岩石薄片图像的分析往往依赖于专业人员在显微镜下观察并给出鉴定结果,不但费时费力,并且受设备影响较大。近些年,针对薄片图像的自动识别方法已经被提出。然而,这些方法大多采用监督学习与深度学习相结合的方式,由于需要大量人工标注...岩石薄片图像的分析往往依赖于专业人员在显微镜下观察并给出鉴定结果,不但费时费力,并且受设备影响较大。近些年,针对薄片图像的自动识别方法已经被提出。然而,这些方法大多采用监督学习与深度学习相结合的方式,由于需要大量人工标注而受到限制,为方法的推广与应用带来巨大困难。此外,模型在不同的地层、岩性等目标应用时,由于不同地质环境中岩石的差异性,其泛化性也受到极大限制。本文针对该问题提出了一种简单线性迭代聚类算法(simple linear iterative cluster,SLIC)与半监督自训练结合的方法,仅依靠6%的人工标注便能够实现岩石图像的自动化分割与组分识别,极大地增强岩石图像自动识别方法在实际应用中的价值。该方法首先使用超像素算法SLIC对岩石图像进行预分割,随后基于分割片的颜色特征进行粗合并,并根据最小外接矩形进行切割;切割下来的岩石组分分割图像作为后续处理的基础数据集,这里仅需要人工标注6%的岩石组分数据;随后,这些数据通过一个改进的半监督自训练方法,以改进的VGG16模型作为主模型、ResNet18模型作为评判模型,不断生成高置信度的伪标签,利用迭代优化调整,将其扩展到整个数据集,最终获得一个具有较高的稳定性、准确性及一致性的组分识别模型。实际数据的测试与分析表明,本文所提出SLIC和半监督自训练结合的方法,对6类岩石组分的识别准确率可达到96%。该方法能够在数据差异不大的条件下,帮助用户基本实现自动化的组分识别。而当数据集产生较大差异时,仅需标注小部分样品即可实现自动组分识别。本方法具有较高的泛化性和可靠性,能够在实际应用提供足够的准确性与便利性。展开更多
In this paper, a new trust region algorithm for unconstrained LC1 optimization problems is given. Compare with those existing trust regiion methods, this algorithm has a different feature: it obtains a stepsize at eac...In this paper, a new trust region algorithm for unconstrained LC1 optimization problems is given. Compare with those existing trust regiion methods, this algorithm has a different feature: it obtains a stepsize at each iteration not by soloving a quadratic subproblem with a trust region bound, but by solving a system of linear equations. Thus it reduces computational complexity and improves computation efficiency. It is proven that this algorithm is globally convergent and locally superlinear under some conditions.展开更多
In asynchronous Multiple-Input-Multiple-Output Orthogonal Frequency Division Multiplexing(MIMO-OFDM) over the selective Rayleigh fading channel,the performance of the existing linear detection algorithms improves slow...In asynchronous Multiple-Input-Multiple-Output Orthogonal Frequency Division Multiplexing(MIMO-OFDM) over the selective Rayleigh fading channel,the performance of the existing linear detection algorithms improves slowly as the Signal Noise Ratio (SNR) increases.To improve the performance of asynchronous MIMO-OFDM,a low complexity iterative detection algorithm based on linear precoding is proposed in this paper.At the transmitter,the transmitted signals are spread by precoding matrix to achieve the space-frequency diversity gain,and low complexity iterative Interference Cancellation(IC) algorithm is used at the receiver,which relieves the error propagation by the precoding matrix.The performance improvement is verified by simulations.Under the condition of 4 transmitting antennas and 4 receiving antennas at the BER of 10-4,about 6 dB gain is obtained by using our proposed algorithm compared with traditional algorithm.展开更多
文摘岩石薄片图像的分析往往依赖于专业人员在显微镜下观察并给出鉴定结果,不但费时费力,并且受设备影响较大。近些年,针对薄片图像的自动识别方法已经被提出。然而,这些方法大多采用监督学习与深度学习相结合的方式,由于需要大量人工标注而受到限制,为方法的推广与应用带来巨大困难。此外,模型在不同的地层、岩性等目标应用时,由于不同地质环境中岩石的差异性,其泛化性也受到极大限制。本文针对该问题提出了一种简单线性迭代聚类算法(simple linear iterative cluster,SLIC)与半监督自训练结合的方法,仅依靠6%的人工标注便能够实现岩石图像的自动化分割与组分识别,极大地增强岩石图像自动识别方法在实际应用中的价值。该方法首先使用超像素算法SLIC对岩石图像进行预分割,随后基于分割片的颜色特征进行粗合并,并根据最小外接矩形进行切割;切割下来的岩石组分分割图像作为后续处理的基础数据集,这里仅需要人工标注6%的岩石组分数据;随后,这些数据通过一个改进的半监督自训练方法,以改进的VGG16模型作为主模型、ResNet18模型作为评判模型,不断生成高置信度的伪标签,利用迭代优化调整,将其扩展到整个数据集,最终获得一个具有较高的稳定性、准确性及一致性的组分识别模型。实际数据的测试与分析表明,本文所提出SLIC和半监督自训练结合的方法,对6类岩石组分的识别准确率可达到96%。该方法能够在数据差异不大的条件下,帮助用户基本实现自动化的组分识别。而当数据集产生较大差异时,仅需标注小部分样品即可实现自动组分识别。本方法具有较高的泛化性和可靠性,能够在实际应用提供足够的准确性与便利性。
文摘In this paper, a new trust region algorithm for unconstrained LC1 optimization problems is given. Compare with those existing trust regiion methods, this algorithm has a different feature: it obtains a stepsize at each iteration not by soloving a quadratic subproblem with a trust region bound, but by solving a system of linear equations. Thus it reduces computational complexity and improves computation efficiency. It is proven that this algorithm is globally convergent and locally superlinear under some conditions.
基金supported by the Hi-Tech Research and Development Program of China under Grant No.2009AA01Z236the National Natural Science Foundation of China under Grants No.60902027,No.60832007 and No.60901018+1 种基金the Funds under Grant No.9140A21030209DZ02the Fundamental Research Funds for the Central Universities under Grants No.ZYGX2009J008,No.ZYGX2009J010
文摘In asynchronous Multiple-Input-Multiple-Output Orthogonal Frequency Division Multiplexing(MIMO-OFDM) over the selective Rayleigh fading channel,the performance of the existing linear detection algorithms improves slowly as the Signal Noise Ratio (SNR) increases.To improve the performance of asynchronous MIMO-OFDM,a low complexity iterative detection algorithm based on linear precoding is proposed in this paper.At the transmitter,the transmitted signals are spread by precoding matrix to achieve the space-frequency diversity gain,and low complexity iterative Interference Cancellation(IC) algorithm is used at the receiver,which relieves the error propagation by the precoding matrix.The performance improvement is verified by simulations.Under the condition of 4 transmitting antennas and 4 receiving antennas at the BER of 10-4,about 6 dB gain is obtained by using our proposed algorithm compared with traditional algorithm.
