Search

We demonstrate a power combining pulse laser system with 1 kW average power output based on spectrum filtering of four subset pulse amplifiers. The output spectrum with and without spectrum filtering are compared. The result shows that, without spectrum filtering, nonlinear spectrum component from subset amplifier will accumulate along the QBH fiber and leads to serious spectrum broadening as well as stimulated Raman scattering (SRS) effect. By spectrum filtering with a band pass filter (BPF) inside the subset amplifier, nonlinear spectrum component is successfully mitigated and high purity spectrum pulse laser output is also achieved at the output power of 1kW.

In order to solve the problems of long time for selecting candidate areas of obstacles and low accuracy of obstacle recognition. In this paper, we proposed an obstacle detection and recognition method based on stereo vision and convolution neural network. First, we use semi-global stereo matching to calculate disparity between left and right images. A method for Stixel computation is used to obtain the candidate regions of obstacles. Then, we use U-disparity map to extract the target obstacles from the candidate regions of obstacles. Then, we propose a new convolutional neural network to recognize the target obstacles. In order to verify the recognition performance of the new convolutional neural network, we compare the results with other convolutional neural networks on CIFAR-10 datasets. The experimental results show that our proposed convolutional neural network improves the real-time performance by 69% when the recognition accuracy is reduced by 4.9%.

Visual-based semantic segmentation for traffic scene plays an important role in intelligent vehicles. In this paper, we present a new real-time deep fully convolution neural network (FCNN) for pixel-wise segmentation with six channel inputs. The six channel inputs include the RGB three channel color image, the Disparity (D) image generated by stereo vision sensor, the image to describe the Height (H) of each pixel above road ground, and the image to describe the Angle (A) between each pixel normal direction and the predicted direction of gravity, which are defined as a RGB-DHA multi-feature map. The FCNN is simplified and modified based on AlexNet to meet the real-time requirements of intelligent vehicle for environmental perception. The proposed algorithm is tested and compared in Cityscapes dataset, yields global accuracies 73.4% and 22ms for $400 \times 200$ resolution image with one Titan X GPU.