The difficulty in establishing the transmission error model of non-circular gears and obtaining the backlash limit its application in precision occasions. In order to improve the transmission accuracy of non-circular gears， the non-circular gear transmission error model and backlash model under bidirectional transmission are established based on the meshing line incremental method considering the eccentricity error. The influence of eccentricity and eccentricity error on bidirectional transmission error and backlash are analyzed， and the correctness of theoretical analysis is verified by transmission test. The analysis shows that the backlash of the non-circular gear can be obtained by the bidirectional transmission error. With the increase of eccentricity and eccentricity error， the two-way transmission error shows an increasing trend. Optimizing the initial phase can reduce the transmission error and backlash to a certain extent. As the load increases， the transmission error and backlash also show an increasing trend. And due to the existence of gear tooth deformation， there is a certain cumulative value of the backlash.

In this study， a nonlinear hierarchical convolutional autoencoder （H-CNN-AE） is employed to sort the energy content of the latent vectors of AE， which is analogous to the method of proper orthogonal decomposition （POD）， and at the same time result in better performance in terms of reduced-order modelling within limits. This method is applied to a turbulent wake behind a three-dimensional circular cylinder bluff body at Re=20 000. We assess the ability of H-CNN-AE with L₂ error and compares it with the results of POD. Furthermore， the concept of grouping AE-modes is extended. We observe the change of mean square error between the reconstructed and the original flow when adding the number of latent vectors of each group. It is demonstrated that when the number of subnetworks and low-dimensional vectors in latent space of each subnetwork is small， H-CNN-AE has better capability to restore the flow field than POD. However， it is also found that the strength of H-CNN-AE will weaken with the increase of the number of subnetworks and latent AE modes and will even be inferior to POD under certain conditions.

In order to study the merging behavior of vehicles in the interweaving area of a freeway， a random parameter model with heterogeneity in variances was used to build a vehicle lane-changing model. Seven explanatory variables that had a significant effect on lane-changing behavior statistically， extracted from NGSIM dataset， was introduced to random parameter model with heterogeneity in variances. The potential heterogeneity was explored by the model. The average margin effects were calculated to quantify the effects of explanatory variables on lane-changing behavior， and individual sample accuracy was established to compare the model accuracy. The results of the study indicate that the space headway between the merging vehicle and the putative leading vehicle， the width of the putative leading vehicle and the speed of the leading vehicle in the auxiliary lane have significant effects on lane-changing behavior， and the longitudinal position of the merging vehicle affects significantly the space headway between the merging vehicle and the putative leading vehicle. Compared with random parameter model that don't consider the variance heterogeneity and binary logit model， random parameter model with heterogeneity in variances has the highest goodness of fit and model accuracy， and can better explain the unobserved heterogeneity during the lane-changing. The results of this paper can be applied to the autonomous driving assistance system and traffic flow simulation software and can shed light on the mechanism of lane-changing behavior.

Aiming at the hybrid transmission system with dual-planetary gear sets， the purely axial-torsional nonlinear dynamic model considering the internal excitations， including the comprehensive time-varying meshing stiffness and meshing error， and the external excitations， including the load torque and output torque of different power sources， is established. On this basis， the fourth-order Runge-Kutta method is applied to study the nonlinear high-frequency vibration characteristics of the system in the pure electric and hybrid driving modes with different torque excitations. The impacts of the torque allocation of different power sources on each component of the planetary gear sets are obtained. Research results demonstrate that in the pure electric driving mode with dual motors， the vibration response of the front planetary gear set is greater than that of the rear one， and the vibration fluctuation of the rear planetary row is greater than that of the front planetary row. The motor MG2 should preferentially work in the working range with large and constant torque； In the hybrid drive mode， the front planetary row is greatly affected by the motor MG1， and the vibration response amount and vibration response fluctuation range are larger than those of the rear planetary row. The motor MG1 should be kept in a relatively constant working range， and the motor MG2 should preferentially work in a larger torque range. The research results will provide a theoretical basis for the dynamic behavior analysis and torque allocation optimization of the power-split HEV based on the dual-planetary gear sets configuration.

A type of two-degree-of-freedom forced vibro-impact system with rigid and elastic compound constraints was studied. Based on the numerical calculation method of two-parameter co-simulation， the mode types and existence regions of the periodic motions of the system were obtained on the parameter plane composed of the key parameters excitation force frequency and the gap value between two masses. The transition laws between impactless motion and adjacent fundamental period motion and between adjacent fundamental period motions were revealed. The effect of the change of the elastic constraint stiffness on the types of mode and the region of existence of the periodic motions of the system was analyzed. The results show that there are two main types of transition regions， including tongue-like regions and hysteresis regions， between impactless motion and adjacent fundamental period motion and between adjacent fundamental periodic motions. There are regular subharmonic motions in the tongue-like regions. These subharmonic motions and the adjacent fundamental period motions form a hysteresis region group near the boundary lines of the tongue-like regions. Increasing the stiffness value at the elastic constraint will significantly increase the existence regions of quasi-periodic motions and chaotic motions in the low frequency domain in the two-parameter plane， and divide the existence regions of periodic motions.

In order to investigate the influence of milling parameters on the milling force and material removal rate when machining TC4 titanium alloy material and to find the best combination of parameters， side milling experiments designed based on the orthogonal experimental method were carried out with a four-edged ball-ended milling cutter. Milling speed， feed per tooth， milling depth and milling width were chosen as variables， while milling force and material removal rate were selected as evaluation indicators. The influence laws of machining parameters on milling force and material removal rate， based on the range analysis method， were found out. The grey relational analysis （GRA） and particle swarm optimization （PSO） algorithms were adopted to optimize the milling parameters respectively， meanwhile a milling force prediction model was established based on the regression analysis method to prepare for the PSO. Finally， the two optimization methods were compared and analyzed and the prediction model was verified through the verification experiments. The results show that the prediction model established in this paper can accurately and efficiently predict the milling force， and the optimization effect of PSO is much better.

The traditional one-dimensional beam research method cannot describe the complex time-varying transient turbulent flow in the working cavity of a high power density torque converter. RANS， LES， DES and SBES models are efficiently coupled with fine hexahedral grids and dynamic physicochemical properties of media by setting turbulence models in CFD calculation， which provides guidance for transient simulation of hydraulic torque converter （TC） to predict its external performance and complex internal flow field distribution. The stress-blended eddy simulation （SBES） method in dynamic hybrid rans-les （DHRL） was found by qualitative and quantitative analysis of the flow structure inside the hydraulic torque converter and the sequential evolution of the quasi-ordered vortex system in the guide vane boundary layer. The boundary layer flow in the working chamber can be fully identified， and the multi-basin coupling complex flow phenomenon can be accurately captured. The maximum error of the original characteristic prediction results is less than 4% through the bench experiment. In addition， the mechanism of flow loss generation in the working chamber is also elucidated， and the flow mechanism of vortex structure generation， development， transport， fragmentation and synthesis is revealed， which provides a calculation method for the efficient development of new products and the improvement of original products.

In-situ （Mg₂Si+Si）/Al composites were prepared by using melting casting with different solidification rates. The microstructure and wear resistance of （Mg₂Si+Si）/Al composites were investigated. The results show that， after P modification， the primary Mg₂Si transformed to polygonal blocks， while the primary Si was mainly complex morphology. With increasing the solidification rate， the number of primary Mg₂Si and primary Si particles increased and their particle sizes decreased. Extraction tests showed that the Mg₂Si crystals had tetrahedral and hexahedral morphologies. Dry sliding wear behaviors of （Mg₂Si+Si）/Al composites with different solidification rates against 45# steel， under conditions of different sliding speeds and loads， were investigated. The results show that， the wear resistance of （Mg₂Si+Si）/Al composites increased with the increasing of the solidification rate. The wear mechanisms of （Mg₂Si+Si）/Al composites are mainly adhesive wear and abrasive wear.

Focusing on the issue of mixed traffic flow composed of autonomous vehicular platoons and human-driven vehicles， and considering the inherent dynamics of vehicular platoon and the coupling with traffic flow， this work aims to investigate the property of mixed traffic flow when accounting for the actual dynamics of vehicular platoon. A model describing the relationship between the randomization probability of platoon and the variables dominating platoon dynamics including time headway and platoon size is proposed. Then a cellular automata （CA） mixed traffic flow model mixed with vehicular platoon and human-driven vehicles under open boundary condition is established. It is demonstrated from numerical simulations that the new model taking account the platoon inherent dynamics presented in this work， can reveal more particular phenomena of mixed traffic flow which cannot be reflected by the existing models. The characteristics of mixed traffic flow when considering the inherent dynamics of vehicular platoon is generally considerably affected by the typical features of platoon such as time headway， platoon size. Also， with the variations of vehicle density， more complex or even nonlinear relationship phenomenon has been observed. Generally， the effects of time headway and platoon size on mixed traffic flow are twofold， and increasing platoon penetration and platoon size can obviously improve the property of mixed traffic flow， and this advantage is more obvious in relatively high vehicle density scenarios to enhance the travelling capability.

To address the problem that distracted driving detection using classification methods can only identify a limited number of distracted driving behavior categories and ignore temporal information， we propose a distracted driving detection method based on improved YOLOv5. First， the Ghost module is introduced on the basis of YOLOv5， and linear transformation is used instead of partial conventional convolution for feature extraction to lighten the network model to achieve fast and accurate detection of cell phone， water bottle， driver's eyes and head region in the image； second， after obtaining the target detection results， the logic algorithm is designed to detect the presence of distracted driving in each frame by combining with head pose estimation. Second， on the basis of obtaining the target detection results， a logic algorithm is designed and integrated into YOLOv5 with head pose estimation to detect the presence of distracted driving in each frame from both cognitive distraction and visual distraction perspectives， which avoids the problem that the classification method is limited by the number of distracted driving categories， and then setting an appropriate time threshold， thus realizing real-time and effective distracted driving detection； finally， three sets of experiments are conducted on the collected driving behavior dataset of 18 drivers to verify the feasibility and effectiveness of the method in this paper.

At present， there is a lack of technical standards and theoretical methods in heterogeneous sensors deployment for road traffic accident detection and prevention. Therefore， this paper presented a deployment method of heterogeneous sensors for accident detection and prevention. Considering the uncertainty of traffic accidents， historical traffic accident data was used to obtain the stable spatial distribution of road accident risk. Then sensors deployment problem was converted into the optimal coverage problem on road traffic accidents risk. The optimal deployment model of heterogeneous sensors was established to maximize the coverage quality with the constraint of sensors deployment cost， accidents detection error， and so on. International open accidents data and road scenarios were used to validate the proposed method. The results show that the sensors layout scheme can effectively match with road accidents risk and achieve the expected coverage performance under different cost. Therefore， the proposed method can meet the requirements of sensors deployment under accidents uncertainty， and realize the optimal layout of heterogeneous sensors.

Understanding the travel behavior characteristics of spectators is the key to solve the problem of the surge in traffic during the event and ensure the normal operation of urban traffic. Combining the 954 survey samples of the individual attributes， travel characteristics， spectators behavior characteristics and travel intention of the spectators， a structural equation model （SEM） can describe the causal relationship between four latent variables and eleven obvious variables from the perspectives of safety， convenience， economy and travel intention. Considering the potential impact of personal preference in the combined departure time decision and travel mode chosen behavior， a SEM Nested Logit（SEM-NLogit）model is developed to investigate the combined decision-choice behavior of spectators. The results show that the developed SEM-NLogit model has the best fit and accuracy， and the prediction accuracy is 87.06%， which is 6.99% and 18.18% improvement compared to the NLogit and multivariate logit models respectively. Moreover， the findings show that the elders prefer to departure early， more companions， higher income， more likely to choose a car to travel， and the spectators more emphasis on safety and economy when traveling， which is the findings that NLogit model cannot obtain， but it is more in line with the actual travel law.

In this paper， we analyze the influencing factors of highway traffic accidents from four aspects： road alignment conditions， traffic operation status， weather environment conditions and aggressive driving behavior， using a random parameter logit model and focusing on the unobserved data heterogeneity. The results showed that plane alignment， longitudinal alignment， congestion index， speed coefficient of variation， weather conditions， wind level， frequency of aggressive acceleration behavior and frequency of aggressive deceleration behavior all had significant effects on freeway traffic accidents； meanwhile， unobserved heterogeneity was found in the four variables of longitudinal alignment， congestion index， frequency of aggressive acceleration behavior and frequency of aggressive deceleration behavior， and the frequency of aggressive deceleration behavior and congestion index， frequency of aggressive There were interactions between the frequency of rapid acceleration behavior and the frequency of rapid deceleration behavior. The paper provides a scientific and reasonable explanation for the causes of highway traffic accidents and provides theoretical support for the scientific management of highway accidents.

Aiming at the problems of folding， tail-flicking and rollover of the semi-trailer train in the dangerous section of the mountain highway in the cold area， taking a domestic semi-trailer train as the research object， the dynamic mathematics of the six-degree-of-freedom semi-trailer train is established based on the vehicle dynamics theory model. Using Truck-sim to complete the establishment of the semi-trailer train vehicle model and dangerous road section model， according to the principle of semi-trailer train rollover and yaw instability， determine the judgment conditions for semi-trailer train rollover and yaw instability. Taking the difference between the ideal state parameters and the actual state parameters of the semi-trailer train as the control target， a control strategy for the anti-rollover and yaw instability of the semi-trailer train based on the tinear quadratic regutator （LQR） control algorithm and differential braking is established with the help of Simulink. For a typical dangerous section of the expressway， changing the road adhesion coefficient corresponds to the road surface under different actual conditions. The scene reconstruction is completed by the co-simulation of Truck-sim and Simulink， and the simulation test of the proposed control strategy is carried out. The results show that： on the S-shaped curve of the K398-K406 section of the Hamu Expressway， when the road adhesion coefficients is 0.1， 0.25 and 0.35， the maximum value of the tractor roll angle can be reduced by 0.05°， 0.04° and 0.1 °. The maximum value of the trailer roll angle decreases by 0.04°， 0.01° and 0.1°. When the road adhesion coefficient is 0.35， the decrease range is the largest， which are 40% and 23.26% respectively. On the curved slope of the K273-K274 section of Hamu Expressway， when the road adhesion coefficient is 0.1， the control strategy effectively prevents the vehicle from rolling over. When the road adhesion coefficient is 0.25 and 0.35， the maximum value of the roll angle is reduced by 0.04° and 0.02° under the control strategy， corresponding to reductions of 16% and 3.92%. The control strategy can effectively reduce the peak value of lateral acceleration， yaw rate and roll angle of the tractor and semi-trailer， especially for the control of roll angle， which can avoid the danger of rolling over when the semi-trailer train is towed on the curved road in icy and snowy weather.

To enable a reliable analysis of the residual behavior of corroded rebars， a computationally-efficient method capable of considering the corrosion morphology characteristics was proposed by the probability distribution function of the cross-section area. The degraded mechanical properties of rebars under different corrosion rates were quantitatively studied based on the combination of Monte Carlo method and numerical tensile tests of corroded rebars. This method can accurately estimate the mechanical properties of corroded rebars verified with previous test results， especially for the deformation capacity. The effects of constitutive laws and gauge lengths were analyzed respectively. The results show that the strength yield ratio of rebars is the dominant issue governing deformation capacity after being corroded， and different gauge lengths are responsible for the different calculation results of the ultimate strain of corroded rebars.

Based on the trial design of the actual continuous prestressed concrete jointless bridge structure， the new type of integral jointless bridge with arch structure behind abutment was proposed. The three-dimensional beam-truss calculation model and the solid finite element model of the new structure was established. The effect of the interaction between the pile foundation and the soil around the pile was considered， and the half bridge organic glass model test was designed according to the principle of similarity ratio to verify the correctness of the simulation method. The verified numerical model was used to establish the full bridge beam-truss model and the full bridge solid model for the arch structure behind the abutment， and the effect of rise-span ratio and pile length on the overall mechanical performance was analyzed. The analysis results show that： The displacement and stress values of the model test are 8% and 21% different from those of the beam-truss model， and -3% and 5.3% different from those of the solid model. With the increase of rise-span ratio， the horizontal constraint on the arch foot decreases， the vertical displacement of the arch crown decreases relatively， the axial force on the main beam decreases gradually， and the bending moment variation range also decreases slightly， the horizontal force generated by the temperature is more transmitted to the arch structure behind the abutment. With the reduction of the pile length， the vertical displacement of the arch crown decreases， the shear force at the pile top decreases relatively， the axial force on the main beam decreases gradually， and the range of bending moment changes slightly decreases. Considering the actual foundation conditions， the jointless bridges with shorter pile length and larger rise-span ratio has better overall mechanical performance.

In order to simulate the mechanical properties of reinforced concrete（RC） columns within nonlinear state， In this paper， an modified Bouc-Wen model is proposed to improve the Bouc-Wen-Baber-Noori （BWBN） model from the aspects of stiffness degradation， pinching effect and asymmetric characteristics. In the meantime， an identification method of model parameters is established by combining parameter normalization processing and chaotic particle swarm optimization algorithm. Furthermore， secondary development for modified Bouc-Wen material is presented in OpenSees. Afterwards， the proposed material is applied to describe the moment-rotation relationship of plastic hinge of RC columns and numerical modeling of RC columns based on concentrated plastic hinge is completed. According to the low-cyclic loading test data of RC columns， the hysteretic characteristics of RC columns are described by directly identifying the horizontal force-displacement curve and numerical modeling based on lumped plastic hinge model. The results show that the accuracy of the modified Bouc-Wen model is significantly higher than that of BWBN model. The calculation accuracy of fiber model depends on the accuracy of components information， which fails to simulate the nonlinear response of RC columns after performance change. The numerical modeling method based on the modified Bouc-Wen model can rationally simulate the degradation， pinching effect and asymmetric characteristics of RC columns after earthquake damage.

The extreme stress data is taken as a time series， an improved Gaussian mixed particle filter（IGMPF） dynamic prediction new approach of bridge extreme stresses is proposed. Firstly， the dynamic nonlinear model， which provides state equation and monitored equation for the particle filter， is built with the monitored bridge extreme stress data； then， the EM algorithm is introduced to estimate the probability density function（PDF） of the target state and embedded in the Gaussian mixed particle filter（GMPF）； further， with the IGMPF prediction approach， structural stresses are dynamically predicted based on the monitored extreme stress data； finally， the monitored stress data of an actual bridge is provided to illustrate the feasibility and application of the proposed models and methods. The result shows that the proposed algorithm has good prediction accuracy， can apply to real engineering.

To reveal the coordinated deformation mechanism of temperature shrinkage between unbound aggregate base layer and asphalt mixture overlay， the dynamic and static strain acquisition system was adopted to obtain the thermal contraction strain of asphalt mixture layer in real time. And the continuum-discrete coupling model of composite structure was constructed based on PFC6.0 Suite and FLAC3D software. The macro- and meso-response law of unbound aggregate layer and asphalt mixture overlay were investigated. The results show that the continuous-discrete coupling model has higher consistency with the laboratory test than the continuum model， and the relative error of temperature shrinkage coefficient is 8.1%. The thermal strain-time curves of asphalt mixture and its composited specimens exhibit a nonlinear change law of first fast and then slow. And the asphalt mixture types and cooling temperature differences have little effect on the constraint action of unbound aggregate layer. The temperature shrinkage deformation occurred within the first hour， where the asphalt mixture layer gradually changed from "warp" to "gentle". The coordinated deformation between unbound aggregate base layer and asphalt mixture layer can be realized by particle contact recombination， the inwardly extruded movement of particles， loose on both ends and the middle compaction. In this paper， the coordinated deformation mechanism of temperature shrinkage was revealed from the macroscopic strain response， voids， coordination number and three-dimensional fabric change. A theoretical foundation for researching low temperature cracking resistant of asphalt pavement with unbound aggregate base layer can be improved.

In view of the one-sided problem in the current research on the bearing capacity of seismic joints with rigid connections between beams and columns， a calculation method for the bearing capacity of seismic joints with rigid connections between beams and columns in multi story buildings is proposed. This method is used to design three types of seismic joint core area of rigid connection between beam and column： only adding steel fiber， only adding stirrup， and adding steel fiber and stirrup at the same time. Nine specimens are made according to the different volume ratio of steel fiber and stirrup ratio. With the nine specimens as the experimental object， after designing the test loading scheme and the distribution of measuring points， the seismic joint specimen model of rigid connection between beam and column of multi story buildings is constructed using finite element software， The model is used to simulate its test loading scheme. At the same time， the calculation of seismic joints of beam column rigid connection of multi story and high-rise buildings is realized by calculating the bending capacity of T-shaped steel， bolt capacity and column flange capacity of the specimens of seismic joints of beam column rigid connection of multi story and high-rise buildings. The experimental results show that the calculation results of the bearing capacity of the steel webs of the seismic joints of beam column rigid connections by this method agree well with the actual bearing capacity； It can effectively calculate the shear bearing capacity and bending bearing capacity of specimens with steel fibers and stirrups added in different core areas， and at the same time， it can clearly show the distribution of different bearing capacity of seismic joints of beam column rigid connection by using the finite element cloud diagram.

Aiming at the problems of stereo image copyright protection and the poor robustness of existing zero watermarking schemes against geometric attacks， a novel zero watermarking algorithm against geometric attacks is proposed for color stereo image based on Tucker decomposition and Fractional-order Jacobi-Eourier Moments （FJFM）. Firstly， the R， G and B channels of the left and right views of the original color stereo image are separated respectively. Then， the separated six channels are combined and treated as a third-order tensor for Tucker decomposition to obtain the first energy map containing the correlation between the left and right views. Secondly， the mixed low-order moment feature （MLMF） of the mean sub-image obtained by 4×4 non-overlapping blocking to the first energy map is calculated based on FJFM. Then the binary robust feature matrix is constructed according to the MLMF. Finally， the final authentication zero watermark is generated using the exclusive-or operation between the watermark encrypted by the double scrambling and encryption methods and the feature matrix scrambled using the proposed chaotic system. Experimental results show that the proposed algorithm shows strong robustness to the common non-geometric attacks such as adding noise， filtering and compression， and the NC values of watermark detection are more than 0.98. In addition， the proposed algorithm is overall superior to the comparison algorithms in resisting rotation， scaling， translation and the cropping geometric attacks.

In view of the problems of semantic sparseness and lack of integration of prior knowledge in the existing methods of psychological assessment using social media data through computer technology， a psychological assessment method based on heterogeneous graph network is proposed. A heterogeneous graph is constructed by extracting users' profile information， text information， and topic information on social media， and the psychological assessment problem is described as a node classification problem. On the one hand， building a knowledge graph of social media users' mental state aims to integrate prior knowledge and represent the users' profile information； On the other hand， constructing a heterogeneous graph to integrate heterogeneous information and make semantic complements， and the hierarchical attention mechanism is used to pay attention to the node weights to obtain the evaluation results. Experiments are carried out on the Twitter and Reddit dataset， and the results show that the model is effective in classifying users' mental states， and the key performance is significantly improved.

To address the problem of lack of multi-scale feature extraction in existing skin lesion image segmentation， which leads to lack of detailed information and incorrect segmentation of skin lesion regions， this paper proposes a fusion multi-scale Transformer encoder-decoder network skin lesion segmentation algorithm.First， a hierarchical encoder is constructed using Transformer Block， which analyses the skin lesion region from the perspective of global feature variation at multiple scales. Then， the multi-scale fusion module， channel attention module and concat layer are used to construct the fusion decoder. The multi-scale fusion module fuses shallow network information and deep network information in the hierarchical encoder to enhance the dependency between spatial and semantic information， and the channel attention module can effectively identify channels containing rich feature information and improve the segmentation accuracy of the algorithm. Finally， an expansion module is introduced to recover the image size to meet the practical requirements. The proposed algorithm was experimentally tested on three public datasets， ISBI2016， ISBI2017 and ISIC2018. The pixel accuracies were 96.70%， 94.50% and 95.39%， respectively， and the mean intersection over union were 91.69%， 85.74% and 89.29%， respectively， with the overall performance of the tested algorithms outperforming existing algorithms.Simulation experiments show that the multi-scale Transformer encoder-decoder network can effectively segment skin lesion images， providing a new window for the diagnosis of modern skin diseases.

In the imaging process of optical waveguide， the center of the array is prone to shift， and the detection accuracy of the center shift of the phased array is low. For this reason， a method to detect the center deviation of integrated optical waveguide phased array is proposed. The optical wave field of integrated optical waveguide is modeled by the beam propagation method， and the scanning characteristics of integrated optical waveguide are analyzed according to the grating diffraction theory. When the center of the phased array is in the deviation state， the phased array planes are not parallel in the integrated optical waveguide. On the basis of the optical autocollimation method and the scanning characteristics of the integrated optical waveguide， the parallelism of the phased array in the integrated optical waveguide is calculated to obtain the parallelism of the phased array plane in the x and y directions， and the deviation degree of the integrated optical waveguide phased array center is detected based on this. The experimental results show that the detection errors of the proposed method in the x and y directions are 0.001 and 0.002， and the detection results are consistent with the actual results， which proves that the proposed method has high detection accuracy.

Due to the dense arrangement and different directions of objects in remote sensing images， the existing detection algorithms are difficult to locate the instance objects accurately. Therefore， this paper proposed an omnidirectional accurate detection algorithm for dense small objects in remote sensing images. In order to improve the ability of feature extraction， the meta-ACON activation function was introduced into the residual structure of the backbone network to adaptively learn the importance of channel features. A new strengthen tconnection feature pyramid network was proposed. The lateral connection part for deep and shallow feature fusion was redesigned， and a jump connection was added between the input and output of the same level feature map to enrich the feature semantic information. The angle prediction branch was introduced， and the circular smooth label method was used to transform the angle regression problem into a classification problem. While realizing the rotation of the object frame， the problem of sudden change of the boundary of the rotation frame was solved. A post-processing method （Rotate-Soft-NMS） for rotation detection frame was designed to remove adjacent repeated rotation detection frames by suppressing the confidence of the detection frame. The experimental results on DOTA dataset show that the mAP of the proposed algorithm is 76.15%， which is 5.22% higher than the benchmark model YOLOv5m， and has achieved the best detection results compared with other advanced algorithms. The algorithm in this paper has achieved better detection results for complex remote sensing images.

The coordinated control of compound power supply is an indispensable step in the normal driving process of hybrid electric vehicles. However， the cooperative control process is easily affected by problems such as battery performance， current effect， and vehicle speed， resulting in unstable current distribution of the composite power supply. Therefore， a hybrid electric vehicle hybrid power cooperative control method considering the state of charge of the battery is proposed. The method first obtains the estimated value of the battery state of charge by constructing a second-order RC equivalent circuit model， then calculates the efficiency characteristics of the hybrid power supply based on the estimated value， and finally uses the adaptive filter power allocation control algorithm to realize the power sharing between the supercapacitor and the battery by constructing the objective function and constraint conditions， thus completing the cooperative control of the hybrid power supply of the hybrid electric vehicle. The simulation results show that the current distribution of the proposed method is good， the energy consumption of the controlled vehicle is low， and the energy consumption is always controlled within the range of 0.6 ~2.7 kW/h.

Aiming at the challenges of unnecessary leap and significant undulations terrains with large steering angles in path planning of multimodal quadruped robots by Rapidly Exploring Random Tree algorithm， a path planning algorithm solution based on discrete sampling is proposed. The path is preprocessed to remove unnecessary leap and a solution set is obtained by discrete sampling and dynamic programming method. B-spline curves are used to define spline segments and quadratic programming method is used to optimize the final path. The simulation results show that paths planned by the proposed method exhibit an average reduction of 31.4% in the adjustment of robot's center of mass height， a 13.4% decrease in undulation of terrain， an 11.4% reduction in terrain slope angle and a 62.7% reduction in steering angle， which affirm the effectiveness of the proposed method.

Microgrid with renewable energy system has high economy and environmental protection， and requires system load to be flexible and controllable， that is， resources can be coordinated for scheduling operation. Therefore， a collaborative optimal scheduling method for renewable energy system based on improved MOEA/D algorithm is proposed. Build key power generation models in renewable energy systems， including photovoltaic array model， wind power generation model and energy storage battery model. With the lowest energy consumption as the goal， build the system's collaborative optimal scheduling function， obtain the functional constraints of electric power and thermal power balance， and use the improved MOEA/D algorithm to solve the optimal solution of the function to achieve the collaborative optimal scheduling of renewable energy systems. The experimental results show that the proposed method has good SOC value control constraint ability， and the SOC value control result is the closest to the ideal state. Under the environmental optimization criterion， the system scheduling energy consumption is low， and the 24-hour energy consumption power is 0.6 kW.

Low density parity check code is a kind of linear block code. Due to the low content of non-zero elements， there is a problem that the node information reliability determination is not accurate enough in the decoding process. Therefore， an improved LDPC decoding algorithm based on self correcting minimum sum is proposed. The iteration rule of BP decoding algorithm is analyzed. The first minimum and the second minimum of the minimum sum algorithm are used to set the correction threshold of variable node information， and the self-corrected min-sum algorithm is improved. The order statistics theory is used to obtain the normalization factor corresponding to the two minimum values to prevent the transmission and diffusion of unreliable variable node information in the iterative decoding process. Simulation results show that when the bit error rate is 10^-5， the proposed algorithm can obtain about 0.2 dB of decoding performance gain， and the average number of iterations can be reduced by 18.2% at most， which proves that the proposed algorithm can effectively improve the decoding performance and iterative convergence performance.

To improve the adsorption performance of the sucker， the study used bionics to design the pit structure on the surface of the normal sucker. So， there were multiple small suction cups on the working surface of the sucker during the adsorption process to improve the suction performance of the sucker. Partial orthogonal polynomial regression analysis was used to explore the influence of the diameter of the pit shape， the number of single-row annularly distributed pits， and the row spacing on the adsorption force of the sucker. The test showed that the sucker with different morphological parameters on the surface had different effects on the adsorption force of the sucker. When the pit diameter was 1.5 mm， the number of single-row annular pits was 40， and the row spacing was 4 mm， the adsorption force of the bionic sucker on the substrate surface was 49.54 N. Compared with the normal sucker， the maximum adsorption force of the bionic sucker on the substrate surface was increased by 49.21%. The mathematical regression model was established between the design factors and the evaluation index. The response surface method was used to analyze the significant factors affecting the pit parameters on the adsorption force of the bionic sucker. The significant order of the influence on the sucker adsorption force was determined as row spacing， pit diameter， and the number of single-row annular distribution pits. The analysis showed that the shape of the pits on the working surface changed the distribution of contact pressure and friction stress on the working surface of the sucker. The pit shape increased the friction stress and contact pressure at the edge of the pit. In addition， the friction stress and contact pressure of the working surface of the bionic sucker were greater than that of the standard sucker， which improved the adsorption force of the bionic sucker on the surface of the substrate. This present study would help to improve the adsorption performance of the sucker by designing bionic pits on the working surface of the sucker.

In view of the problem that the arable plough could not be used for deep ploughing due to the narrow space of facility agriculture and the serious soil diseases and insect pests， a method of turning soil on the spot was put forward.Based on the reverse engineering， the reverse surface of the plough body of the local soil turning plough was obtained. Based on the existing forming principle of the plough body， the forward surface of the plough body of the share plough body was established by referring to the ploughing width， ploughing depth， installation angle of the plough body， soil cutting angle， guiding curve lifting angle， opening and length of the plough body.The soil relative humidity was 34.57% and the soil accumulation angle was 39°. The contact parameter range was obtained by GEMM， the contact parameter database of EDEM software. The surface energy between soil particles was 4.19 J/m²， the collision recovery coefficient was 0.282， the dynamic friction coefficient was 0.051， and the static friction coefficient was 0.629.The concept of soil fall rate in original gully was introduced. The interaction model between soil fall rate and surface parameters was established， and the optimal surface parameters of soil fall rate were obtained.Using Design-Expert software， plackett-Burman experimental design method was used to carry out multi-factor and two-level significance screening test with soil fall rate as the response index. Three significant factors affecting soil fall rate of the original ditch were obtained， namely， plow width， plowshare installation angle and soil cutting angle.Box-Behnken experimental design was carried out for the three factors， and the optimal design parameters of 278.392 mm plow width， 40.522°plow share installation angle and 23.211°soil cutting angle were obtained.Based on the optimized results， the error between the experimental results and the predicted results is 3.13%. The experimental results show that the tilling plow can achieve the aim of tilling in situ， and provide a reference for solving the problem of soil deep tilling in facility agriculture.

In view of the problem that the inertia force was large and the vibration was difficult to solve， the model optimization design method was studied on the compression mechanism. With the development of intelligent agricultural machinery， big square baler density of bale need controlled based on the compression force， but there was no research on the compression force time domain waveform of big square baler in China. In order to design a big square baler and test the compression force of the bale density， the minimum average acceleration of sliders in a cycle was taken as the optimal design target， the transmission angle， velocity uneven coefficient and motion non-interference was taken as the constraints. The model shows that the crank length has a greater impact on the inertial force of the crank slider mechanism， and the longer the crank， the greater the inertial force. Based on the above parameters， YTO 9YFA-220 baler was designed and the compression force was tested. The results showed that the bale density was up to 230 kg/m³. The maximum compression force was 410 kN. The compression force appears shorter times in the whole period， about a quarter of the full cycle. The compression mechanism studied is of important engineering application value for the research and development of big baler.