Image alignment is a fundamental problem in the field of computer vision and an important prerequisite for carrying out many other tasks. Firstly, the theoretical basis and realization method of image alignment as well as the process and the method of alignment are introduced to provide alignment ideas. Subsequently, an image alignment method based on the union of multi-scale features is proposed, and a new loss term is introduced to the small-scale features therein, which further improves the distinguishability of the small-scale feature descriptors while guaranteeing the invariance of the large-scale feature descriptor matching therein. Three common alignment algorithms (RIFT algorithm, HAPCG algorithm, and SAR-SIFT algorithm) are selected for stability assessment and quantitative evaluation on the dataset, and an image enhancement algorithm with histogram equalization is used to enhance the dataset. The results show that the feature stability of this paper’s method is described as 99.1%, which is better than other algorithms. Meanwhile the desired effect is achieved on the dataset.

At present, the evaluation of spoken English in domestic universities is affected by the evaluation teachers’ personal cognition, preference, time, energy and other factors, and it is difficult to unify the standard of oral evaluation in the implementation, and the evaluation frequency and timeliness are insufficient to meet the students’ willingness to improve their oral language. In this paper, multimodal speech recognition technology is utilized to firstly collect students’ speech signals through microphone arrays, secondly extract acoustic and linguistic features of speech, and construct multimodal feature vectors by combining visual information such as students’ lip movements and facial expressions. Subsequently, the feature vectors are input into a deep neural network model for training and recognition, fusing LSTM network with attention mechanism to analyze the speech emotion and capture the emotional changes in speech. Meanwhile, the interaction behavior in speech is analyzed by combining temporal convolutional network. Construct a deep reinforcement learning model, introduce a user item interaction layer, design a user interaction simulator, and obtain user feedback on the smart English classroom. Using multimodal speech recognition technology, the temporal waveform of classroom speech is analyzed for sound pressure value, and the normalized sound pressure value range fluctuates around [-1.5,1.5].The average recognition rate of the six emotions rises to 67.86% with the joint effect of LSTM and attention mechanism. By comparing the experiment, analyzing the difference between the experimental class and the control class before and after the reading aloud ability, the average score of the experimental class is 23.945, and the average score of the control class is 21.464, at the same time, the post-test of reading aloud ability corresponding to the experimental class and the control class P=0.005<0.05. It can be seen that the intelligent interactive classroom of English language constructed in this paper has a facilitating effect in the process of teaching reading aloud in the aspect of reading aloud ability of students The classroom can be seen that the intelligent English interactive classroom constructed in this paper has a promoting effect in the process of teaching reading aloud in terms of students' reading ability.

Intellectualization of agricultural machinery can effectively improve the efficiency and quality of operations, and has an important role in promoting agricultural development. Based on AR technology, this paper introduces the key technology to build the interactive control system of agricultural machinery, uses NURBS to realize virtual agricultural machinery modeling, uses VRML technology to design a prototype of the scene environment of interactive farmland virtual reality, and details the methods of virtual modeling, virtual roaming, interactive control and collision detection in the process of system development. A four-degree-of-freedom simulation test bed is established to realize the simulation of the tractor’s attitude when walking in the field. The position information of the crop rows is extracted from the virtual scene, and the control signals are given according to this information to carry out the speed, direction and balance control of the traveling of the agricultural machine, so that the tractor travels along the crop rows. The maximum deviations of the roll angle, pitch angle and yaw angle are within 0.36°, and the maximum deviations of the elevation and traveling speed are 2.11 mm and 0.14 km/h. The simulation analysis and the physical test show the feasibility of the interactive control system of the farm machine.

In this paper, we use a large language model for business English translation and context analysis, and propose an adaptive parameter unfreezing method based on the quantization difference between adjacent layers within the decoder to fine-tune the layers of the language model related to the translation task, and to understand the behavior of the model in the relevant layers. Then the method of combining different encoders is proposed as a dual encoding-decoding framework on top of the traditional encoding-decoding framework, which is applied to the task of context analysis in business English translation. The fine-tuning method in this paper significantly improves the text translation quality of the language model, especially in the English-X tri-lingualization, which improves the COMET and BLEU metrics by 3.22 and 2.58 points respectively. In addition, the dual encoding-decoding model proposed in this paper is applicable to the task of contextual analysis in business English translation, which significantly improves the performance of contextual analysis in business English, and the F1 value on the HIT-CDTB dataset is improved by 11.60% compared with that of Rutherford’s model. The experiment proves that the proposed method of text has made progress in the research of the task of analyzing textual contextual relations in business English.

In this paper, the 3D reconstruction of the finite element model of the knee joint is completed by first generating and editing the 3D images of the martial arts movements through Mimics software. After that, Hypermesh and Abaqus software are used to pre- and post-process the properties of materials in the knee joint biomechanical finite element model. Visual 3D software and low-pass filter smoothing technique were used to obtain and process the kinematic and kinetic data of the martial arts maneuvers, and the processed data were used as boundary and loading conditions to import the data of the three martial arts maneuvers, namely, horse stance, lunge stance, and servant stance into the finite element model for calculating and comparing the biomechanical responses of the articular cartilage and meniscus. The results showed that the movement pattern of horse stance has a larger knee range of motion and a smaller peak ground reaction force compared to the lunge and servant stance movements in the martial arts maneuvers. Finite element simulations showed that the straddling knee stance produced smaller peak contact stresses on the knee cartilage and meniscus, and the peak stress area changed more during the movement. Three-dimensional finite element simulation analysis obtained four characteristic moments, namely: the first peak ground reaction force moment, the maximum external rotation-external rotation moment, the maximum dorsiflexion moment, and the second peak ground reaction force moment, which corresponded to a greater difference in ground reaction force values. Therefore, it is recommended to wear protective equipment in advance for the injury-prone areas to reduce the risk of injury before the wushu performance.

Under the background of big data era, big data mining technology is widely used, through data mining technology, deeper exploration of data, discovering the relevance of data, can provide decision support for decision makers. This paper analyzes the Internet big data of college students’ employment decision-making based on big data mining technology, uses Apriori algorithm to mine the influencing factors of college students’ vocational skills generation, meanwhile applies ID3 decision tree algorithm to analyze the college students’ tendency of vocational choice, and explores the relevant factors affecting college students’ employment through correlation analysis and clustering analysis. The results of the study show that students’ personal, family and school have strong correlation with students’ vocational skills generation, which affects the improvement of students’ personal job-seeking ability. Meanwhile, the ID3 decision tree algorithm is applied to the employment consulting service for graduates to construct a career decision tree for individual college students, which visualizes their career choice paths under the influence of career values and helps them make more appropriate career choices. In addition, qualification certificates, social practice experience, academic performance, expected salary, ideal employment unit and other factors will affect the employment choice of college students, and there are individual differences among different students.

With the continuous promotion of the integration of industry and education, constructing a quality evaluation system for the integration of industry and education in vocational education has become a key issue to improve the level of vocational colleges and universities’ curricula. Based on the CIPP model, the article builds a quality evaluation system of vocational education industry-teaching integration that includes 4 first-level indicators, 12 second-level indicators and 34 third-level indicators, and empirically analyzes the quality of industry-teaching integration in three higher vocational colleges, H1, H2 and H3, using the fuzzy comprehensive evaluation method through the questionnaire survey from the viewpoint of empirical application. According to the results of the fuzzy comprehensive evaluation, the quality of industry-teaching integration in H1 and H2 higher vocational colleges and universities belongs to the good level, and its comprehensive judgment value is 78.2 and 78.395 respectively.The comprehensive judgment value of the quality of industry-teaching integration in H3 higher vocational colleges and universities is 82.037, which belongs to the excellent level. The three sample higher vocational colleges have achieved outstanding results in the integration of industry and education, providing an example for the development of integration of industry and education for higher vocational colleges in the region.

In the digital campus network security construction, the existence of potential security vulnerabilities can easily cause serious threats to campus information security, resulting in significant losses. In order to prevent and mitigate the risk, the article designs a security vulnerability identification system. Firstly, the URL similarity is compared by machine learning in order to scan the vulnerability information. The SeCF embedding layer is utilized to improve the input speed and the discard layer is designed to improve the overfitting problem during the training process. Finally, TextACBL security vulnerability identification model is proposed by combining CA, 1D-CNN and BiLSTM techniques and analyzed numerically. The average recognition rate of this paper’s method is as high as 80% for 10 common security vulnerabilities, which achieves better security vulnerability recognition results compared with existing methods such as cppcheck, deepbugs, flawfinder and vuldeepecker. The experimental results verify the effectiveness and feasibility of the method in this paper, which provides ideas for safeguarding campus network security during the construction of digital campus.

This paper focuses on the characteristics of multilevel information extraction, based on the convolutional neural network model (CNN), introduces the multi-scale feature fusion and multilevel feature fusion strategy to study the multilevel information extraction method, and proposes the full convolutional neural network based on the attention mechanism and residual connection to form the multilevel information extraction model. Aiming at the gradient disappearance and saddle point problem of convolutional neural network, an activation gradient (AG) algorithm is proposed to optimize its training, which is improved to a class of activation gradient convolutional neural network (AG-CNN). The practical application effect of the multilevel information extraction model in this paper is verified by the information extraction work of net-pen culture in river-type reservoirs. Compared with the classical models such as UNet and ResUNet, the intersection and integration ratio (IoU), recall rate, precision rate, and F1 score of this paper’s model reach the highest 80.28%, 91.02%, 87.18%, and 89.03% among all the models, which possesses a stronger extraction capability. And in the multilevel information extraction experiments on Cifar100 and Caltech256 datasets, when the number of batch training data is greater than 100, the accuracy rate and performance of the experimental group basically remain stable.

Scientific and efficient curriculum design and teaching activity plan is the key to the quality of teaching in higher vocational colleges and universities. Based on the principle of SPOC segmented teaching, this paper proposes a “two-line hybrid” language teaching model. Combined with the implementation process of blended teaching, a blended teaching quality evaluation index system for higher vocational colleges is constructed, which includes the dimensions of rule of law and ethics, professionalism, learning ability, skills and technology. Using the standardization principle of hierarchical analysis, the judgment matrix was constructed by comparing two by two to achieve the empowerment of the indicator system. Introducing cloud model comprehensive evaluation, combining the weights of indicators from the forward cloud generator to get the cloud diagram, and derive the evaluation results. The initial matrix is constructed according to the scores of experts, and all the items passed the consistency test, which verifies that the index system has high reliability and validity. The obtained cloud diagram shows that the cloud model parameter Ex = 5.462, in which the A rule of law ethical dimension Ex is about 5.58, closest to the medium level. This paper makes a useful exploration for actively promoting the teaching reform of higher vocational discipline courses.