In order to solve the problems of large training parameters and low text recognition rate of convolutional recurrent neural networks (CRNN) handwritten Chinese character recognition network model, a novel method for handwritten Chinese character recognition based on attention bi-directional long short-term memory network(AT-BLSTM) and knowledge distillation (KD) technology was proposed. By assigning different weights to the input vector features of AT-BLSTM network, the model training data set was more efficient and accurate. Through KD technology, the knowledge acquired from a large high-performance model was transferred to a small model, which ensured the accuracy of the model, reduced the training parameters and internal storage ratio, and obtained a lightweight training model with better performance. Through the comparison of multiple groups of experiments, the accuracy of Chinese character recognition is increased by 6.7%, and the training parameters are reduced by 15.94 M. The recognition accuracy of this network model reaches 97.9%, and the recognition effect of Chinese characters is better.

Remote sensing image target detection is one of great significance in military reconnaissance, intelligent agriculture and other fields, especially small target detection has been gaining continuous attention. However, small targets in remote sensing images face the problems of insufficient feature information and difficult detection, which have become the biggest obstacles plaguing the development of remote sensing applications. To this end, the you only look once-hybrid feature(YOLO-HF) algorithm was proposed, which introduced a hybrid attention mechanism of channel attention and self-attention in the network of the traditional YOLOv7 model to extract the target’s deep features, and fused the shallow and deep features to increase the richness of local features; to further strengthen the attention to the global information, a global attention mechanism was added for the small-scale targets after the extraction of the features, to achieve the ability of global feature expression enhancement. In order to avoid that the traditional loss function was sensitive to the positional deviation of small targets, which leaded to poor detection effect, a new metric was selected for use, which was embedded into the computation of the bounding box loss function, so as to accelerated the convergence of the loss function and realized the enhancement of the detection accuracy of small targets. The experimental results show that compared with the traditional YOLOv7 algorithm, the proposed algorithm shows superiority on both RSOD and NWPU VHR-10 datasets, and in particular, the mean average accuracy on RSOD dataset is improved by 2.90%, and the mean average accuracy on NWPU VHR-10 dataset realizes an improvement of 3.61%.

Accurately identifying the connecting handle of the train coupler is of great significance for the operation of the uncoupling robot. A train connection handle target recognition algorithm based on improved YOLOv5 was proposed to address this issue. The C_switchable atrous convolution(C_SAC) module was integrated into the backbone feature extraction network, and the wise intersection over union(WIOU) function was introduced as a new bounding box loss function to enhance the feature extraction ability of the backbone network, improve the model’s generalization ability and convergence rate. Then, images of the connecting handles of train couplers in different environments and positions on the production site were collected for recognition. The experimental results show that the improved YOLOv5 algorithm achieves a target recognition rate of 96.6% for the connecting handles of train couplers. Compared with the original algorithm, it shows significant improvements in accuracy, recall, average accuracy, and other aspects. Finally, it was applied in the development of an automatic uncoupling robot for train carriages, greatly improving the accuracy and effectiveness of automatic uncoupling.

In order to realize the automatic optimization of hyperparameters of YOLO model, the hyperparameter optimization of you only look once (YOLO) model based on orthogonal optimization strategy (OOS) was proposed. Firstly, based on the principle of statistical orthogonal test, the orthogonal search method of population and the hyperparameter contribution analysis strategy were proposed to improve the optimization efficiency of the algorithm. Then, the uniform orthogonal search strategy and the neighborhood orthogonal search strategy were designed to alleviate the problem of the YOLO model falling into the local optimum and premature convergence. Finally, YOLOv5, YOLOv5s-Transformer and YOLOv7 were used as optimization objects to test on two target detection datasets, NWPU VHR-10 and Pascal VOC. Test results show that the recognition accuracy of the YOLO model is improved by the OOS hyperparameter optimization method in all cases. The average recognition accuracy mAP@0.5 on two datasets is improved to 93.94%, 93.18%, 93.45%, and 85.81%, 84.59%, 89.96%. The mAP@0.5-0.95 is improved to 60.00%, 60.08%, 56.98%,and 62.27%, 58.89%, 70.77%. It can provide a new intelligent method for hyperparameter optimization of object detection model.

A lightweight and efficient two-stage video flame detection algorithm was designed to address issues of high false positive rates, poor adaptability, and low efficiency in complex scenes. In the first stage, an improved adaptive Gaussian mixture model (AGMM) was employed for rapid background modeling of video image sequences. Suspicious candidate regions were extracted from the sequences by leveraging the flickering and surging characteristics of flames. In the second stage, a residual deep normalization and convolutional neural network (ResDN) was used to discriminate these suspicious candidate regions. A simplified residual block was introduced to replace the original convolutional layers for a lightweight design, enabling accurate flame detection and localization. Compared with traditional classification algorithms, the proposed two-stage video flame detection algorithm effectively overcomes environmental interference in complex scenes, rapidly and accurately identifies flames, and demonstrates higher detection rates and adaptability.

Aiming at the problem of low tension control accuracy existing in the warp yarn tension control of carbon fiber corner link loom, a warp yarn tension control method considering the effect of beating-up was proposed. Firstly, the elongation of the warp yarn during beating-up was analyzed, and then a new tension control model of the warp feeding system was established by combining Hooke’s law with the existing tension model. Secondly, a command filter backstepping sliding mode controller was proposed, which estimated the first-order differentiation of the virtual control law through the filter to avoid the problem of “differential expansion”, and adopted the radial basis function(RBF) neural network to adaptively estimate the unmodeled part of the system, and at the same time, the sliding mode control was introduced to enhance the robustness of the system. Finally, MATLAB/SIMULINK software was used to carry out simulation experiments on the tension system. The results show that the filtered backstepping sliding mode control considering the effect of beating-up in tension control compared with the traditional backstepping sliding mode control in the case of similar response time, the stabilization time is shortened by nearly 16.3%, the amount of overshooting is reduced by 24.6%. Compared with the fuzzy proportional integral derivative(PID) control, the stabilization time is shortened by 51.7%, the amount of overshooting is reduced by 49.2%.

When utilizing double skin composite shear walls as the primary components for resisting lateral forces in buildings, it is crucial to ensure proper horizontal connection of the walls. Currently, traditional bolt connections and welding are the main methods used for horizontal connection of these walls. However, these methods present significant challenges during construction and do not fully exploit the structural advantages of the wall. Based on the structural characteristics of this type of wall, a new type of horizontal joint connection node was designed for the upper and lower layers of the wall. This new design featured a socket-type square semi-grouted sleeve connection. The influence of steel bar diameter, sleeve length, grouting material strength, and sleeve form on the tensile performance of this new connection node was explored using finite element simulation. The results indicate that this new node effectively connects the horizontal joints of double skin composite shear walls with several advantages including convenient construction and reliable performance. Furthermore, it is observed that failure mainly occurs on connecting steel bars which achieves an “equivalent cast-in-place” goal for prefabricated double skin composite shear walls.

In order to improve the engineering characteristics of silty soil in yellow plain area with low strength, easy deformation and poor bonding ability, mechanical testing and scanning electron microscope (SEM) were used to add different contents of xanthan gum(XG), The mechanical properties and improvement mechanism of XG, lignin fiber (LF) and curing age were studied. The results show that both XG and LF as improved materials can increase the compressive strength of silty sand. With the increase of XG content, the compressive strength of silty sand first increases and then decreases. With the increase of LF content, the compressive strength of silt will increase, and the improvement effect will be weakened by adding too much LF. When the two materials are added to the silt simultaneously, the compressive strength of the silt is higher than that of one material alone. XG produces high viscosity gel when it encounters water, the loose silty soil is tightly cemented together, and the strength of the soil is improved. LF contains large molecular groups, forming a spatial network structure with surrounding soil particles, which strengthens the joint force between soils. The research results can provide reference values for the silty soil subgrade improvement project in the yellow plain area.

In order to speed up the engineering progress, early strength additives are usually added to improve the early strength of cement grout. Calcium chloride, sodium sulfate, triethanolamine, synthetic calcium formate, HR-SA1 and other early strength additives were selected, the flowability, shrinkage rate, and compressive strength of cement slurry under the action of different early strength agents were explored through indoor experiments. In this way, the influence of different types of early strength agents on the early physical and mechanical properties of cement slurry were revealed. The results show that all early strength agents can meet the needs of on-site grouting requirements, ensuring that the fluidity of the cement slurry can be stable for a long time. The flowability of the synthesized calcium formate-based cement slurry significantly increases within 60 minutes, indicating a notable retarding effect. The fluidity of the other early strength agents cement slurry slightly increases within 60 minutes, but the amplitude is not significant. The compressive strength of (HR-SA1)-cement slurry specimens at all ages is significantly higher than that of ordinary cement slurry specimens, while the other four early strength agents have no significant effect on improving the compressive strength of cement slurry. The average shrinkage and shrinkage rate of (HR-SA1)-cement slurry specimens after final setting are very low, far lower than the average shrinkage and shrinkage rate of other early strength agent cement slurry specimens. It is verified by the construction site of prestressed anchor cable of a subway station in Qingdao City. The test values of the physical and mechanical properties of cement slurry are basically consistent with the results of laboratory test, which lays a foundation for the popularization and application of HR-SA1 early strength agent in practical engineering.

Microbial induced calcium carbonate precipitation (MICP) technology is an emerging green reinforcement technology for geotechnical engineering, which has a good application prospect in the reinforcement of loess slopes. The reinforcement of loess by MICP is affected by a variety of factors, in addition to the external environment, material properties and reinforcement methods and other factors, the calcium source, the concentration of binder, the age of the maintenance and the maintenance methods also play a decisive role in the microbial reinforcement of loess. The loess in Longxi area was taken as the research object, bacillus subtilis-induced calcium carbonate precipitation technology was adopted to consolidate loess, and a comparative experimental study on the shear strength of MICP-consolidated loess under the conditions of different calcium sources, binder concentration, age of maintenance and maintenance methods was carried out. The results show that the MICP technology is more effective in consolidating loess specimens when the calcium source is calcium chloride, the binder concentration is 1.0 mol/L and the specimens are cured for 7 d. The cohesion and internal friction are increased by 4.95 and 1.34 times, respectively, compared with the vegetal loess soil. The research results have certain reference value for the roadbed reinforcement and slope management in the Loess Plateau area.