Rule–based reasoning and neural network perception for safe off–road robot mobility

Operational safety and health monitoring are critical matters for autonomous field mobile robots such as planetary rovers operating on challenging terrain. This paper describes relevant rover safety and health issues and presents an approach to maintaining vehicle safety in a mobility and navigation context. The proposed rover safety module is composed of two distinct components: safe attitude (pitch and roll) management and safe traction management. Fuzzy logic approaches to reasoning about safe attitude and traction management are presented, wherein inertial sensing of safety status and vision–based neural network perception of terrain quality are used to infer safe speeds of traversal. Results of initial field tests and laboratory experiments are also described. The approach provides an intrinsic safety cognizance and a capacity for reactive mitigation of robot mobility and navigation risks.     

L-UNet：轻量化云遮挡道路提取网络

目的 道路提取是常见的遥感应用之一。现有的基于深度卷积网络的道路提取方法往往未考虑云遮挡给道路提取带来的影响，且提取网络模型较大，不利于在移动端部署，同时缺乏用于云遮挡场景下的道路提取数据集。对此，本文提出一种轻量化的UNet网络（lightweight UNet，L-UNet），高效地实现云遮挡下的道路提取。方法 通过柏林噪声模拟云层以扩展现有道路提取数据集，进而训练L-UNet。使用移动翻转瓶颈卷积模块作为特征提取的主要结构，在深度可分离卷积的基础上加入扩展卷积和压缩激励模块，在减少参数量的同时大幅提升了分割效果。结果 在DeepGlobe道路提取扩展数据集的测试中，与D-LinkNet相比，L-UNet的交并比（intersection over union，IoU）提升了1.97%，而参数量仅为D-LinkNet的1/5。在真实云遮挡遥感图像道路提取测试中，L-UNet的性能仍然最优，与D-LinkNet和UNet相比，IoU值分别提高19.47%和31.87%。结论 L-UNet网络具有一定的云遮挡区域下道路标签生成能力，虽然在模拟云遮挡数据集下训练得到，但对于真实云遮挡仍具有较强的鲁棒性。L-UNet模型参数量很小，易于嵌入移动端。     

A fast and accurate online self-organizing scheme for parsimonious fuzzy neural networks

In this paper, we present a fast and accurate online self-organizing scheme for parsimonious fuzzy neural networks (FAOS-PFNN), where a novel structure learning algorithm incorporating a pruning strategy into new growth criteria is developed. The proposed growing procedure without pruning not only speeds up the online learning process but also facilitates a more parsimonious fuzzy neural network while achieving comparable performance and accuracy by virtue of the growing and pruning strategy. The FAOS-PFNN starts with no hidden neurons and parsimoniously generates new hidden units according to the proposed growth criteria as learning proceeds. In the parameter learning phase, all the free parameters of hidden units, regardless of whether they are newly created or originally existing, are updated by the extended Kalman filter (EKF) method. The effectiveness and superiority of the FAOS-PFNN paradigm is compared with other popular approaches like resource allocation network (RAN), RAN via the extended Kalman filter (RANEKF), minimal resource allocation network (MRAN), adaptive-network-based fuzzy inference system (ANFIS), orthogonal least squares (OLS), RBF-AFS, dynamic fuzzy neural networks (DFNN), generalized DFNN (GDFNN), generalized GAP-RBF (GGAP-RBF), online sequential extreme learning machine (OS-ELM) and self-organizing fuzzy neural network (SOFNN) on various benchmark problems in the areas of function approximation, nonlinear dynamic system identification, chaotic time-series prediction and real-world regression problems. Simulation results demonstrate that the proposed FAOS-PFNN algorithm can achieve faster learning speed and more compact network structure with comparably high accuracy of approximation and generalization.     

A fast and accurate face detector based on neural networks

Detecting faces in images with complex backgrounds is a difficult task. Our approach, which obtains state of the art results, is based on a neural network model: the constrained generative model (CGM). Generative, since the goal of the learning process is to evaluate the probability that the model has generated the input data, and constrained since some counter-examples are used to increase the quality of the estimation performed by the model. To detect side view faces and to decrease the number of false alarms, a conditional mixture of networks is used. To decrease the computational time cost, a fast search algorithm is proposed. The level of performance reached, in terms of detection accuracy and processing time, allows us to apply this detector to a real world application: the indexing of images and videos     

Convolutional neural network based heterogeneous transfer learning for remote-sensing scene classification

ABSTRACT

Deep convolutional neural network (CNN) transfer has recently shown strong performance in scene classification of high-resolution remote-sensing images. However, the majority of transfer learning solutions are categorized as homogeneous transfer learning, which ignores differences between target and source domains. In this paper, we propose a heterogeneous model to transfer CNNs to remote-sensing scene classification to correct input feature differences between target and source datasets. First, we extract filters from source images using the principal component analysis (PCA) method. Next, we convolute the target images with the extracted PCA filters to obtain an adopted target dataset. Then, a pretrained CNN is transferred to the adopted target dataset as a feature extractor. Finally, a classifier is used to accomplish remote-sensing scene classification. We conducted extensive experiments on the UC Merced dataset, the Brazilian coffee scene dataset and the Aerial Images Dataset to verify the effectiveness of the proposed heterogeneous model. The experimental results show that the proposed heterogeneous model outperforms the homogeneous model that uses pretrained CNNs as feature extractors by a wide margin and gains similar accuracies by fine-tuning a homogeneous transfer learning model with few training iterations.     

A fast, three-layer neural network for path finding

A path-planning algorithm is proposed to find a path based on local rules applied to a three-layer artificial neural network. Each layer consists of two-dimensionally arranged neurons with recurrent connections within a limited neighbourhood. The output of one layer determines the weights of the connections in the next layer. In principle, the method is based on a diffusion process, but is modified such that it does not suffer from several drawbacks involved in this algorithm. By application of a nonlinear transformation in layer 2, the diffusion front has the qualitative properties of a propagation wave. Therefore, limited resolution of the units is not critical, in contrast to classical diffusion algorithms. Furthermore, the algorithm generally does not suffer from the superposition of diffusion gradients when several paths are possible. The diffusion takes place in a space covered with 'obstacle potentials' which decrease the velocity of the diffusion front. In this way the path can maintain an adjustable safety margin in relation to the obstacles, for example, to cope with problems of incomplete knowledge of the obstacle's position. The algorithm thus combines the advantages of the diffusion algorithm, namely avoidance of local minima, of wave propagation, i.e. coping with limited resolution, and the potential field approach, i.e. maintaining a safety margin in relation to obstacles. The distributed architecture also allows for 'spatial interpolation' between the units (coarse coding), thereby providing smooth path forms. A comparison with paths developed by human subjects shows some similarity on the qualitative level, but there are also obvious differences.     

基于深度卷积神经网络的道路场景理解

在无人驾驶技术中,道路场景的理解是一个非常重要的环境感知任务,也是一个很具有挑战性的课题。提出了一个深层的道路场景分割网络（Road Scene Segmentation Network,RSSNet）,该网络为32层的全卷积神经网络,由卷积编码网络和反卷积解码网络组成。网络中采用批正则化层防止了深度网络在训练中容易出现的“梯度消失”问题;在激活层中采用了Maxout激活函数,进一步缓解了梯度消失,避免网络陷入饱和模式以及出现神经元死亡现象;同时在网络中适当使用Dropout操作,防止了模型出现过拟合现象;编码网络存储了特征图的最大池化索引并在解码网络中使用它们,保留了重要的边缘信息。实验证明,该网络能够大大提高训练效率和分割精度,有效识别道路场景图像中各像素的类别并对目标进行平滑分割,为无人驾驶汽车提供有价值的道路环境信息。     

Adaptive fuzzy color segmentation with neural network for road detections

In this paper a new color space, called the RGB color ratio space, is proposed and defined according to a reference color such that an image can be transformed from a conventional color space to the RGB color ratio space. Because a color in the RGB color ratio space is represented as three color ratios and intensity, the chrominance can be completely reserved (three color ratios) and the luminance can be de-correlated with the chrominance. Different from traditional distance measurement, a road color model is determined by an ellipse area in the RGB ratio space enclosed by the estimated boundaries. A proposed adaptive fuzzy logic in which fuzzy membership functions are defined according to estimated boundaries is introduced to implement clustering rules. Therefore, each pixel will have its own fuzzy membership function corresponding to its intensity. A basic neural network is trained and used to achieve parameters optimization. The low computation cost of the proposed segmentation method shows the feasibility for real time application. Experimental results for road detection demonstrate the robustness to intensity variation of the proposed approach.     

LRCN-RetailNet: A recurrent neural network architecture for accurate people counting

Multimedia Tools and Applications - Measuring and analyzing the flow of customers in retail stores is essential for a retailer to better comprehend customers’ behavior and support...     

Pooling spike neural network for fast rendering in global illumination

Neural Computing and Applications - The generation of photo-realistic images is a major topic in computer graphics. By using the principles of physical light propagation, images that are...     

Composite adaptive control with fast convergence for multilayer neural network

A composite adaptive control (CAC) that combines the benefits of direct and indirect adaptive controls has better parameter adaptation and control response. Multilayer neural networks (NNs) can be employed to enhance a model's representation capacity, but previous composite adaptive approaches cannot easily train the model due to its nonlinearities. A novel CAC is therefore developed in this study to tackle the above limitations. A modified robust version is adopted by focusing on the direct adaptive part to enhance robustness of adaption. Then, the indirect parameter adaptive law is improved by adopting a small learning rate in which a multistep adaption update is executed in one control interval. Moreover, multistep prediction errors are implemented to guarantee the consistency of the approximation errors, and an experience replay technique is adopted to attenuate the requirement of persistent excitation conditions. These improvements not only accelerate the convergence process but also smoothen the updating of NN parameters. Given that a nonlinear plant with MIMO strict‐feedback structure is considered, the proposed CAC is integrated into the backstepping framework. The uniformly bounded property of the tracking errors and the approximation errors is proven by Lyapunov theory. The superiority of the proposed method and the roles of these improvements are demonstrated by comparative simulations.     

Boosting a fast neural network for supervised land cover classification

It is demonstrated that the use of an ensemble of neural networks for routine land cover classification of multispectral satellite data can lead to a significant improvement in classification accuracy. Specifically, the AdaBoost.M1 algorithm is applied to a sequence of three-layer, feed-forward neural networks. In order to overcome the drawback of long training time for each network in the ensemble, the networks are trained with an efficient Kalman filter algorithm. On the basis of statistical hypothesis tests, classification performance on multispectral imagery is compared with that of maximum likelihood and support vector machine classifiers. Good generalization accuracies are obtained with computation times of the order of 1 h or less. The algorithms involved are described in detail and a software implementation in the ENVI/IDL image analysis environment is provided.     

Indoor scene segmentation algorithm based on full convolutional neural network

Neural Computing and Applications - With the leaps and bounds of computer performance and the advent of the era of big data, deep learning has drawn more and more attention from all walks of life....     

基于改进时空残差卷积神经网络的城市路网短时交通流预测

城市路网交通流预测受到历史交通流和相邻路口交通流的影响,具有复杂的时空关联性.针对传统时空残差模型缺乏对交通流数据进行相关性分析、捕获微小变化而容易忽略长期时间特征等问题,提出一种基于改进时空残差卷积神经网络(CNN)的城市路网短时交通流预测模型.该模型将原始交通流数据转化成交通栅格数据,利用皮尔逊相关系数(PCC)对...     

An optimized discrete neural network in embedded systems for road recognition

A type of optimized neural networks with limited precision weights (LPWNN) is presented in this paper. Such neural networks, which require less memory for storing the weights and less expensive floating point units in order to perform the computations involved, are better suited for embedded systems implementation than the real weight ones. Based on analyzing the learning capability of LPWNN, Quantize Back-propagation Step-by-Step (QBPSS) algorithm is proposed for such neural networks to overcome the effects of limited precision. Methods of designing and training LPNN are represented, including the quantization of non-linear activation function and the selection of learning rate, network architecture and weights precision. The optimized LPWNN performance has been evaluated by comparing to conventional neural networks with double-precision floating-point weights on road recognition of image for intelligent vehicle in ARM 9 embedded systems, and the results show the optimized LPWNN has 7 times faster than the conventional ones.     

多路径卷积神经网络的轮廓感知

目的 引入视觉信息流的整体和局部处理机制,提出了一种多路径卷积神经网络的轮廓感知新方法。方法 利用高斯金字塔尺度分解获得低分辨率子图,用来表征视觉信息中的整体轮廓;通过2维高斯导函数模拟经典感受野的方向选择性,获得描述细节特征的边界响应子图;构建多路径卷积神经网络,利用具有稀疏编码特性的子网络（Sparse-Net）实现对整体轮廓的快速检测;利用具有冗余度增强编码特性的子网络（Redundancy-Net）实现对局部细节特征提取;对上述多路径卷积神经网络响应进行融合编码,以实现轮廓响应的整体感知和局部检测融合,获取轮廓的精细化感知结果。结果 以美国伯克利大学计算机视觉组提供的数据集BSDS500图库为实验对象,在GTX1080Ti环境下本文Sparse-Net对整体轮廓的检测速度达到42幅/s,为HFL方法1.2幅/s的35倍;而Sparse-Net和Redundancy-Net融合后的检测指标数据集尺度上最优（ODS）、图片尺度上最优（OIS）、平均精度（AP）分别为0.806、0.824、0.846,优于HED （holistically-nested edge detection）方法和RCF （richer convolution features for edge detection）方法,结果表明本文方法能有效突出主体轮廓并抑制纹理背景。结论 多路径卷积神经网络的轮廓感知应用,将有助于进一步理解视觉感知机制,并对减弱卷积神经网络的黑盒特性有着重要的意义。     

基于多分支网络的道路场景实时语义分割方法

针对目前实时语义分割方法存在大目标分割不准确、小目标信息丢失的问题,提出一种基于多分支网络的实时语义分割算法。首先,对双边分割网络进行优化,设计了金字塔分支扩大感受野,以覆盖视野内的大目标,充分地将上下文信息结合起来;其次,设计双边指导融合模块,为深层和浅层的特征映射提供指导信息,弥补小目标信息的损失。最后在Cityscapes数据集上进行验证,实验结果表明所提模型以51.3 fps的推理速度使平均交并比达到77.8%,与基准相比,精度提高了2.5个百分点。所提方法采用金字塔分支,在扩大感受野的同时,获取不同尺度的语义边缘区域特性,增强对语义边界的建模能力,且提出的双边指导融合模块可以更有效地融合不同层次的特征,弥补下采样造成的信息丢失,能够更好地指导模型学习。     

Flexible memristor based neuromorphic system for implementing multi-layer neural network algorithms

Abstract

This paper describes a memristor-based neuromorphic system that can be used for ex situ training of various multi-layer neural network algorithms. This system is based on an analogue neuron circuit that is capable of performing an accurate dot product calculation. The presented ex situ programming technique can be used to map many key neural algorithms directly onto the grid of resistances in a memristor crossbar. Using this weight-to-crossbar mapping approach along with the memristor based circuit architecture, complex neural algorithms can be easily implemented using this system. Some existing memristor based circuits provide an approximated dot product based on conductance summation, but neuron outputs are not directly correlated to the numerical values obtained in a traditional software approach. To show the effectiveness and versatility of this circuit, two different powerful neural networks were simulated. These include a Restricted Boltzmann Machine for character recognition and a Multilayer Perceptron trained to perform Sobel edge detection. Following these simulations, an analysis was presented that shows how both memristor accuracy and neuron circuit gain relates to output error.     

全卷积网络结合改进的条件随机场循环神经网络用于SAR图像场景分类

传统合成孔径雷达（SAR）图像基于粗分割像素块提取相关特征,后接支持向量机（SVM）和马尔可夫随机场（MRF）或条件随机场（CRF）进行分类,该方法存在同一像素块内部不同类别像素的误差,而且只考虑邻近区域未充分用到全局信息和结构信息。故考虑基于像素点引入全卷积网络（FCN）,以ESAR卫星图像为样本,基于像素点级别构建卷积网络进行训练,得到各像素的初始类别分类概率。为了考虑全局像素类别的影响后接CRF-循环神经网络（CRF-RNN）,利用FCN得到的初始概率,结合CRF结构得到全局像素类别转移结果,之后进行RNN的迭代进一步优化实验结果。由于基于像素点和考虑了全局信息与结构信息,克服了传统分类的部分缺点,使正确率较传统SVM或CRF方法平均提高了约6.5个百分点。由于CRF-RNN的距离权重是用高斯核人为拟合的,不能随实际训练样本来改变和确定,故存在一定误差,针对该问题提出可训练的全图距离权重卷积网络来改进CRF-RNN,最终实验结果表明改进后方法的正确率较未改进的CRF-RNN又提高了1.04个百分点。     

Approximating support vector machine with artificial neural network for fast prediction

Support vector machine (SVM) is a powerful algorithm for classification and regression problems and is widely applied to real-world applications. However, its high computational load in the test phase makes it difficult to use in practice. In this paper, we propose hybrid neural network (HNN), a method to accelerate an SVM in the test phase by approximating the SVM. The proposed method approximates the SVM using an artificial neural network (ANN). The resulting regression function of the ANN replaces the decision function or the regression function of the SVM. Since the prediction of the ANN requires significantly less computation than that of the SVM, the proposed method yields faster test speed. The proposed method is evaluated by experiments on real-world benchmark datasets. Experimental results show that the proposed method successfully accelerates SVM in the test phase with little or no prediction loss.