基坑安全管理是大型建筑基坑施工的关键内容,基坑结构位移预测是预防基坑支护事故的重要手段.但是由于基坑局部基坑位移成因复杂,现有的支持向量回归(SVR)、随机森林(RF)方法忽略了基坑位移随空间位移局部减弱、随时间局部位移加快增长的特点,导致预测精度不高.因此,本文提出一种融合时空注意力机制的GA-BP神经网络(A-GA-BP)方法,通过时空特征准确表示基坑位移预测的时空维度及其特征相关性,提高基坑位移预测的有效性.最后,本文以苏州市某大型工程为实例,对基坑的水平与垂直位移监测数据进行模型训练与评估,按时域特征、空域特征、多阶时域空域特征进行量化分析与研究,并与现有方法进行比较.实验结果表明,本文方法的拟合指数比其他几种方法分别提高29.19％与41.25％,多阶时空域特征相较于单独的时间域或空间域特征分别提高3.08％与1.83％.

Path planning research can effectively solve the problem of finding free parking space in multi-storey parking lots. This paper takes advantage of the decision ability of reinforcement learning and the perception ability of deep learning to improve the algorithm based on traditional DQN. On the one hand, Q value is updated with qualification trace; On the other hand, different loss functions are set for the main network and the target network, and the two are combined to improve the accuracy of path planning. The experimental results show that the improved DQN model can accomplish the path planning task of multi-storey parking lot more accurately and efficiently.

In the field of medical image processing today, there are more and more medical image categories, such as cell images, tissue images, etc. The wide variety of images is of great help in medical diagnosis, not only for visual observation but also for precise analysis of various causes of disease. Due to the development of medicine, the requirements for images are also higher and the amount of data is becoming larger, and the images have reached tens of thousands of pixels, for the current computer, the current environment can no longer meet the needs of image loading display. In response to the above problems, this paper proposes a method for storing and displaying oversized medical images based on centralized points of interest, which achieves fast loading and displaying of oversized cell images, and has been practically applied in relevant medical institutions, achieving certain results in compressed storage and real-time display of cell images, showing the effectiveness and advancement of the method.