The video question answering task aims to understand video content deeply and answer natural language questions. Current methods based on Multimodal Large Language Models （MLLMs） struggle to model global semantics and temporal dependencies across scenes effectively due to dual constraints of context length and computational complexity， resulting in degraded model inference performance. To address this issue， a Scene-Driven Adaptive Keyframe Sampling （SD-AKS） method and a Question-Driven Video Summarization （QD-VS） method were proposed. First， an iterative video scene clustering based on K-means was introduced， and a Scene Separation Score （SSS） was designed， so as to evaluate feature differences between scenes quantitatively， thereby enhancing the accuracy and robustness of scene segmentation. Second， a problem semantic-guided adaptive keyframe sampling was designed to achieve dynamic keyframe selection based on quantified scene information density and fine coverage of information-dense regions. Furthermore， based on the keyframe method， Large Language Models （LLMs） were employed to summarize the video-related question set， so that the task-oriented textual summaries were generated to improve global model inference capability. Experimental results show that compared to LLaVA?Video， the best?performing baseline model， the proposed method improves prediction accuracy by 5.3 and 1.8 percentage points on the EgoSchema and NExT?QA datasets， respectively， which verifies its effectiveness in cross?dataset scenes.

To address the problems of blurry generation results， detail loss， and uneven feature distribution caused by insufficient local information extraction of the existing models in sign language generation tasks， a sign language generation model based on Kolmogorov-Arnold Network （KAN） and Diffusion Transformer （KDT） was proposed. Firstly， the nonlinear approximation capability of the KAN was utilized to fit complex data distribution， so as to enhance the detail representation and motion fluency between video frames， thereby addressing the blurriness problem of videos generated by traditional Multilayer Perceptron （MLP） models. Then， Contrast Normalization （ContraNorm） was used to replace the original normalization， so as to address the uneven feature distribution problem by calibrating differences in feature scales， thereby ensuring the model’s stability with poor data quality and interference. Finally， diffusion Transformer was employed to achieve refined evolution from random noise to the target sequence through multi-step iterative optimization， thereby addressing the detail loss problem of traditional models. Experimental results on the validation set of RWTH-Phoenix-2014T continuous sign language dataset show that compared to the Sign-IDD （Sign-Iconicity Disentangled Diffusion） model， this model has the BLEU-1 （Bilingual Evaluation Understudy 1-gram） and ROUGE （Recall-Oriented Understudy for Gisting Evaluation） metrics improved by 8.1% and 5.9%， respectively， and the Word Error Rate （WER） metric reduced by 4.5%. The above results verify the effectiveness of this model in enhancing the richness of video details and the fluency of sign language movements.

With the development of wireless communication technology and high-precision mobile positioning technology， Internet of Vehicles （IoV） has become deeply embedded in everyday life. While IoV brings convenience to people， it also brings privacy risks. Typically， in IoV， vehicle driving information interacts with information of other vehicles and infrastructure in real time. During the interaction process， privacy issues such as the leakage of sensitive information may occur. Firstly， the location privacy architecture and privacy risks of IoV were introduced. Secondly， the dynamic noise allocation mechanism， multi-dimensional differential privacy trajectory protection and data perturbation technology in differential privacy were presented. Thirdly， the spatial generalization based on anonymization and the K-anonymity， as well as the asymmetric encryption， symmetric encryption， and homomorphic encryption of encryption mechanism were introduced. Finally， the advantages， disadvantages， limitations and other aspects of differential privacy， anonymity， and encryption mechanisms were analyzed and evaluated.

In order to improve the accuracy of loan financial customer behavior prediction, aiming at the incomplete problem of dealing with non-numerical factors in data analysis of traditional K-Nearest Neighbors (KNN) algorithm, an improved KNN algorithm based on Value Difference Metric (VDM) distance and iterative optimization of clustering results was proposed. Firstly the collected data were clustered by KNN algorithm based on VDM distance, then the clustering results were analyzed iteratively, finally the prediction accuracy was improved through joint training. Based on the customer data collected by Portuguese retail banks from 2008 to 2013, it can be seen that compared with traditional KNN algorithm, FCD-KNN (Feature Correlation Difference KNN) algorithm, Gauss Naive Bayes algorithm, Gradient Boosting algorithm, the improved KNN algorithm has better performance and stability, and has great application value in the customer behavior prediction from bank data.

For the color quantization problem of color images, the K-means clustering algorithm has strong dependence on initial conditions and is easy to fall into local optimum, and the traditional intelligent optimization algorithms only consider the mutual competition between individuals in the population layer and ignores the mutual cooperation between the population layers. To solve the problems, a K-means-based PES (Pyramid Evolution Strategy) color image quantization algorithm was proposed. Firstly, the clustering loss function in K-means clustering algorithm was used as the fitness function of the new algorithm; secondly, PES was used for the population initialization, layering, exploration, acceleration and clustering of the colors; finally, the new algorithm was used to quantify four standard color test images at different color quantization levels. The experimental results show that the proposed algorithm can improve the defects of the K-means clustering algorithm and the traditional intelligent algorithm. Under the criterion of intra-class mean squared error, the average distortion rate of the image quantized by the new algorithm is 12.25% lower than that quantized by the PES-based algorithm, 15.52% lower than that quantized by the differential evolution algorithm, 58.33% lower than that quantized by the Particle Swarm Optimization (PSO) algorithm, 15.06% lower than that quantized by the K-means algorithm; and the less the color quantization levels, the more the image distortion rate reduced quantized by the new algorithm than that quantized by other algorithms. In addition, the visual effect of the image quantized by the proposed algorithm is better than that quantized by other algorithms.

For a kind of linear bi-level programming problems with interval coefficients in the upper level objective, a Genetic Algorithm (GA) was proposed by using a double fitness function evaluation technique, which was characterized by simultaneously obtaining the best optimal solution as well as the worst one in one run of the genetic algorithm. Firstly, individuals were encoded by using the vertices of the constraint region, and a double fitness was constructed by the upper and lower bounds of the upper level objective coefficients. Secondly, fitness functions were used to sort all individuals in populations. According to the order, the feasibility of individual was checked one by one until a feasible individual was found. Finally, the feasible individual was updated in executing algorithm. The simulation results on four computational examples show that the proposed algorithm is feasible and efficient.

The newly developed botnet defense technologies pose a severe challenge to botnet survivability. In order to improve the survivability of the botnet, from an attackers perspective, this article proposed a new anti-strike mechanism based on regroup, which was suitable for semi-distributed botnet. In the case that semi-distributed botnet suffered a severe blow, which caused topology broken, this mechanism could perceive the state of botnet, detect survival nodes, recover survival node and reassemble them into a new botnet. The experiments verify the effectiveness of the mechanism to effectively enhance the survivability of the semi-distributed botnet.

An adaptive filter control method based on Recursive Least Squares (RLS) was proposed for solving the low convergence speed of Filtered-X Least Mean Square (FXLMS) and Filtered-U Least Mean Square (FULMS) algorithms. It was roughly composed of two parts: Infinite Impulse Response (IIR) filter and RLS algorithm. IIR filter was as the main frame of the whole algorithm and adjusted the filter weights in real-time to realize the adaptive filter control. Seen from the analysis and comparison, the algorithm has higher convergence speed and the overall vibration response of the controlled object drops by about 65%, which full proves the validation and feasibility of the algorithm.

At present there are many kinds of technologies which can track, detect and counter botnet effectively, which are serious threats to botnet. In order to improve the survivability of botnets, with the analysis on the existing anti-botnet technology, the paper proposd a new P2P-botnet based on Kademlia protocol from an attacker's prospective. A communication encryption and node authentication mechanism was designed. The theoretical analysis shows that the mechanism can effectively address improper command attack and sybil attack. Eventually, the experimental results verify that this botnet has high survivability.

In order to improve the control precision and to speed up the convergence rate of the controlled system, a kind of fuzzy PD type iterative learning control algorithm was put forward based on genetic algorithm. In the proposed approach, the iterative learning controller was designed by fuzzy Takagi-Sugeno-Kang （TSK） system, the parameters of fuzzy TSK system were calculated by genetic algorithm, and then appropriate updating law was created. Appropriate iterative learning control algorithm of controlled system was designed and compared with PD iterative learning control algorithm and fuzzy PID iterative learning control algorithm, and then the proposed algorithm was used in double joint manipulator simulation. The simulation results show the effectiveness of the proposed algorithm.

Concerning the lack of refined and quantized real-time evaluation of rainstorm disaster risk level, this paper applied meteorological data, historical disaster data, height and distance from the sea of Guangxi, and confirmed the identification technique and data sequence building method of hazard-formative factors of rainstorm of Guangxi based on hazard-bearing body, the hazard-formative environment, hazard-formative factors, and anti-disaster capability. Real-time evaluation model and grade index of rainstorm disaster risk level based on risk, subsequently environment fragile degree, vulnerability and anti-disaster capability were constructed for different hazard-bearing body, such as agriculture and social economy. And then the rainstorm risk level of real-time evaluation system was developed, with real-time evaluation model as the core. By using the Geographic Information System (GIS) secondary development techniques, the operating process of rainstorm risk level real-time evaluation was simplified and standardized. By using the proposed system to evaluate the violent typhoon named Neuchatel on September 29, 2011, the experimental results show that it is consistent with disaster condition.

A new method to design a 3D surface movement simulation system for A-SMGCS based on Linux operation system and FilightGear flight simulator was proposed. The simulation system was driven by real-time track data of ADS-B. An airport model was built up according to a real airport prototype by modeling of aircrafts, airport layout, terminal, tower and terrain. Because ADS-B track data had no flight attitude data, an algorithm to extract flight attitude from a space vector, which was connected by two pre and post track points, was proposed. Six degrees of freedom (6-DOF) aircraft movement simulation was realized by importing position and attitude data into FlightGear through its multiplayer interface to drive aircraft models. The simulation results show that, the system can simulate surface movement accurately and realistically. The simulation system is low-cost because it is based on free, open source software.

For a kind of fractional bilevel programming problems with more than one linear follower, a genetic algorithm based on a new encoding scheme was proposed. Firstly, the dual theory was applied to transform the original problem to a singlelevel nonlinear programming; secondly, all individuals were encoded by considering feasible base of the follower's dual problem. For any individual given, the dual variables can be solved, which makes the nonlinear problem become a linear fractional programming; finally, the resulting linear fractional programming was resolved and the objective value was taken as the fitness of this individual. Based on the encoding scheme and fitness evaluation, the efficiency of the genetic algorithm is improved, which is also illustrated by the simulation of four examples.

In order to solve the problem of geometric distortion and low efficiency in the process of biased skeleton trim, a new algorithm of biased skeleton trim based on intersecting cortical model was proposed. At first, according to inherent features of skeleton biased branch, definitions of endpoint and junction point were introduced and revised in the algorithm to accurately locate skeleton branch and biased branch. Then, with that information and the iteration number of intersecting cortical model, flameout condition of neurons spreading was set up. Finally, guided by that condition, the biased skeleton branch can be judged fast and trimmed accurately, with the aid of impulse dynamically generated by ignition neurons, which has biological nature of parallel transmission. Compared with conventional methods based on mathematical morphology, the experimental results show that the proposed algorithm has good performance in structural integrity of skeleton, as well as computation speed and anti-noise ability.

Concerning the shortcomings of classical Hidden Markov Model (HMM) in handwritten Uighur characters recognition, such as largly varied width of characters, slow convergent speed and premature convergence, a new Uighur characters recognition algorithm was proposed in combination with Locality Preserving Projection (LPP) and HMM. Firstly, the aspect ratio of original image was maintained by a highly-normalized method. Sub-images were obtained by using sliding window, and observation sequences were extracted from these windows. Secondly, the observation sequences were mapped into low-dimensional space based on LPP, and the scale of adjacency matrix was reduced via the random sampling technique. Finally, HMM was trained by adopting obtained observation sequences. The algorithm decreases dimension of observation vectors, accelerates the convergence, and prevents premature convergence effectively. The simulation results show the LPP-HMM algorithm is efficient and robust, which decrease average convergence steps as well as errors.

The strip packing problem was studied. The algorithm divided the strip into layers with horizontal cuts. The width of a layer was equal to the strip width, and the height was the same as that of the leftmost item in the layer. The algorithm used a recursive procedure to arrange the layers on the strip, selected the leftmost item for each layer, and determined the item layout on each layer. Branch-and-bound techniques were used to shorten the searching space. The computational results indicate that the algorithm is more efficient than several recent algorithms.