Single-image night haze removal based on color channel transfer and estimation of spatial variation in atmospheric light

The visible-light imaging system used in military equipment is often subjected to severe weather conditions, such as fog, haze, and smoke, under complex lighting conditions at night that significantly degrade the acquired images. Currently available image defogging methods are mostly suitable for environments with natural light in the daytime, but the clarity of images captured under complex lighting conditions and spatial changes in the presence of fog at night is not satisfactory. This study proposes an algorithm to remove night fog from single images based on an analysis of the statistical characteristics of images in scenes involving night fog. Color channel transfer is designed to compensate for the high attenuation channel of foggy images acquired at night. The distribution of transmittance is estimated by the deep convolutional network DehazeNet, and the spatial variation of atmospheric light is estimated in a point-by-point manner according to the maximum reflection prior to recover the clear image. The results of experiments show that the proposed method can compensate for the high attenuation channel of foggy images at night, remove the effect of glow from a multi-color and non-uniform ambient source of light, and improve the adaptability and visual effect of the removal of night fog from images compared with the conventional method.

An Approach to Checking 3D Model with Related Engineering Drawings

For some reasons, engineers build their product 3D mo del according to a set of related engineering drawings. The problem is how we ca n know the 3D model is correct. The manual checking is very boring and time cons uming, and still could not avoid mistakes. Thus, we could not confirm the model, maybe try checking again. It will effect the production preparing cycle greatly , and should be solved in a intelligent way. The difficulties are quite obvious, unlike word checking in a word processing package, the checking described above is not a comparison between same items. One is 2D drawing, the another is 3D mo del, they are not in the same dimension. So, we should make a change for compari son in the same dimension. If we can rebuild a 3D model through related 2D drawi ngs automatically, that’s great. We can not only compare two 3D models to check and correct, but also omit the manual process itself completely. Unfortunately, we can not build such a 3D model automatically right now. So only one way left: compare two 2D drawings, one is the original, the another is processed from tha t manual built one.The method is to select a drawing as a background, rotate th e 3D model and make projections, compare projection with the background automati cally to find a case which they meet each other in certain amount of error ( tolerance), otherwise alarm. This process can be repeated many times if needed t o fulfil the checking task. Also, this is a man-machine system, computer does h ard working, man keeps final decision. The project involved in CAD, VRML, patter n recognition, image capture and comparison, artificial intelligence.

Computer generated hologram from full-parallax 3D image data captured by scanning vertical camera array(Invited Paper)

Full-parallax light-field is captured by a small-scale 3D image scanning system and applied to holographic display. A vertical camera array is scanned horizontally to capture full-parallax imagery, and the vertical views between cameras are interpolated by depth image-based rendering technique. An improved technique for depth estimation reduces the estimation error and high-density light-field is obtained. The captured data is employed for the calculation of computer hologram using ray-sampling plane. This technique enables high-resolution display even in deep 3D scene although a hologram is calculated from ray information, and thus it makes use of the important advantage of holographic 3D display.

Efcient autofocus method for sequential automatic capturing of high-magnification microscopic images

This letter presents an autofocus （AF） method to position a high-magnification microscope lens that automatically captures hundreds of images from a single moving slide. These images are taken by a mobile clinic unit in a rural location, and are later automatically processed and revised by a remote specialist. This process requires high focus precision to enable image processing techniques to achieve proper results. Low focusing times are also required for the system to be operative. We propose a novel method that combines two focus measures with an adapted searching scheme to cope with both constraints.

Kiwifruit recognition at nighttime using artificial lighting based on machine vision

Most researches involved so far in kiwifruit harvesting robot suggest the scenario of harvesting in daytime for taking advantage of sunlight.A robot operating at nighttime can overcome the problem of low work efficiency and would help to minimize fruit damage.In addition,artificial lights can be used to ensure constant illumination instead of the variable natural sunlight for image capturing.This paper aims to study the kiwifruit recognition at nighttime using artificial lighting based on machine vision.Firstly,an RGB camera was placed underneath the canopy so that clusters of kiwifruits could be included in the images.Next,the images were segmented using an R-G color model.Finally,a group of image processing conventional methods,such as Canny operator were applied to detect the fruits.The image processing results showed that this capturing method could reduce the background noise and overcome any target overlapping.The experimental results showed that the optimal artificial lighting ranged approximately between 30-50 lx.The developed algorithm detected 88.3%of the fruits successfully.