This article addresses the challenges faced by Gm-APD detectors in lidar systems when detecting weak signals. We use a triggering model based on the negative binomial distribution, which replaces the traditional Poiss...
详细信息
ISBN:
(数字)9798331542283
ISBN:
(纸本)9798331542290
This article addresses the challenges faced by Gm-APD detectors in lidar systems when detecting weak signals. We use a triggering model based on the negative binomial distribution, which replaces the traditional Poisson distribution model, allowing for a more accurate description of the echo characteristics from rough surface targets. Building on this foundation, we have improved the conventional matched filtering algorithm by integrating the laser pulse waveform with the actual triggering response of the detector, thereby enhancing the detection rate. Experimental results demonstrate that the improved algorithm significantly increases the detection probability under low signal-to-noise ratios and with a limited number of frames, confirming the effectiveness of the new triggering model and the enhanced algorithm.
To solve the problem that current dynamic intention recognition methods fail to make full use of time domain variation information between multi-temporal group targets, which leads to low accuracy of intention recogni...
To solve the problem that current dynamic intention recognition methods fail to make full use of time domain variation information between multi-temporal group targets, which leads to low accuracy of intention recognition. This paper proposes a bidirectional convolutional long short term memory-attention network for marine formation target intention recognition. The network takes the multi-source trajectory data of the marine ship formation target as the input, extracts and uses the target change rule and time domain characteristics of the Marine formation data, and trains the model to have the ability to independently learn the weight of information in different time periods. The simulation results show that the method has good performance and can meet the needs of practical application.
GM-APD lidar has single-photon sensitivity, so that it can obtain 3D information of distant targets. On the other hand, the laser echo of targets is easily drowned by noise. In order to solve this problem, an interval...
详细信息
ISBN:
(数字)9798331542283
ISBN:
(纸本)9798331542290
GM-APD lidar has single-photon sensitivity, so that it can obtain 3D information of distant targets. On the other hand, the laser echo of targets is easily drowned by noise. In order to solve this problem, an interval-peak-based algorithm for range image reconstruction is proposed in this paper. Firstly, in the photon counts histogram, the x-axis is divided into small intervals and the peak value of each interval is taken. This step weakens the bad characteristic caused by the trigger probability. Secondly, according to the data feature, the trigger probability ratio coefficient and the neighborhood correlation coefficient are obtainned. Finally, the range image can be extracted by the peak method after the peak value of each interval is multiplied with this two coefficients. It is verified by the experimental data that when the SNR of target echo data is 0.022, the recovery ratio of the proposed algorithm can reach 80%, which is 25% higher than the matching filter algorithm. The proposed algorithm has a good application prospect in the range image reconstruction of GM-APD lidar.
Visual-Inertial Odometry(VIO) fuses measurements from camera and Inertial Measurement Unit(IMU) to achieve accumulative performance that is better than using individual *** VIO is an extended Kalman filter-based solut...
详细信息
Visual-Inertial Odometry(VIO) fuses measurements from camera and Inertial Measurement Unit(IMU) to achieve accumulative performance that is better than using individual *** VIO is an extended Kalman filter-based solution which augments features with long tracking length into the state vector of Multi-State Constraint Kalman Filter(MSCKF). In this paper, a novel hybrid VIO is proposed, which focuses on utilizing low-cost sensors while also considering both the computational efficiency and positioning precision. The proposed algorithm introduces several novel contributions. Firstly, by deducing an analytical error transition equation, onedimensional inverse depth parametrization is utilized to parametrize the augmented feature *** modification is shown to significantly improve the computational efficiency and numerical robustness, as a result achieving higher precision. Secondly, for better handling of the static scene,a novel closed-form Zero velocity UPda Te(ZUPT) method is proposed. ZUPT is modeled as a measurement update for the filter rather than forbidding propagation roughly, which has the advantage of correcting the overall state through correlation in the filter covariance matrix. Furthermore, online spatial and temporal calibration is also incorporated. Experiments are conducted on both public dataset and real data. The results demonstrate the effectiveness of the proposed solution by showing that its performance is better than the baseline and the state-of-the-art algorithms in terms of both efficiency and precision. A related software is open-sourced to benefit the community.
Oriented towards the requirements for reliable positioning and navigation of aircraft under the condition of rejection of navigation satellite, we proposed cross-domain guide positioning methods based on multi-layer n...
详细信息
Zero-shot object detection aims to localize and recognize objects of unseen classes. Most of existing works face two problems: the low recall of RPN in unseen classes and the confusion of unseen classes with backgroun...
Zero-shot object detection aims to localize and recognize objects of unseen classes. Most of existing works face two problems: the low recall of RPN in unseen classes and the confusion of unseen classes with background. In this paper, we present the first method that combines DETR and meta-learning to perform zero-shot object detection, named Meta-ZSDETR, where model training is formalized as an individual episode based meta-learning task. Different from Faster R-CNN based methods that firstly generate class-agnostic proposals, and then classify them with visual-semantic alignment module, Meta-ZSDETR directly predict class-specific boxes with class-specific queries and further filter them with the predicted accuracy from classification head. The model is optimized with meta-contrastive learning, which contains a regression head to generate the coordinates of class-specific boxes, a classification head to predict the accuracy of generated boxes, and a contrastive head that utilizes the proposed contrastive-reconstruction loss to further separate different classes in visual space. We conduct extensive experiments on two benchmark datasets MS COCO and PASCAL VOC. Experimental results show that our method outperforms the existing ZSD methods by a large margin.
Zero-shot object detection aims to localize and recognize objects of unseen classes. Most of existing works face two problems: the low recall of RPN in unseen classes and the confusion of unseen classes with backgroun...
详细信息
The simulation of aircraft guidance and controlsystem has characteristics such as multiple model classification, large parameter influence,complex information interaction and so on. The traditional code-level model d...
详细信息
Synthetic aperture radar (SAR) image registration is widely used in integrated navigation system with development of guidance system in sensors and other applications. However, there exists some difficulty because of ...
Synthetic aperture radar (SAR) image registration is widely used in integrated navigation system with development of guidance system in sensors and other applications. However, there exists some difficulty because of the quality and imaginary principle of SAR images. In this article, a brand-new method based on feature points using improved CSP-DenseNet is proposed to solve the problems of SAR image registration with weak and noisy texture. Deep features of interest points in SAR images are extracted using the proposed network from the crops of search image and template image respectively. The method has an advantage of preserving abundant feature information in SAR images under the influence of insufficient image information and noise of SAR images. The CSP-DenseNet architecture has a cross-stage construction optimizing fused by CSP-Net and DenseNet to extract matching features. The network is partially connected in specific convolutional layers for feature reusing and computation resources saving. Then, Brute Force Matcher and RANSAC voting are successively applied for a larger number of matching pairs and high-precise matching vertexes. Experimental results on various SAR image matching methods show that the proposed method provides better performance than other approaches compared.
Due to the strong scattering intensity of the apex of the carrier, a scheme of rounding the apex of the carrier is proposed. In this design, it is embodied in face fillet and line fillet. Based on the above designs, t...
详细信息
暂无评论