An adaptive segmentation method combining MSRCR and mean shift algorithm with K-means correction of green apples in natural environment

During the recognition and localization process of green apple targets,problems such as uneven illumination,occlusion of branches and leaves need to be solved.In this study,the multi-scale Retinex with color restoration(MSRCR)algorithm was applied to enhance the original green apple images captured in an orchard environment,aiming to minimize the impacts of varying light conditions.The enhanced images were then explicitly segmented using the mean shift algorithm,leading to a consistent gray value of the internal pixels in an independent fruit.After that,the fuzzy attention based on information maximization algorithm(FAIM)was developed to detect the incomplete growth position and realize threshold segmentation.Finally,the poorly segmented images were corrected using the K-means algorithm according to the shape,color and texture features.The users intuitively acquire the minimum enclosing rectangle localization results on a PC.A total of 500 green apple images were tested in this study.Compared with the manifold ranking algorithm,the K-means clustering algorithm and the traditional mean shift algorithm,the segmentation accuracy of the proposed method was 86.67%,which was 13.32%,19.82%and 9.23%higher than that of the other three algorithms,respectively.Additionally,the false positive and false negative errors were 0.58%and 11.64%,respectively,which were all lower than the other three compared algorithms.The proposed method accurately recognized the green apples under complex illumination conditions and growth environments.Additionally,it provided effective references for intelligent growth monitoring and yield estimation of fruits.

IMAGE SEGMENTATION BASED ON MARKOV RANDOM FIELD AND WATERSHED TECHNIQUES

This paper presented a method that incorporates Markov Random Field(MRF), watershed segmentation and merging techniques for performing image segmentation and edge detection tasks. MRF is used to obtain an initial estimate of x regions in the image under process where in MRF model, gray level x , at pixel location i , in an image X , depends on the gray levels of neighboring pixels. The process needs an initial segmented result. An initial segmentation is got based on K means clustering technique and the minimum distance, then the region process in modeled by MRF to obtain an image contains different intensity regions. Starting from this we calculate the gradient values of that image and then employ a watershed technique. When using MRF method it obtains an image that has different intensity regions and has all the edge and region information, then it improves the segmentation result by superimpose closed and an accurate boundary of each region using watershed algorithm. After all pixels of the segmented regions have been processed, a map of primitive region with edges is generated. Finally, a merge process based on averaged mean values is employed. The final segmentation and edge detection result is one closed boundary per actual region in the image.

融合多语义特征的精读式文档级事件抽取

为解决文档级事件抽取任务依赖实体识别、忽略先验语义和参数分散的问题,提出一种融合多语义特征的精读式抽取方法。结合“三阶段”阅读特点,根据事件与角色交互、角色类型及释义特征构建外部语义模板,提出窗口切分算法切割文档语义;基于预训练模型BERT融合外部与窗口语义;多轮精读文档避免实体依赖,设计记忆网络对精读结果建模,完成跨句定位参数和事件路径扩展。引入噪声扰动防止模型过拟合。实验结果表明,该模型性能优于当前主流方法,验证了其可行性和有效性。

Segmented thermoelectric generator modelling and optimization using artificial neural networks by iterative training

Renewable energy technologies are central to emissions reduction and essential to achieve net-zero emission.Segmented thermoelectric generators(STEG)facilitate more efficient thermal energy recovery over a large temperature gradient.However,the additional design complexity has introduced challenges in the modelling and optimization of its performance.In this work,an artificial neural network(ANN)has been applied to build accurate and fast forward modelling of the STEG.More importantly,we adopt an iterative method in the ANN training process to improve accuracy without increasing the dataset size.This approach strengthens the proportion of the high-power performance in the STEG training dataset.Without increasing the size of the training dataset,the relative prediction error over high-power STEG designs decreases from 0.06 to 0.02,representing a threefold improvement.Coupling with a genetic algorithm,the trained artificial neural networks can perform design optimization within 10 s for each operating condition.It is over 5,000 times faster than the optimization performed by the conventional finite element method.Such an accurate and fast modeller also allows mapping of the STEG power against different parameters.The modelling approach demonstrated in this work indicates its future application in designing and optimizing complex energy harvesting technologies.

一种基于改进AOD-Net的航拍图像去雾算法

针对航拍图像易受雾气影响,AOD-Net(All in one dehazing network)算法对图像去雾后容易出现细节模糊、对比度过高和图像偏暗等问题,本文提出了一种基于改进AOD-Net的航拍图像去雾算法.本文主要从网络结构、损失函数、训练方式三个方面对AOD-Net进行改良.首先在AOD-Net的第二个特征融合层上添加了第一层的特征图,用全逐点卷积替换了传统卷积方式,并用多尺度结构提升了网络对细节的处理能力.然后用包含有图像重构损失函数、SSIM(Structural similarity)损失函数以及TV(Total variation)损失函数的复合损失函数优化去雾图的对比度、亮度以及色彩饱和度.最后采用分段式的训练方式进一步提升了去雾图的质量.实验结果表明,经该算法去雾后的图像拥有令人满意的去雾结果,图像的饱和度和对比度相较于AOD-Net更自然.与其他对比算法相比,该算法在合成图像实验、真实航拍图像实验以及算法耗时测试的综合表现上更好,更适用于航拍图像实时去雾.

基于LabVIEW和IMAQ的动车车号自动识别

随着人工智能的快速发展,机器视觉系统所具有的数字化、智能化、实时化等特点逐渐被应用到实际项目中.此系统以实现对过往动车车号识别为研究目的,采用NI公司的LabVIEW软件作为开发平台,调用IMAQ专业控件及Vision Assistant函数库,设计一款智能化,并同时包括图像采集、图像预处理、车号定位与字符识别等完整的动车车号识别系统.此系统包含背景校正算法、基本形态学算法以及字符间距分割算法等,且动车车号识别率高达99.81%,解决实际应用问题,为进一步研究奠定基础.

基于遗传优化支持向量机的开关磁阻电机非线性建模方法

将有限元法(FEM)与非线性映射技术相结合,得到了开关磁阻电机(SRM)动态仿真模型。利用FEM获取了SRM的磁化特性和转矩特性数据,并依此对支持向量机进行了训练,进而在MATLAB中建立了仿真模型。采用改进型遗传算法对支持向量机的超参数进行全局寻优,提高了其逼近和泛化能力。基于对磁化特性数据的分析,引入了分段训练的思想,进一步提高了模型在小电流下的仿真精度。将所建模型的动态仿真结果与FEM分析结果相比较,验证了建模方法的有效性。

人工智能深度学习的激光图像分割研究

激光成像受到环境、设备自身等干扰,使得激光图像含有噪声,当前图像分割方法对噪声干扰鲁棒性差,误分割现象出现概率高,重要信息丢失严重,为了克服当前激光图像分割的弊端,提出了基于人工智能深度学习的激光图像分割方法。首先采用小波变换对激光图像进行特征提取,并对噪声干扰进行抑制处理,然后引入人工智能学习算法对激光图像特征向量进行训练,并根据训练结果对激光图像像素点进行分类,从而实现激光图像分割,最后采用含噪和不含噪的激光图像进行仿真测试。结果表明,对含噪和不含噪的激光图像,人工智能深度学习的分割精度分别达到了91%和95%以上,精度明显高于经典激光图像分割方法,分割效率可以满足激光图像向大规模方向发展的要求。

A deep semantic segmentation-based algorithm to segment crops and weeds in agronomic color images

In precision agriculture,the accurate segmentation of crops and weeds in agronomic images has always been the center of attention.Many methods have been proposed but still the clean and sharp segmentation of crops and weeds is a challenging issue for the images with a high presence of weeds.This work proposes a segmentation method based on the combination of semantic segmentation and K-means algorithms for the segmenta-tion of crops and weeds in color images.Agronomic images of two different databases were used for the segmentation algorithms.Using the thresholding technique,everything except plants was removed from the images.Afterward,semantic segmentation was applied using U-net followed by the segmentation of crops and weeds using the K-means subtractive algorithm.The comparison of segmentation performance was made for the proposed method and K-Means clustering and superpixels algorithms.The proposed algorithm pro-vided more accurate segmentation in comparison to other methods with the maximum accuracy of equivalent to 99.19%.Based on the confusion matrix,the true-positive and true-negative values were 0.9952 and 0.8985 representing the true classification rate of crops and weeds,respectively.The results indicated that the proposed method successfully provided accurate and convincing results for the segmentation of crops and weeds in the images with a complex presence of weeds.

包围圆分割在铁路货车车号字符的应用

针对铁路货车车号的断裂、倾斜、变形等特点,传统字符分割方法分割精度低的问题,提出一种基于改进包围圆的分割方法。基于铁路货车单行、双行2种排列方式,采用自适应游程算法进行双行车号的分割,鉴于游程算法背景像素前景化的处理特点,预先采用游程算法进行断裂消除,再使用包围圆方法进行字符分割。实验结果表明:在图像质量不高情况下,可以实现良好的分割精度且在分割准确率和鲁棒性方面均优于传统算法。

改进的K均值分割算法在关键词检测中的应用

鉴于K均值分割算法中隐马尔可夫模型 (HMM)参数重估公式简单、实用 ,目前大多数基于HMM的关键词检测系统都采用此算法训练参考模型。为了提高参考模型的有效性和解决该算法在具体实现时所遇到的问题 ,本文提出了改进的K均值分割 (MSKM)算法。MSKM算法以关键词检测系统的检出率为模板收敛的判决依据 ,使HMM参数调整从一定程度上而言是以检测系统性能为目标函数 ;同时引入了基于HMM的聚类方法 ,使聚类和参数估计融为一体。实验结果表明 ,采用MSKM算法比原算法可使关键词检测系统的平均检出率提高 1 8%。

IMAGE ANALYSIS BASED ON EDGE DETECTION TECHNIQUES

A method that incorporates edge detection technique, Markov Random field (MRF), watershed segmentation and merging techniques was presented for performing image segmentation and edge detection tasks. It first applies edge detection technique to obtain a Difference In Strength (DIS) map. An initial segmented result is obtained based on K means clustering technique and the minimum distance. Then the region process is modeled by MRF to obtain an image that contains different intensity regions. The gradient values are calculated and then the watershed technique is used. DIS calculation is used for each pixel to define all the edges (weak or strong) in the image. The DIS map is obtained. This help as priority knowledge to know the possibility of the region segmentation by the next step (MRF), which gives an image that has all the edges and regions information. In MRF model, gray level l , at pixel location i , in an image X , depends on the gray levels of neighboring pixels. The segmentation results are improved by using watershed algorithm. After all pixels of the segmented regions are processed, a map of primitive region with edges is generated. The edge map is obtained using a merge process based on averaged intensity mean values. A common edge detectors that work on (MRF) segmented image are used and the results are compared. The segmentation and edge detection result is one closed boundary per actual region in the image.

Optimal Energy-Efficient Operation of a Metro Train on a Long and Steep Downhill Segment

Because the standard four-stage operation and con-trol strategy cannot fully utilize the gravitational potential energy of a train operating on a long and steep downhill segment,this paper further improves the method for train operation and control strategy.The improved operation includes three stages of acceleration,coasting-speed limit cruising,and brak-ing.Taking the speed limit,time limit,and distance limit as the constraints,the coasting condition switching point,braking condition switching point,traction coefficient,and braking force coefficient are used as the decision variables.Then,an improved train traction energy consumption model is constructed,and an improved differential evolution algorithm is designed to solve this model.The improved method is used to simulate two long and steep downhill segments of the Nanning metro.The results show that the improved method can meet the requirements of speed limit,time limit,and distance limit.Compared with the standard four-stage operation,the improved train operation and control strategy can reduce train energy consumption by more than 40%on the two long and steep downhill segments;compared with other similar algorithms,the improved algorithm is more suitable for solving the model.

Image Segmentation via Fischer-Burmeister Total Variation and Thresholding

Image segmentation is a significant problem in image processing.In this paper,we propose a new two-stage scheme for segmentation based on the Fischer-Burmeister total variation(FBTV).The first stage of our method is to calculate a smooth solution from the FBTV Mumford-Shah model.Furthermore,we design a new difference of convex algorithm(DCA)with the semi-proximal alternating direction method of multipliers(sPADMM)iteration.In the second stage,we make use of the smooth solution and the K-means method to obtain the segmentation result.To simulate images more accurately,a useful operator is introduced,which enables the proposed model to segment not only the noisy or blurry images but the images with missing pixels well.Experiments demonstrate the proposed method produces more preferable results comparing with some state-of-the-art methods,especially on the images with missing pixels.

基于协同训练的指纹图像分割算法

指纹图像分割是自动指纹识别系统预处理中最关键的技术之一.精确、可靠地将指纹图像从背景中分割出来,能够加快后续工作的处理速度,提高识别算法的准确性.传统的分割算法需要大量已标记的指纹图像作为训练数据,但实际应用中获取标记样本比较繁琐和耗时.为综合利用已标记和未标记的指纹图像,提出一种基于协同训练的半监督指纹图像分割算法:CoSeg.该算法在基于像素水平的Coherence、Mean、Variace(CMV)特征体系下,使用标记盒和支持向量机作为基分类器进行协同训练.在FVC2002指纹库上的实验结果表明,CoSeg能够在标记信息较少的情况下取得较好的性能,并在处理低质量指纹图像时表现出较强的鲁棒性.

Multi-layer collaborative optimization fusion for semi-supervised learning

Recently,the Cooperative Training Algorithm(CTA),a well-known Semi-Supervised Learning(SSL)technique,has garnered significant attention in the field of image classification.However,traditional CTA approaches face challenges such as high computational complexity and low classification accuracy.To overcome these limitations,we present a novel approach called Weighted fusion based Cooperative Training Algorithm(W-CTA),which leverages the cooperative training technique and unlabeled data to enhance classification performance.Moreover,we introduce the K-means Cooperative Training Algorithm(km-CTA)to prevent the occurrence of local optima during the training phase.Finally,we conduct various experiments to verify the performance of the proposed methods.Experimental results show that W-CTA and km-CTA are effective and efficient on CIFAR-10 dataset.