Enhancing Hyper-Spectral Image Classification with Reinforcement Learning and Advanced Multi-Objective Binary Grey Wolf Optimization

Hyperspectral(HS)image classification plays a crucial role in numerous areas including remote sensing(RS),agriculture,and the monitoring of the environment.Optimal band selection in HS images is crucial for improving the efficiency and accuracy of image classification.This process involves selecting the most informative spectral bands,which leads to a reduction in data volume.Focusing on these key bands also enhances the accuracy of classification algorithms,as redundant or irrelevant bands,which can introduce noise and lower model performance,are excluded.In this paper,we propose an approach for HS image classification using deep Q learning(DQL)and a novel multi-objective binary grey wolf optimizer(MOBGWO).We investigate the MOBGWO for optimal band selection to further enhance the accuracy of HS image classification.In the suggested MOBGWO,a new sigmoid function is introduced as a transfer function to modify the wolves’position.The primary objective of this classification is to reduce the number of bands while maximizing classification accuracy.To evaluate the effectiveness of our approach,we conducted experiments on publicly available HS image datasets,including Pavia University,Washington Mall,and Indian Pines datasets.We compared the performance of our proposed method with several state-of-the-art deep learning(DL)and machine learning(ML)algorithms,including long short-term memory(LSTM),deep neural network(DNN),recurrent neural network(RNN),support vector machine(SVM),and random forest(RF).Our experimental results demonstrate that the Hybrid MOBGWO-DQL significantly improves classification accuracy compared to traditional optimization and DL techniques.MOBGWO-DQL shows greater accuracy in classifying most categories in both datasets used.For the Indian Pine dataset,the MOBGWO-DQL architecture achieved a kappa coefficient(KC)of 97.68%and an overall accuracy(OA)of 94.32%.This was accompanied by the lowest root mean square error(RMSE)of 0.94,indicating very precise predictions with minimal error.In the case of the Pavia University dataset,the MOBGWO-DQL model demonstrated outstanding performance with the highest KC of 98.72%and an impressive OA of 96.01%.It also recorded the lowest RMSE at 0.63,reinforcing its accuracy in predictions.The results clearly demonstrate that the proposed MOBGWO-DQL architecture not only reaches a highly accurate model more quickly but also maintains superior performance throughout the training process.

DGConv: A Novel Convolutional Neural Network Approach for Weld Seam Depth Image Detection

We propose a novel image segmentation algorithm to tackle the challenge of limited recognition and segmentation performance in identifying welding seam images during robotic intelligent operations.Initially,to enhance the capability of deep neural networks in extracting geometric attributes from depth images,we developed a novel deep geometric convolution operator(DGConv).DGConv is utilized to construct a deep local geometric feature extraction module,facilitating a more comprehensive exploration of the intrinsic geometric information within depth images.Secondly,we integrate the newly proposed deep geometric feature module with the Fully Convolutional Network(FCN8)to establish a high-performance deep neural network algorithm tailored for depth image segmentation.Concurrently,we enhance the FCN8 detection head by separating the segmentation and classification processes.This enhancement significantly boosts the network’s overall detection capability.Thirdly,for a comprehensive assessment of our proposed algorithm and its applicability in real-world industrial settings,we curated a line-scan image dataset featuring weld seams.This dataset,named the Standardized Linear Depth Profile(SLDP)dataset,was collected from actual industrial sites where autonomous robots are in operation.Ultimately,we conducted experiments utilizing the SLDP dataset,achieving an average accuracy of 92.7%.Our proposed approach exhibited a remarkable performance improvement over the prior method on the identical dataset.Moreover,we have successfully deployed the proposed algorithm in genuine industrial environments,fulfilling the prerequisites of unmanned robot operations.

Applying Source Parameter Imaging (SPI) to Aeromagnetic Data to Estimate Depth to Magnetic Sources in the Mamfe Sedimentary Basin

Aeromagnetic data over the Mamfe Basin have been processed. A regional magnetic gridded dataset was obtained from the Total Magnetic Intensity (TMI) data grid using a 3 × 3 convolution (Hanning) filter to remove regional trends. Major similarities in magnetic field orientation and intensities were observed at identical locations on both the regional and TMI data grids. From the regional and TMI gridded datasets, the residual dataset was generated which represents the very shallow geological features of the basin. Processing this residual data grid using the Source Parameter Imaging (SPI) for magnetic depth suggests that the estimated depths to magnetic sources in the basin range from about 271 m to 3552 m. The highest depths are located in two main locations somewhere around the central portion of the study area which correspond to the area with positive magnetic susceptibilities, as well as the areas extending outwards across the eastern boundary of the study area. Shallow magnetic depths are prominent towards the NW portion of the basin and also correspond to areas of negative magnetic susceptibilities. The basin generally exhibits a variation in depth of magnetic sources with high, average and shallow depths. The presence of intrusive igneous rocks was also observed in this basin. This characteristic is a pointer to the existence of geologic resources of interest for exploration in the basin.

Three-Dimensional Space Interpolation of Grey / Depth Image Sequence-A New Technique of Computer Graphics Synthesis

This paper advances a three-dimensional space interpolation method of grey / depth image sequence, which breaks free from the limit of original practical photographing route. Pictures can cruise at will in space. By using space sparse sampling, great memorial capacity can be saved and reproduced scenes can be controlled. To solve time consuming and complex computations in three-dimensional interpolation algorithm, we have studied a fast and practical algorithm of scattered space lattice and that of 'Warp' algorithm with proper depth. By several simple aspects of three dimensional space interpolation, we succeed in developing some simple and practical algorithms. Some results of simulated experiments with computers have shown that the new method is absolutely feasible.

基于Depth-wise卷积和视觉Transformer的图像分类模型

图像分类作为一种常见的视觉识别任务,有着广阔的应用场景。在处理图像分类问题时,传统的方法通常使用卷积神经网络,然而,卷积网络的感受野有限,难以建模图像的全局关系表示,导致分类精度低,难以处理复杂多样的图像数据。为了对全局关系进行建模,一些研究者将Transformer应用于图像分类任务,但为了满足Transformer的序列化和并行化要求,需要将图像分割成大小相等、互不重叠的图像块,破坏了相邻图像数据块之间的局部信息。此外,由于Transformer具有较少的先验知识,模型往往需要在大规模数据集上进行预训练,因此计算复杂度较高。为了同时建模图像相邻块之间的局部信息并充分利用图像的全局信息,提出了一种基于Depth-wise卷积的视觉Transformer(Efficient Pyramid Vision Transformer,EPVT)模型。EPVT模型可以实现以较低的计算成本提取相邻图像块之间的局部和全局信息。EPVT模型主要包含3个关键组件:局部感知模块(Local Perceptron Module,LPM)、空间信息融合模块(Spatial Information Fusion,SIF)和“+卷积前馈神经网络(Convolution Feed-forward Network,CFFN)。LPM模块用于捕获图像的局部相关性;SIF模块用于融合相邻图像块之间的局部信息,并利用不同图像块之间的远距离依赖关系,提升模型的特征表达能力,使模型学习到输出特征在不同维度下的语义信息;CFFN模块用于编码位置信息和重塑张量。在图像分类数据集ImageNet-1K上,所提模型优于现有的同等规模的视觉Transformer分类模型,取得了82.6%的分类准确度,证明了该模型在大规模数据集上具有竞争力。

基于改进FeatDepth的足球运动场景无监督单目图像深度预测

为了在降低成本的同时提高图像深度信息预测的精确度,并将深度估计应用于足球运动场景,提出一种基于改进FeatDepth的足球运动场景无监督单目图像深度预测方法。首先,对原FeatDepth引入注意力机制,使模型更加关注有效的特征信息;其次,将FeatDepth中的PoseNet网络和DepthNet网络分别嵌入GAM全局注意力机制模块,为网络添加额外的上下文信息,在基本不增加计算成本的情况下提升FeatDepth模型深度预测性能;再次,为在低纹理区域和细节上获得更好的深度预测效果,由单视图重构损失与交叉视图重构损失组合而成最终的损失函数。选取KITTI数据集中Person场景较多的部分进行数据集制作并进行仿真实验,结果表明,改进后的FeatDepth模型不仅在精确度上有所提升,且在低纹理区域及细节处拥有更好的深度预测效果。最后,对比模型在足球场景下的推理效果后得出,改进后的模型在低纹理区域(足球、球门等)及细节处(肢体等)有更好的深度预测效果,实现了将基于无监督的单目深度估计模型应用于足球运动场景的目的。

3D depth image analysis for indoor fall detection of elderly people

This paper presents a new fall detection method of etderly people in a room environment based on shape analysis of 3D depth images captured by a Kinect sensor. Depth images are pre- processed by a median filter both for background and target. The sithouette of moving individual in depth images is achieved by a subtraction method for background frames. The depth images are converted to disparity map, which is obtained by the horizontal and vertical projection histogram statistics. The initial floor plane information is obtained by V disparity map, and the floor ptane equation is estimated by the least square method. Shape information of human subject in depth images is analyzed by a set of moment functions. Coefficients of ellipses are calculated to determine the direction of individual The centroids of the human body are catculated and the angle between the human body and the floor plane is calculated. When both the distance from the centroids of the human body to the floor plane and the angle between the human body and the floor plane are tower than some threshotds, fall incident will be detected. Experiments with different failing direction are performed. Experimental results show that the proposed method can detect fall incidents effectively.

Optical coherence tomography enhanced depth imaging of chorioretinal folds in patients with orbital tumors

AIM:To characterize spectral-domain optical coherence tomography(SD-OCT)features of chorioretinal folds in orbital mass imaged using enhanced depth imaging(EDI).METHODS:Prospective observational case-control study was conducted in 20 eyes of 20 patients,the uninvolved eye served as a control.All the patients underwent clinical fundus photography,computed tomography,EDI SDOCT imaging before and after surgery.Two patients with cavernous hemangiomas underwent intratumoral injection of bleomycin A5;the remaining patients underwent tumor excision.Patients were followed 1 to 14mo following surgery(average follow up,5.8mo).RESULTS:Visual acuity prior to surgery ranged from 20/20 to 20/200.Following surgery,5 patients’visual acuity remained unchanged while the remaining 15 patients had a mean letter improvement of 10(range 4 to 26 letters).Photoreceptor inner/outer segment defects were found in 10 of 15 patients prior to surgery.Following surgical excision,photoreceptor inner/outer segment defects fully resolved in 8 of these 10 patients.CONCLUSION:Persistence of photoreceptor inner/outer segment defects caused by compression of the globe by an orbital mass can be associated with reduced visual prognosis.Our findings suggest that photoreceptor inner/outer segment defects on EDI SD-OCT could be an indicator for immediate surgical excision of an orbital mass causing choroidal compression.

Technique of Embedding Depth Maps into 2D Images

This paper proposes a new technique that is used to embed depth maps into corresponding 2-dimensional （2D） images. Since a 2D image and its depth map are integrated into one type of image format, they can be treated as if they were one 2D image. Thereby, it can reduce the amount of data in 3D images by half and simplify the processes for sending them through networks because the synchronization between images for the left and right eyes becomes unnecessary. We embed depth maps in the quantized discrete cosine transform （DCT） data of 2D images. The key to this technique is whether the depth maps could be embedded into 2D images without perceivably deteriorating their quality. We try to reduce their deterioration by compressing the depth map data by using the differences from the next pixel to the left. We assume that there is only one non-zero pixel at most on one horizontal line in the DCT block because the depth map values change abruptly. We conduct an experiment to evaluate the quality of the 2D images embedded with depth maps and find that satisfactory quality could be achieved.

Analysis of Object Depth Effects on Accuracy of Dimensional Shape in X and Y Directions Using Single Non-metric Image

In general, to reconstruct the accurate shape of buildings, we need at least one stereomodel （two photographs） for each building. In most cases, however, only a single non-metric photograph is available, which is usually obtained either by an amateur, such as a tourist, or from a newspaper or a post card. To evaluate the validity of 3D reconstruction from a single non-metric image, this study analyzes the effects of object depth on the accuracy of dimensional shape in X and Y directions using a single non-metric image by means of simulation technique, as this was considered to be, in most cases, a main source of data acquisition in recording and documenting buildings.

A method to generate foggy optical images based on unsupervised depth estimation

For traffic object detection in foggy environment based on convolutional neural network(CNN),data sets in fog-free environment are generally used to train the network directly.As a result,the network cannot learn the object characteristics in the foggy environment in the training set,and the detection effect is not good.To improve the traffic object detection in foggy environment,we propose a method of generating foggy images on fog-free images from the perspective of data set construction.First,taking the KITTI objection detection data set as an original fog-free image,we generate the depth image of the original image by using improved Monodepth unsupervised depth estimation method.Then,a geometric prior depth template is constructed to fuse the image entropy taken as weight with the depth image.After that,a foggy image is acquired from the depth image based on the atmospheric scattering model.Finally,we take two typical object-detection frameworks,that is,the two-stage object-detection Fster region-based convolutional neural network(Faster-RCNN)and the one-stage object-detection network YOLOv4,to train the original data set,the foggy data set and the mixed data set,respectively.According to the test results on RESIDE-RTTS data set in the outdoor natural foggy environment,the model under the training on the mixed data set shows the best effect.The mean average precision(mAP)values are increased by 5.6%and by 5.0%under the YOLOv4 model and the Faster-RCNN network,respectively.It is proved that the proposed method can effectively improve object identification ability foggy environment.

Case Depth Measurement of Induction Hardening Using Image Processing

Case depth measurement of the induction hardened steel parts is necessary for quality control. Vickers microhardness test is the most industrially accepted method to identify the case depth. But this method is a time consuming one and it requires expensive equipment. The aim of this study is to develop a different method to determine the case depth using image processing. The surface hardened steel samples were cross cut, ground and etched with Nital. The etched macrosectioned specimens were scanned by a scanner. The scanned images were evaluated by the developed software. The principle of the software is to identify the gray level difference. The effective case depths of the surface hardened specimens obtained by Vickers microhardness test and the developed method were compared. It was found that the deviation of the developed method was ±0.12 mm at the case depth range of 0.6 - 2.0 mm and mm at the case depth range of 2.1 - 4.3 mm. The measuring time was only 20% of Vickers microhardness test. The deviation range is much lower than the tolerance case depth specification for induction hardening in general.

Simulation of Structural Characteristics and Depth Filtration Elements in Interconnected Nanofibrous Membrane Based on Adaptive Image Analysis

Due to their unique structural features, electrospun membranes have gained considerable attention for use in applications where quality of depth filtration is a dominant performance factor. To elucidate the depth filtration phenomena it is important to quantify the intrinsic structural properties independent from the dynamics of transport media. Several methods have been proposed for structural characterization of such membranes. However, these methods do not meet the requirement for the quantification of intrinsic structural properties in depth filtration. This may be due to the complex influence of transport media dynamics and structural elements in the depth filtration process. In addition, the different morphological architectures of electrospun membranes present obstacles to precise quantification. This paper seeks to quantify the structural characteristics of electrospun membranes by introducing a robust image analysis technique and exploiting it to evaluate the permeation-filtration mechanism. To this end, a nanostructured fibrous network was simulated as an ideal membrane using adaptive local criteria in the image analysis. The reliability of the proposed approach was validated with measurements and comparison of structural characteristics in different morphological conditions. The results were found to be well compatible with empirical observations of perfect membrane structures. This approach, based on optimization of electrospinning parameters, may pave the way for producing optimal membrane structures for boosting the performance of electrospun membranes in end-use applications.

Image-Pro Plus技术在沟槽滤棒特征参数测量中的应用

为准确描述并测量沟槽滤棒的特征参数,采用Image-Pro Plus(IPP)技术建立了一种沟槽滤棒特征参数测量系统。将5组已知沟槽面积比的沟槽滤棒标准件和10支裁切后的自制沟槽滤棒放置在显微镜载物台上,沟槽端面朝上,调整光圈亮度及显微镜焦距,按编号顺序拍摄成像,并使用IPP软件计算沟槽特征参数,运用MINITAB软件分析IPP测量系统的准确性和精密度。结果表明:1IPP测量系统操作方便,测量过程简单,能快速测量沟槽滤棒的沟槽面积比、深度、数目等特征参数。2系统对沟槽面积比、沟槽深度的测量结果不存在线性偏倚。沟槽面积比精密度指标Gage R&R=24.66%<30%,沟槽深度精密度指标Gage R&R=26.40%<30%,测量数据精密度好。3系统手动和自动两种测量模式对测量结果无显著差异(P>0.05),可以根据不同需求选择测量模式。该测量方法为研究沟槽滤棒降焦机理和提高生产过程控制能力提供了技术支持。

Removing fog from traffic image sequence

Aiming at removing fog from traffic images, a distance field is built according to the characteristics of traffic images, and a novel parameter estimation method based on the traffic image sequence is proposed. The fog model is derived from atmospheric scattering models. The direction of the distance field is parallel to the center line of the road, which increases along a line from the observer to the horizon, and the normalization is carried out to improve the distribution of the distance field model. After parameter initialization, the variations of the average gray values of reference regions are taken as the determining conditions to adjust the parameters. Finally, restorations are made by the fog model. Experimental results show that the proposed method can effectively remove fog from traffic images.

SEM-backscattered Electron Imaging and Image Processing for Evaluation of Unhydrated Cement Volume Fraction in Slag Blended Portland Cement Pastes

Proper parameters for image taking and minimum field number for image processing were investigated to evaluate volume fraction of unhydrated cement（UHC） in both neat cement paste and slag blended cement paste. Our research suggested that magnification 250x was sufficient for the two pastes, and accelerating voltage should be set as 15 kV and 20 kV for BSE image taking of neat cement paste and slag blended cement paste respectively; the minimum field number increased while the total imaging area stayed the same as the magnification increased within certain statistical bias.

An Image Classification Method Based on Deep Neural Network with Energy Model

The development of deep learning has revolutionized image recognition technology.How to design faster and more accurate image classification algorithms has become our research interests.In this paper,we propose a new algorithm called stochastic depth networks with deep energy model(SADIE),and the model improves stochastic depth neural network with deep energy model to provide attributes of images and analysis their characteristics.First,the Bernoulli distribution probability is used to select the current layer of the neural network to prevent gradient dispersion during training.Then in the backpropagation process,the energy function is designed to optimize the target loss function of the neural network.We also explored the possibility of using Adam and SGD combination optimization in deep neural networks.Finally,we use training data to train our network based on deep energy model and testing data to verify the performance of the model.The results we finally obtained in this research include the Classified labels of images.The impacts of our obtained results show that our model has high accuracy and performance.

Improving sampling depth of laser speckle imaging by topical optical clearing:A theoretical and in vivo study

The effect of optical cleaning method combined with laser speckle imaging(LSI)was discussed to improve the detection depth of LSI due to high scattering characteristics of skin,which limit its clinical application.A double-layer skin tissue model embedded with a single blood vessel was established,and the Monte Carlo method was used to simulate photon propagation under the action of light-permeating agent.808 nm semiconductor and 632.8 nm He–Ne lasers were selected to study the e®ect of optical clearing agents(OCAs)on photon deposition in tissues.Results show that the photon energy deposition density in the epidermis increases with the amount of tissue°uid replaced by OCA.Compared with glucose solution,polyethylene glycol 400(PEG 400)and glycerol can considerably increase the average penetration depth of photons in the skin tissue,thereby raising the sampling depth of the LSI.After the action of glycerol,PEG 400,and glucose,the average photon penetration depth is increased by 51.78%,51.06%,and 21.51%for 808nm,68.93%,67.94%,and 26.67%for 632.8 nm lasers,respectively.In vivo experiment by dorsal skin chamber proves that glycerol can cause a substantial decrease in blood°ow rate,whereas PEG 400 can signicantly improve the capability of light penetration without a®ecting blood velocity,which exhibits considerable potential in the monitoring of blood°ow in skin tissues.

Influence of limited-view scanning on depth imaging of photoacoustic tomography

We study the influence of limited-view scanning on the depth imaging of photoacoustic tomography. The situation, in which absorbers are located at different depths with respect to the limited-view scanning trajectory, is called depth imaging and is investigated in this paper. The results show that limited-view scanning causes the reconstructed intensity of deep absorbers to be weaker than that of shallow ones and that deep absorbers will be invisible if the scanning range is too small. The concept of effective scanning angle is proposed to analyse that phenomenon. We find that an effective scanning angle can well predict the relationship between scanning angle and the intensity ratio of absorbers. In addition, limited-view scanning is employed to improve image quality.

Optical coherence tomography with or without enhanced depth imaging for peripapillary retinal nerve fiber layer and choroidal thickness

AIM：To assess peripapillary retinal nerve fiber layer（RNFL）and choroidal thickness obtained with enhanced depth imaging（EDI）mode compared with those obtained without EDI mode using Heidelberg Spectralis optical coherence tomography（OCT）.METHODS：Fifty eyes of 25 normal healthy subjects and32 eyes of 20 patients with different eye diseases were included in the study.All subjects underwent 3.4 mm diameter peripapillary circular OCT scan centered on the optic disc using both the conventional and the EDI OCT protocols.The visualization of RNFL and choroidoscleral junction was assessed using an ordinal scoring scale.The paired t-test,intraclass correlation coefficient（ICC）,95%limits of agreement（LoA）,and Bland and Altman plots were used to test the agreement of measurements.RESULTS：The visibility score of RNFL obtained with and without EDI was of no significant difference（P=0.532）,the visualization of choroidoscleral junction was better using EDI protocol than conventional protocol（P〈0.001）.Peripapillary RNFL thickness obtained with EDI was slightly thicker than that obtained without EDI（103.25±9.42μm vs 101.87±8.78μm,P=0.010）.The ICC of the two protocols was excellent with the value of 0.867 to 0.924,the 95%LoA of global RNFL thickness was between-10.0 to 7.4μm.Peripapillary choroidal thickness obtained with EDI was slightly thinner than that obtained without EDI（147.23±51.04μm vs 150.90±51.84μm,P〈0.001）.The ICC was also excellent with the value of 0.960 to 0.987,the 95%LoA of global choroidal thickness was between-12.5 to 19.8μm.CONCLUSION：Peripapillary circular OCT scan with or without EDI mode shows comparable results in the measurement of peripapillary RNFL and choroidal thickness.