Lanes play a very critical and important role in maintaining the orderly operation of road traffic. Therefore, the automatic lane detection is very important and has significant potential value. Combining the existing lane detection methods based on deep learning, this paper proposes a new lane detection model, Bi-Lanenet, to solve the problems that still exist in the current methods, aiming to overcome the disadvantages in these methods and improve the practicality of the lane detection algorithm. For the task of lane detection, this paper improves the segmentation accuracy and improves the traditional lanenet model while ensuring that the network model is lighter weight and more robust. Then, we propose a new bilateral lane recognition network based on semantic segmentation and details, and use the random sample consensus (RANSAC) method to optimize the post-processing process. We conduct experiments on the TuSimple and CULane datasets and prove that the method can detect lanes in the image efficiently with 110 frame-per-second (FPS) and accurately at 97.08%.

In mobile crowd sensing, the mobile terminal integrates a variety of widely distributed sensing devices and communication ports. Sensing devices and communication ports can collect and share all kinds of perception data. However, inherent contradictions exist among perceived ability, communication port, and moving rule while collecting real-time and accurate sensing information. This article mainly focused on recruited and selected mobile nodes and assigned sensing tasks to improve the quality of sensing information. The optimization of the implementation stage of the sensing task is beyond the scope of this study. This article proposes a utility-based sensing task decomposition and subcontract algorithm, which is a method of sensing data acquisition that establishes direct collaboration between mobile nodes. A mobility model based on Markov chain is established to forecast the spatial distribution of sensing nodes. A utility function is designed to estimate the sensing task execution capacity of sensing nodes based on spatial distribution and tempo-spatial coverage of the collected data. The sensing tasks are then decomposed and subcontracted to neighboring nodes according to the utilities of the neighboring nodes to the decomposed sensing tasks. This method improves the quality of sensing data in terms of sensing data coverage and finished ratio of sensing task.

Online learning is gradually becoming popular with the continuous development of Internet technology and the rapid development of educational informatization. It plays a key role in predicting students’ course performance based on their online learning behavior. It can optimize the effects of teaching and improve teaching strategies. Student performance prediction models that are built with a single algorithm currently have limited prediction accuracy. Meanwhile, model fusion improvement technology can combine many algorithms into a single model, thereby enhancing the overall effect of the model and providing better performance. In this paper, a stacking fusion model based on RF-CART–XGBoost–LightGBM is proposed. The first layer of the model uses a decision tree (CART), random forest, XGBoost and LightGBM as the base models. The second layer uses the LightGBM model. We used the Kalboard360 student achievement dataset, and features related to online learning behavior were selected as the model’s input for model training. Finally, we employed five-fold cross-validation to assess the model’s performance. In comparison with the four single models, the two fusion models based on the four single models both show significantly better performance. The prediction accuracies of the bagging fusion model and stacking fusion model are 83% and 84%, respectively. This proves that the proposed stacking fusion model has better performance, which helps to improve the accuracy of the performance prediction model further. It also provides an effective basis for optimizing the effects of teaching.

Currently, the on-site measuring of the size of a steel pipe cross-section for scaffold construction relies on manual measurement tools, which is a time-consuming process with poor accuracy. Therefore, this paper proposes a new method for steel pipe size measurements that is based on edge extraction and image processing. Our primary aim is to solve the problems of poor accuracy and waste of labor in practical applications of construction steel pipe inspection. Therefore, the developed method utilizes a convolutional neural network and image processing technology to find an optimum solution. Our experiment revealed that the edge image that is proposed in the existing convolutional neural network technology is relatively rough and is unable to calculate the steel pipe’s cross-sectional size. Thus, the suggested network model optimizes the current technology and combines it with image processing technology. The results demonstrate that compared with the richer convolutional features (RCF) network, the optimal dataset scale (ODS) is improved by 3%, and the optimal image scale (OIS) is improved by 2.2%. At the same time, the error value of the Hough transform can be effectively reduced after improving the Hough algorithm.

To cope with the challenges posed by the complex linguistic structure and lexical polysemy in ancient texts, this study proposes a two-stage translation model. First, we combine GujiBERT, GCN, and LSTM to categorize ancient texts into historical and non-historical categories. This categorization lays the foundation for the subsequent translation task. To improve the efficiency of word vector generation and reduce the limitations of the traditional Word2Vec model, we integrated the entropy weight method in the hopping lattice training process and spliced the word vectors with GujiBERT. This improved method improves the efficiency of word vector generation and enhances the model’s ability to accurately represent lexical polysemy and grammatical structure in ancient documents through dependency weighting. In training the translation model, we used a different dataset for each text category, significantly improving the translation accuracy. Experimental results show that our categorization model improves the accuracy by 5% compared to GujiBERT. In contrast, the Entropy-SkipBERT improves the BLEU scores by 0.7 and 0.4 on historical and non-historical datasets. Ultimately, the proposed two-stage model improves the BLEU scores by 2.7 over the baseline model.