In order to address the challenges of urban traffic congestion and safety，an ACLR model was proposed. By integrating ConvLSTM，attention mechanisms，and residual structures，the ACLR model effectively enhanced the extraction of spatio-temporal traffic features.The time，space and other characteristics of taxi traffic were processed respectively，and the influence of regional point of interest(POI) data on taxi traffic was mined. Additionally，a specialized learning component was incorporated to capture the impact of external factors and point-of-interest density on traffic flow. Using taxi trajectory data from Beijing，the ACLR model demonstrates superior prediction accuracy compared to other models such as the autoregressive integrated moving average (ARIMA) model，long short-term memory (LSTM)，deep spatio-temporal residual networks (ST-ResNet)，convolutional neural network(CNN)-ResNet-LSTM (CRL)，and attentive crowd flow machines (ACFM) in urban traffic flow forecasting，which is helpful to improve the prediction performance of the model without POI density or considering POI density. The predicted value of the ACLA model is basically consistent with the real value，and it can also be in good agreement with the real value during peak hours，which effectively improves the ability to extract traffic temporal and spatial characteristics，reduces the prediction error，and optimizes the traffic flow prediction performance.

In order to improve the high-precision detection and early warning of crane safety operation and enhance the safety management ability of enterprises，focusing on the needs of unmanned industrial safety incident analysis and monitoring and early warning，an inspection robot that combines ground and air flight in hoisting scene was customized to intelligentize hoisting safety monitoring，pop-up image recording and safety alarm.A lifting dataset Cranes-Dataset (CRN-Dataset) containing 3 120 images was made，and DPIM algorithm was proposed to enhance the rapid detection ability of multi-scale objects.Based on corner detection and density-based spatial clustering of applications with noise and considering the safety attributes of the space distance between cranes and workers，the process of triggering alarms based on safety rules was developed to record real-time illegal operation image and popup alarm.The results show that，after actual deployment and verification，the DPIM algorithm significantly improves target identification ability compared with other traditional algorithms，and it is suitable for real-time calculation and data transmission of embedded edge intelligent analysis nodes to complete field deployment.

In order to assist earthquake rescue personnel in enhancing disaster response speed and adapting to diverse search and rescue needs，an intelligent management method for earthquake rescue equipment information based on a knowledge graph was proposed. Through the top-down knowledge graph construction method，earthquake rescue knowledge was first obtained from various information sources to serve as the basis for knowledge modeling. Next，a rule-based method was used to extract search and rescue knowledge，which was then integrated based on cosine similarity. The integrated knowledge was stored in the form of Resource Description Framework (RDF) triples. Subsequently，the open-source graph database Neo4j was employed to organize the triples into a visualized knowledge graph. Finally，a question-and-answer system was built based on the knowledge graph，allowing users to query the knowledge on the graph using natural language. The results indicate that the knowledge graph includes five categories of entities and relationships: disasters，secondary disasters，environmental factors，rescue needs，and rescue equipment. It facilitates quick matching of equipment based on rescue needs. The knowledge graph-based method can effectively manage and schedule rescue equipment information，improving the efficiency of the preparation phase of rescue operations.

To improve the efficiency of disaster response，the "Hebei rainstorm" and "Heilongjiang rainstorm" were adopted as illustrative cross-regional research cases，and text-image-audio multimodal data were collected from short videos. In the face of massive unstructured data，deep learning technology was employed to realize the extraction of multimodal emotional features，cross-modal integration and intelligent sentiment classification in short videos. By comprehensively using spatial and temporal big data，the multimodal emotional characteristics of short video of rainstorm disaster were deeply mined and analyzed in the spatial and temporal dimension. The results indicate that the model's accuracy exceeds 85%，efficiently fulfilling the objectives set for short video analysis. From the temporal perspective，the emotional fluctuations of netizens broadly align with the cycle of rainstorm disasters，providing a basis for assessing disaster severity and public opinion trends. Furthermore，the intervention of media and government entities plays a significant role in shaping the emotional evolution surrounding rainstorm disasters. In terms of spatial dimensions，negative emotions exhibit a "low-high-low" trend as disasters shift locations，and the resonance and diffusion of these emotions display distinct regional characteristics. Therefore，it is imperative to prioritize public opinion guidance in disaster-stricken areas，as well as in some eastern regions of China and non-disaster areas experiencing similar phenomena.

To solve the crowd congestion problems caused by a large number of phubbers in enclosed one-way long passages in public spaces such as the subway，an intervention model was proposed to calculate the intervention critical value. Experiments were performed to analyze the behavioral characteristics of phubbers and normal pedestrians in enclosed one-way long corridors. Then，different distribution functions were used to propose a small-scale behavioral model. Furthermore，a large-scale congestion intervention decision-making model was proposed based on the proportion of phubbers and crowd density. Finally，the critical value curve between passenger density and the proportion of phubbers was validated against a one-way long passage in a Beijing subway station. The results indicated that behavioral characteristics of phubbers presented as slow following，while normal pedestrians tended to speed up and overtake whenever possible. The simulations calculated the critical value curve between the passenger's density and the phubber's proportion. If the calculated value was lower than the critical curve，it was a low-risk area without any intervention strategies. Otherwise，intervention strategies were required to avoid serious congestion.

In order to solve the problem of determining the distribution of different types of objects in system faults，a method to determine the distribution of objects was proposed. Firstly，the characteristics of the system fault evolution process and object distribution were discussed. Secondly，the method flow chart and implementation process were given. Finally，an example was analyzed. The example studied the basic data matrix composed of 6 factors and 50 objects，and the maximum training set cross-correlation was 0.8，the test set cross-correlation was 1，and the optimal object label distribution (object distribution) was obtained. Finally，the advantages and disadvantages of the method were described. The analysis shows that the database for studying the evolution process is the object set. Methods based on UKSR，combined with K-means and mutual information methods，a randomly distributed object label set is constructed，and the criteria for the optimal object label set are proposed. The optimal object label set is determined through a loop when the correlation between object labels and object data is the largest. The label value of objects in the set is the optimal object distribution. The method overcomes the problem of unsupervised learning and nonlinear mapping. It is concluded that the method can classify the measured objects in the system fault evolution process under unsupervised and nonlinear conditions，and the distribution of class labels of all objects with evolution time. The disadvantage is that it can only be used to study the system fault evolution process represented by two-dimensional.

To solve the problem that traditional safety design methods based on manual inspection were difficult to cope with the explosion of optional residence solutions caused by the large-scale integration of avionics systems，an avionics system partition model，task model and safety criticality level quantification model were constructed，and the comprehensive design optimization considering safety was modeled as an MDP problem. An optimization method of Soft Action-Critic (SAC) algorithm based on Actor-Critic framework was proposed. In order to obtain the correlation between the parameter selection and training results of SAC algorithm，the sensitivity of the algorithm parameters was studied. At the same time，to verify the superiority of the optimization method based on the SAC algorithm in optimizing the comprehensive design considering safety，optimization comparison experiments were carried out with the Deep Deterministic Policy Gradient (DDPG) algorithm and the traditional allocation algorithm as the objects. The results show that under the optimal parameter combination，the maximum reward after using convergence of SAC algorithm increases by nearly 8% compared with other parameter combinations，and the convergence time is shortened by nearly 16.6%. Compared with the DDPG algorithm and the traditional allocation algorithm，the optimization method based on SAC algorithm has improved approximately 62%，7464%，8370%，2123% and 775% in terms of the maximum reward，cumulative constraint violation rate，partition balance risk effect，partition resource utilization and solution time

Aiming at the problem of UAV aerial photography parameters relying on manual experience when collecting image of 3D real scene reconstruction at traffic accident sites，which led to large model measurement errors and low precision，an automatic calculation method of key parameters of UAV aerial photography was proposed. First，the key parameters of aerial photography used by single-lens UAV for images acquisition of traffic accident site were altitude，gimbal angle and shooting interval angle. The numerical relationships of the key parameters with the imaging range，imaging accuracy and overlap rate were analyzed. Then，the aerial photography key parameters computation model was constructed. The key input parameters were the given accident site，UAV technical parameters，image aspect ratio and overlap requirements. On the premise that the accident site was in the effective imaging range and there was no imaging blind zone，the UAV photography parameters were automatically calculated with the goal of improving the accuracy and presentation effect of the image utilization rate model. Finally，combined with case application，the UAV aerial photography parameters calculated by this method were applied to complete the image acquisition at the accident sites，and the constructed 3D real scene model could clearly and completely present the overview of the accident site，with an average measurement error of 1.72%，and a measurement accuracy of 3.54 cm. Compared with the manual empirical method，the average error of the method was reduced by 47.56%，and the accuracy was improved by 48.40%. The study shows that this method can realize the automatic quantitative calculation of aerial photography parameters for 3D real scene reconstruction of traffic accident sites，construct the model with centimeter-level error，and improve the parameterization and automation of UAV aerial photography.

Frequent heavy rainfall events cause a dramatic increase in attached elevator scaffolding accidents. In order to improve construction safety and reduce the accident rate under heavy rainfall scenarios，an accident causation analysis model based on combination of IFRAM and BN was proposed. The model first qualitatively identified accident mechanisms and explored system functional resonance using IFRAM. Next，IFRAM was mapped to a BN quantitative analysis model，and the prior probabilities of each root node were computed using cloud optimization. Finally，taking the Xi'an "9.10" accident as an example，empirical research was conducted，and corresponding preventive measures were proposed. The results indicate that accidents are most likely to occur when the safety status is IV. The core causes of climbing accidents include workers violating regulations，failure to conduct mandatory supervision such as standing by and heavy rainfall. The combination of factors such as heavy rainfall environment and overloading of the frame after rain is the key to inducing frame climbing accidents.

In order to investigate the influence of internal and external factors on the safety state of the tailings dam，a method for analysing the safety state of the tailings dam based on heterogeneous hierarchical diagrams was proposed. Firstly，a hierarchical causal graph was constructed based on a priori knowledge to link key factors such as environment，seepage field and stress field with the safety status of tailing dams，and an evaluation index system was established by combining the attribute characteristics of heterogeneous nodes. Secondly，the cloud model and set-pair analysis theory were used to quantitatively calculate the potential logical relationship between the heterogeneous causal graph and the safety stability of the tailings dam. A dynamic interval calculation method for quantitative indicators and a safety status level calculation model were proposed to convert the fuzziness and uncertainty of complex and diverse evaluation indicators into quantitative expressions. Finally，a tailings dam in Luoyang was used as an example to verify the scientificity of the model.The results show that the model can quantitatively analyse the link between factors and states and identify the causes of negative changes in stacked dams. The conclusions of the model analysis can be used for the safety management of the dam-building process.