The importance of panoramic traffic perception tasks in autonomous driving is increasing, so shared networks with high accuracy are becoming increasingly important. In this paper, we propose a multi-task shared sensing network, called CenterPNets, that can perform the three major detection tasks of target detection, driving area segmentation, and lane detection in traffic sensing in one go and propose several key optimizations to improve the overall detection performance. First, this paper proposes an efficient detection head and segmentation head based on a shared path aggregation network to improve the overall reuse rate of CenterPNets and an efficient multi-task joint training loss function to optimize the model. Secondly, the detection head branch uses an anchor-free frame mechanism to automatically regress target location information to improve the inference speed of the model. Finally, the split-head branch fuses deep multi-scale features with shallow fine-grained features, ensuring that the extracted features are rich in detail. CenterPNets achieves an average detection accuracy of 75.8% on the publicly available large-scale Berkeley DeepDrive dataset, with an intersection ratio of 92.8% and 32.1% for driveableareas and lane areas, respectively. Therefore, CenterPNets is a precise and effective solution to the multi-tasking detection issue.

Aiming at the problems of the difficult extraction of small target feature information, complex background, and variable target scale in unmanned aerial vehicle (UAV) aerial photography images. In this paper, an anchor-free target detection algorithm based on fully convolutional one-stage object detection (FCOS) for UAV aerial photography images is proposed. For the problem of complex backgrounds, the global context module is introduced in the ResNet50 network, which is combined with feature pyramid networks (FPN) as the backbone feature extraction network to enhance the feature representation of targets in complex backgrounds. To address the problem of the difficult detection of small targets, an adaptive feature balancing sub-network is designed to filter the invalid information generated at all levels of feature fusion, strengthen multi-layer features, and improve the recognition capability of the model for small targets. To address the problem of variable target scales, complete intersection over union (CIOU) Loss is used to optimize the regression loss and strengthen the model’s ability to locate multi-scale targets. The algorithm of this paper is compared quantitatively and qualitatively on the VisDrone dataset. The experiments show that the proposed algorithm improves 4.96% on average precision (AP) compared with the baseline algorithm FCOS, and the detection speed is 35 frames per second (FPS), confirming that the algorithm has satisfactory detection performance, real-time inference speed, and has effectively improved the problem of missed detection and false detection of targets in UAV aerial images.

This paper reports a novel approach to synthesize pazopanib. In our synthetic route, the potently mutagenic alkylating agents such as dimethyl sulfate and methyl iodide are avoided. A novel regioselective methylation of the 2- position of 3-methyl-6-nitro-1H-indazole was reported. This novel route is one step shorter than the previously reported route.

Aiming at the problems of large amount of point cloud data, long registration time and low registration accuracy in the registration process of point cloud, a point cloud registration algorithm based on intrinsic shape signatures (ISS) and 3D shape context (3DSC) was proposed. Firstly, the down-sampling of point cloud was carried out by using a voxel grid filter. Then, the ISS algorithm was adopted to extract the feature points, which were described by 3DSC, and the rough matching was performed according to the improved random sample consensus (RANSAC) algorithm. Finally, the improved iterative closest point (ICP) algorithm was utilized to accurately match the point cloud. The experimental results show that compared with 3D normal distribution transformation (NDT) algorithm based on ISS+3DSC and ICP algorithm based on sample consensus initial aligment (SAC-IA), the proposed algorithm has higher registration accuracy and efficiency, and also has a better matching effect on point clouds with large data volume.

Video semantic segmentation methods are widely used in autonomous driving, traffic management, drone driving and other fields. Aiming at the problems of inaccurate segmentation results and too long segmentation time to achieve real-time performance in video semantic segmentation, a memory network segmentation algorithm based on spatio-temporal dual-branch attention mechanism is proposed. The algorithm is mainly composed of memory storage module, feature integration module and spatio-temporal dual branch attention module. The Historical frame feature information and the sequence time information are stored in the memory storage module. The feature integration module uses large-core decomposition convolution to expand the convolution receptive field without increasing the amount of calculation, and captures long-distance dependent information and context information well to achieve channel compression and feature reduction. The spatio-temporal dual-branch attention module captures the feature information of the same object in adjacent frames and performs feature fusion, using the timing of video frames to improve segmentation accuracy. The algorithm achieves 76.92% and 73.68% mIoU on the public datasets Cityscapes and CamVid, and the calculation speed reaches 38 frames/s. The experiment proves that this method has reached the optimal state in segmentation accuracy and calculation efficiency compared with other video semantic segmentation algorithms.