Transferring the style from source domain to target domain for learning target models is a widely used strategy in domain adaptive segmentation. Although diffusion-based image translation has enabled flexible style transfer, it is often difficult to maintain the original structure of the image realistically during the reverse diffusion, which provides very little control over the generated image. To tackle this issue, we present the diffusion-based approach toward domain adaptive segmentation of general retinal image, which conditions diffusion models with carefully crafted input noise artifacts as explicit guidance at the inference step. Concretely, the input cross-domain image and the segmentation map of source domain are merged by summing the output of two encoders. Then, the encoder-decoder framework is adopted to iteratively refine the segmentation map by using a diffusion model. In order to enhance the general understanding of target domain distribution, we also establish the frequency-adaptive conditions for each sampling step. Moreover, this paper takes the channel-wise information and coarse semantic mask with noise of target image as guidance in the denoising process, which is different from existing approaches that input Gaussian noise and further establishes controllable conditions at the inference step. Extensive experiments on domain adaptation (DA)-based retinal image segmentation demonstrate the superiority of our approach over some state-of-the-art methods.

RGB-D semantic segmentation is attracting wide attention due to its better performance than conventional RGB methods. However, most of RGB-D semantic segmentation methods need to acquire the real depth information for segmenting RGB images effectively. Therefore, it is extremely challenging to take full advantage of RGB-D semantic segmentation methods for segmenting RGB images without the depth input. To address this challenge, a general depth removal distillation method is proposed to remove depth dependence from RGB-D semantic segmentation model by knowledge distillation, which can be employed to any CNN-based segmentation network structure. Specifically, a depth-aware convolution is adopted to construct the teacher network for getting sufficient knowledge from RGB-D images. Then according to the structure consistency between depth-aware convolution and general convolution, the teacher network is used to transfer the learned knowledge to the student network with general convolutions by sharing parameters. Next, the student network makes up for the lack of depth in manner of learning by RGB images. Meantime, a Variable Temperature Cross Entropy (VTCE) loss function is proposed to further increase the accuracy of the student model by soft target distillation. Extensive experiments on NYUv2 and SUN RGB-D datasets demonstrate the superiority of our proposed approach.

A human-computer interaction nursing system is developed for the severely paralyzed patient, wh o cannot move their limb and body and has language barriers, but has clear consciousness and can make head and face movements. If the system can obtain the patient's real intention, the system will enable the patient own self-care ability to some extent through the rehabilitation mechanical device. There are two ways to obtain the real intention of the patient: the active and passive ways. The first way, the patient with clear consciousness can send demands to the system through the head and face movements; the second way, the system can caluculate the patient's real needs according to the patient's emotional state. During the study, the key problems need to be solved include: identifying the patient's expression, head and face movements accurately. Aiming at these problems, the system recognizes the head movement by the offset direction of the face's center position; the face movement is recognized using the Uniform Local Binary Pattern texture feature map; about the facial expression recognition of the patient, the system trains the patient's expression recognition model through the convolutional neural networks, and then identifies the patient's facial expression by the model. The experiment results indicate that the system can detect the human's face, cut the human's eye and mouth areas, identify the human's facial expression, head and face movements accurately.

In this paper, we propose a novel palmprint recognition algorithm. Firstly, the palmprint images are represented by the anisotropic filter. The filters are built on Gaussian functions along one direction, and on second derivative of Gaussian functions in the orthogonal direction. Also, this choice is motivated by the optimal joint spatial and frequency localization of the Gaussian kernel. Therefore,they can better approximate the edge or line of palmprint images. A palmprint image is processed with a bank of anisotropic filters at different scales and rotations for robust palmprint features extraction. Once these features are extracted, subspace analysis is then applied to the feature vectors for dimension reduction as well as class separability. Experimental results on a public palmprint database show that the accuracy could be improved by the proposed novel representations, compared with Gabor.

The satellite TIR remote sensing has become a promising technique for monitoring tectonic activities and detecting earthquake precursors due to the advantages of large observation area and short observation period. In order to identify and to extract the TIR anomaly, several presented methods are tested and compared in this paper. The comparison results indicate that Robust AVHRR Technology (RAT) is a better method for detecting pre-earthquake thermal anomaly. Anyway RAT is not able to quantitatively analyze the TIR anomaly before shock. RAT method is hence improved for obtaining the value of area and average temperature of TIR anomaly. Based on the analysis of some tectonic earthquakes, an empirical quantitative model for tectonic activity analysis and earthquake magnitude predication is established.

The phenomenon of a satellite thermal infrared (TIR) anomaly before earthquakes has been reported since the late 1980s. The reported increase of surface temperatures reaches 2-4 degC, occasionally higher. Usually, the anomaly appears one month to several days before the earthquake. Several mechanisms of hypothesis have been put forward to interpret the reported temperature increase. In this paper, the satellite TIR anomaly features of four earthquakes in the Taiwan region are first analyzed. To study the mechanisms of infrared anomaly a group of physical simulation experiments are carried out. The mechanism of the satellite Infrared anomaly before an earthquake in the Taiwan region is discussed based on the experimental results. Furthermore, a preliminary model for tectonic activity analysis and for short-term earthquake prediction based on the analysis of the satellite infrared anomaly before an earthquake in the Taiwan region is presented.

This paper briefly introduces the general scientific facts from the reports on satellite thermal infrared (TIR) anomaly before earthquake and the presented mechanism and hypothesis for interpretation the TIR anomaly. The spatio-temporal features of infrared (IR) radiation from loaded rock, stick-sliding rock and simulated active fault motion based on experimental IR detection are also introduced. Especially, the experiments on simulated fault activity shows that the TIR anomaly is likely to develop along the primary fault and the both sides of wedge-shaped acute geo-block in condition of intersected fault system, and that the intersection location is usually the coming epicenter. The theoretical mechanism of TIR anomaly before tectonic earthquake is hence suggested to be stress-thermal effect accompanied and strengthened by hydro-geological effect, greenhouse effect, P-hole effect and so on. As a case, the spatio-temporal features of TIR before Zhangbei Ms6.2 1998 earthquake is theoretically analyzed based on the overlay of active fault system on NOAA-AVHHR satellite TIR images. It was discovered that TIR anomaly is controlled by a potential great deep active fault and that the epicenter is exactly the intersection point of the primary great fault and two secondary faults being tow sides of an acute wedge-shaped active geo-block. It is concluded that although the earthquake is a complex and uncertain process of crust motion, the predication of earthquake is not impossible. Referring to the developing GEOSS and generalized remote sensing (GRS), a preliminary procedure for comprehensive multiple parameters analysis on fault activities and earthquake early warning is presented, which is based on massive data fusion.

Face detection and location are widely used in practice. Although there are lots of related researches, the results are not satisfactory, esp. for the situation of side-face and multiscale in complex background. Among these techniques, the method based on skin-color is an important approach, so far, however, the effect of detection and location is not favorable due to the unsatisfactory segmentation. In order to improve the segmentation, we employ a new proposed classifier - support vector machine (SVM), based on the output of which we put forward a new scheme to realize the face detection and location. Experimental results indicate the algorithm can work well on wide range of color, light and size variations in still images.

A universal mathematical scheme for hidden information detection from various original n-dimensional signals is presented. The key points include the establishment of functional expression and the employment of cascade neural networks. The former indicates the detection of hidden information is dependent upon the shape of whole signals, not the specific values and the sampling number of the signals, the latter refers to the architecture of neural networks consisting of two neural networks, the first extracts main features of signals, and the second detects hidden information from the extracted features. Since wavelet functions play important role in signal analysis and feature extraction, thus wavelet neural networks constitute the first neural networks. The general scheme of feature extraction from original signals in L/sup 2/(R/sup M/) by wavelet neural networks is presented. The application in signal processing of chemical chromatographic spectra of solution shows satisfactory results.

In this paper, a military map online publishing system is described which make browsing military map and querying army symbol attribute come true. This system is composed of publishing server and Web page template. Publishing server realized the mapping of army symbol and the display of military map. Page template is used to realize the Web page layout and the interface of querying the attribute of army symbol

A Web mapping system is described based on OpenGIS style layer specification and pipeline pattern. The patterns used to design the system are classified into three levels that are realized in architecting, designing and implementing of the system each. The whole system is constructed on the pipeline pattern. The pattern is divided into four filters to finish the selecting of spatial data, the generating of map feature style, and the rendering of map layer and presentation of map image. Map style data saved in a custom XML document is decoupled from map feature data that associate feature style generator filter. The system runs on Microsoft COM+ distribute implementing pattern.

Crop model simulates growth and yield of crop in field scale, and requires a lot of initial conditions and parameters as input data such as soil, water, climate data etc. WebGIS is a powerful tool for obtaining, managing and analysing spatial data on the Web. With the support of WebGIS, crop models based on Web services can be easily run on the Web by remote client and provide decision information and measures for crop production in the remote field. In this paper we describe such a system for remote crop production manage. For each cell of the grid where remote client lives on, WebGIS tool will extract the input variables (climate, crop, soil, ...) for the crop model from the corresponding data layers, then the model will be run and the database will be updated with the results of the simulation. Spatial analysis and representation of crop model input data and output results can also be done effectively on the Web. Both the input data for the model and the outputs from the model can be displayed on a map. The soil, water, climate data sets are organized in GIS layers. Users will also be able to create thematic maps for the results of the simulations.