To solve the problem of blind travel, a guide system based on image processing is designed. The visual information collected by a CCD camera. The voice information is used to communicate with the user. The overall design of this system is given. Recognize the zebra line by the bipolar system value. Recognize...

Traditional dehazing methods always use the dark channel a priori method to recover the original image, which leads to the calculation of the transmission is large and takes too much time. Therefore, in this paper, we present a fog-day-based image sharpening algorithm based on depth of field estimation....

This paper proposes a method for welding process surveillance, using the CCD camera and the spectrograph for the acquisition of the welding arc image and the spectral information. The spectral information inferred the temperature distribution and the metallic vapor density in the welding arc space. The...

In this paper, we present a new framework to detect pedestrians in infrared images. The framework consists of a candidate generation module and a classification module, both of which are implemented based on convolution neural network. Specifically, we learned effective segmentation threshold by deep...

The aim of this study is detecting disease edge of the medical CT (Computed Tomography) image. This paper describes the method by using 3-Tap bandpass filter signal and two high frequency sub-band CT images in wavelet domain for detecting part of disease. Simulation result show that the proposed method...

When talking about auditory perception, it remains unclear how people could achieve sound identification and localization efficiently even in the middle of chaos. Though the dual-pathway model is widely held, there are still controversies in specific brain regions. Moreover, some machine learning methods...

In order to solve the problem of insufficient accuracy of the existing person re-identification methods. We propose a neural network model for identifying pedestrian properties and pedestrian ID. Compared with the existing methods, the model mainly has the following three advantages. First, our network...

We propose a method to perform face detection and facial keypoint detection in parallel. On the basis of the traditional object detection framework Faster R-CNN for face detection, we have designed a facial keypoint detection network with the same base network. And the two networks share the weights...

To aim at the challenge of face recognition to uncontrolled situations, an ensemble method of global and local representation for face recognition is presented in this paper. shearlets transform is firstly employed to decompose a image into subimages. Then directional information is utilized along with...

SRGAN has successfully applied the Generative Adversarial Network to the single image super-resolution reconstruction, which has achieved good results. But the loss function based on feature space in SRGAN objectively sacrifices the pursuit of high peak signal-to-noise-ratio (PSNR), which is the result...

In the multi-spectral image registration, because of the different imaging mechanism, the distribution of image's pixel gray value is non-linear. So the traditional matching algorithm based on gray correlation and image feature cannot be used simply. Since the mutual information measure does not need...

Grid cells in the MEC (media entorhinal cortex) of rats show repetitive hexagon firing pattern while moving through the environment. In the deeper layer of the MEC, a kind of interneurons have both the head-directing (HD) and grid-like firing properties. The predecessors' observations indicate the grid...

The vehicle kinematic parameter is important evaluation index of vehicle performance. But the measurement technique for vehicle kinematics parameters is still immature. Practical and efficient measuring system is now in badly demand. With the development of machine vision technique, more and more measurement...

Emotion recognition rapidly gains interest from research community, which gives computer the ability to recognize, understand, express, and adapt to human emotions to help computer more intelligent. Traditional emotion recognition modalities (e.g. face and voice) always could not identify the totally...

Named Entity Recognition(NER) is a very important part of many Natural Language Processing(NLP) tasks, but the accuracy rate of NER has not reached our expectation, especially in Chinese. Therefore, we studied NER task in Chinese social media. Compared to the previous papers on this dataset, we propose...

Multi-target tracking is a popular research topic in the field of computer vision. Most recent works based on tracking-by-detection paradigm have trouble when dealing with tracking errors (i.e. identity switches) that are caused by overlapping of different pedestrians. In this work, we propose a novel...