During the inspection of transmission lines using Unmanned Aerial Vehicle (UAV), there are many dropped anti-vibration hammers that become obstructed by wires or are shot from a distance. This challenge leads to the occlusion of target features and low resolution. In addition, the close proximity of a number of hammers due to sliding poses challenges to the accuracy of target identification. To address the above problems, a deep neural network based on an improved Cascade RCNN was proposed to identify the dropped anti-vibration hammers. The proposed network mainly achieved improvements from the following four aspects. First of all, a contrastive learning network was designed to compare the features of positive and negative samples with real samples. By utilizing a contrastive loss function during network training, the network became more attentive to the blocked dropped anti-vibration hammers and enhanced its feature extraction ability. Secondly, the classifier was enhanced. The selection of interested regions with better regression performance in the cascade structure was filtered and input directly into the final classification regression queue. This improved the classification performance of the classifier, thereby enhancing the classification scores of the detected targets. Thirdly, a parallel attention mechanism module was designed to integrate the extracted features from the network, increasing the weights of key features and directing the network’s attention to more critical features in the image. In addition, in the process of feature fusion of the feature pyramid, the bilinear interpolation method was replaced with deconvolution to enhance the feature restoration capability. The experimental results demonstrated that the improved model achieved a recall rate of 97.5%, precision of 91.0%, and average precision of 92.0%, an improvement of 6.9%, 28.4%, and 8.0%, respectively, compared with the baseline model.

Depth images contain scene depth information and exhibit strong robustness to variations in color and lighting, making them widely used in fields such as stereo vision. However, due to the limitations in depth sensor performance and the complexity of imaging environments, it is challenging to directly obtain high-quality, high-resolution depth images. To address the problem of unclear edge details in reconstructed depth images, a reference-based Transformer texture transfer method for deep image super-resolution reconstruction was proposed. For the preprocessed low-resolution depth images (LR_D) and reference images (Ref) feature blocks, similarity calculation was performed using normalized inner product. The method integrated Transformer to calculate the confidence of similarity positions, and combined it with an attention mechanism for texture transfer. Finally, the method combined the features of the low-resolution depth images to improve image detail clarity and further accurately reconstruct the results. The experimental results demonstrated that compared to other methods, the proposed method could achieve higher structural similarity (SSIM) values, and that both subjective visual effects and objective evaluation indicators have been significantly improved, indicating the excellence of the reconstruction performance.

To address the problems of existing heatmap-based human pose estimation network models, such as high complexity, intensive computing power requirements, and challenges in deployment on embedded platforms and UAV mobile platforms, a lightweight human pose estimation network was proposed based on YOLOv5s6-Pose-ti-lite without using heatmaps. By replacing the backbone network with GhostNet, it enabled the output of more effective feature information with reduced computing resources. This resulted in faster network detection and alleviated issues related to network redundancy. Within the backbone network, a lightweight coordinate attention (CA) attention module was integrated to gather the position information of human keypoints in the picture to the channel, thus enhancing the ability of feature extraction. BiFPN (weighted bidirectional feature pyramid network) module was introduced to enhance the feature fusion ability of the model and balance the feature information across different scales. Finally, the CIoU loss function was replaced with wise-IoU (WIoU) to enhance the performance of the model for human keypoint regression. The results demonstrated that on the COCO2017 human keypoint dataset, the parameters of the optimized network model were reduced by 26.2%, the calculation was decreased by 30.0%, the average precision was increased by 1.7 percentage points, and the average recall rate was boosted by 2.7 percentage points. These improvements could enable real-time performance, verifying the feasibility and effectiveness of the proposed model.

Early diagnosis and treatment are pivotal in elevating the chances of lung cancer survival. Early-stage lung cancer often manifests through lung nodules. However, their heterogeneity poses a challenge in their detection of lung nodules via computed tomography, subsequently diminishing the accuracy of segmentation results. To improve the completeness and accuracy of lung nodule segmentation results, a 3D channel residual nested U-network (CR U²Net) was proposed for lung nodule segmentation. The shallow information processing U-structure (SIPU) was proposed to address the challenge of managing the interference of noise information while simultaneously incorporating key lesion details within shallow features. To enhance the interaction across different layers of feature information, and to enrich feature expression and transfer, the Channel Residual structure was introduced in conjunction with the nested U-structure to extract and optimize feature information. Acknowledging the spatial detail information found in shallow features and the semantic abstraction in deep features, the channel extrusion U-structure (CEU) was designed to effectively fuse features at different semantic levels. By integrating the proposed modules into UNet, a lung nodule segmentation model based on nested U-structures was constructed. The proposed model was trained on the Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset. And chieved the best Dice Similarity Coefficient performance, reaching 83.83%. This outperformed UNet, UNet++, and PCAMNet networks by 3.98%, 1.96%, and 1.26%, respectively. In addition, ablation experiments were conducted to evaluate the structural validity of the proposed CR U²Net, demonstrating that each module within the proposed segmentation algorithm contributes to achieving optimal performance while adhering to acceptable parameter and computational constraints.

In response to the challenge of detecting small-scale, occluded pedestrians in dense scenes, where they are prone to being missed, we proposed an improved YOLOv8 detection algorithm. First, to address the issue of extracting features from small-scale pedestrians, a backbone network improved by deformable convolution was employed to enhance the feature extraction capability of the network, and an occlusion-aware attention mechanism was designed to enhance the visible part of the occluded pedestrian features. Second, to address imprecise localization of the detection head in dense pedestrian scenes, a dynamic decoupling head was designed to enhance attention to multi-scale pedestrian features, thereby improving the expression capability of the detection head. Finally, to address the problem of low model training efficiency, the regression loss that combined Wise-IoU with distributed focus loss was utilized for training, thereby enhancing the convergence ability of the model. Through the analysis of experimental results, the improved YOLOv8 algorithm demonstrated exceptional performance on two challenging and dense pedestrian datasets, namely CrowdHuman and WiderPerson, achieving an AP₅₀ of 90.6% and 92.3% and an AP_50:95 of 62.5% and 68.2%, respectively. In contrast to the original algorithm, the improvements were substantial, establishing robust competitiveness when compared with other state-of-the-art pedestrian detection models. The proposed algorithm exhibited a wide range of applications in dense pedestrian detection tasks.

To address the limitations of existing stereo matching algorithms in both edge regions and regions of discontinuous disparity, a deep learning stereo matching algorithm fusing structural information was proposed. By limiting the convolution kernel size and replacing the BatchNorm layer and activation function layer with the Inplace-ABN layer, the efficiency of convolution to extract image features was enhanced. The local similarity pattern module combined with an attention mechanism was employed to extract image structural features, and the features extracted by convolution were fused to enrich image feature information. The correlation volume and connection volume of the output feature were calculated. By utilizing the correlation volume to generate attention weights, the algorithm filtered out the redundant information of the connection volume and improved the accuracy of the stereo matching cost volume. In order to expedite network cost aggregation, a simplified hourglass network was employed. The algorithm was tested against the Scene Flow dataset, CREStereo dataset, and KITTI dataset. The experimental results demonstrated that the algorithm had an overall region endpoint error of 0.45 pixels. In the first frame image, only 1.55% of regions were incorrectly predicted, and merely 6.87% of pixels exhibited prediction errors greater than 1 pixel. These results demonstrated the excellent performance of the proposed algorithm compared to other algorithms in terms of matching accuracy. Furthermore, it validated the effectiveness and advantages of the algorithm in matching problematic areas.

The adaptive segmentation network represents an efficacious method for cross-domain semantic segmentation within urban scenes. However, the challenge arises from the distinct appearance distributions among cross-domain datasets, leading to domain gaps and unsatisfactory network segmentation accuracy for small targets. To address these issues, a domain adaptive segmentation method based on a dual-source discriminator was proposed. Firstly, the new source domain S' was obtained using the style translation technology FastPhotoStyle for the source domain S, thereby reducing the domain gaps at the image level. Next, the generator was employed to extract segmentation feature maps from the source domain S, the new source domain S', and the target domain T, respectively. The feature map of the new source domain served as an intermediate bridge for the channel-wise fusion between the source and target domains feature maps. The two fused feature maps were input into the dual-source discriminator, with both the dual-source discriminator and the generator undergoing iterative training. Since the discriminator input of the proposed model consists of dual-source features, it is referred to as a dual-source discriminator. The two features from the dual-source input contained similar feature information, which further reduced domain differences at the feature level. To enhance segmentation accuracy, a self-training pseudo-label was introduced. At the same time, to address class imbalance issues during training, a class balance factor was incorporated into the loss function of the target domain, thereby enhancing the network’s ability to segment small targets. Experiments on two segmentation tasks GTA5→Cityscapes and SYNTHIA→Cityscapes demonstrated the advancement and effectiveness of the proposed method.

To address the challenge of detecting tiny-size and dense occlusion objects in the task of multi-fitting detection for transmission lines, a transmission line multi-fitting detection method based on implicit spatial knowledge fusion was proposed. First, in order to mine the implicit spatial knowledge among transmission line fittings and assist the model in detection, the spatial box setting module and the spatial context extraction module were proposed to set the spatial box and extract the spatial context information. Then, the spatial context memory module was designed to filter and remember the spatial context information to aid the positioning of multi-fitting detection model. Finally, the post-processing part of the model was enhanced to further mitigate the issue of low detection accuracy stemming from dense occlusion by fittings. The experimental results demonstrated the efficacy of the proposed model in enhancing the detection of various kinds of fittings, especially those of tiny size and dense occlusion. Compared with the baseline model, the AP⁵⁰ evaluation index and the more stringent AP⁷⁵ evaluation index were increased by 3.5% and 5.7%, respectively. It laid a foundation for the application of fitting detection and further fault diagnosis.

To address the current problems of chromatic aberration and unsatisfactory defogging effects in the single-image defogging algorithm, we proposed a single-image defogging algorithm based on YUV color space. This method applied the idea of GAN image coloring task to recolor haze images from a positive perspective. The haze image was converted to the YUV color space, and the dense residual module was employed to collect the brightness features of the image from the Y channel. Additionally, the brightness information of the haze image was adjusted according to the characteristics, mitigating the impact of haze on the image. Four residual modules were used on the UV channel to extract image color information multiple times, and recolored the image through model prediction based on the extracted color information. A feature fusion network, including a skip connection structure, was utilized to fuse low-level features with high-level ones. Furthermore, the addition of an attention module during the fusion process led to more refined dehazing. The experimental results demonstrated the algorithm’s efficacy, showcasing remarkable performance in terms of RMSE, SSIM, and PSNR on the synthetic haze image datasets. On the real haze image, the algorithm displayed excellent performance on dense fog and thin fog images, ultimately leading to an outstanding defogging effect and ensuring a high level of stability.

In response to the wide distribution of defects, the variety of appearance defect features, and the challenges in extracting features during the substation instrument appearance defect detection task, the appearance defect detection algorithm of substation instrument based on improved YOLOX was proposed. Through the contrast sample strategy, the discriminative features of various appearance defects could be identified and extracted in depth. Additionally, the Global Context Feature Module was added to the network backbone to enhance the learning ability of the model for appearance defect features and improve the network performance. Finally, the SIoU loss function was integrated in the prediction part to fully consider the influence of the direction frame’s angle on model optimization, thereby improving the accuracy of detecting appearance defects in substation instrument and reducing the occurrence of missed detection and false detection. In the experiment, four typical types of appearance defects of substation instruments were selected as the experimental objects. The analysis of experimental results showed an average accuracy rate increase of 6.2% compared with the baseline model. This substantial improvement effectively enhanced the detection effect for appearance defects in substation instruments, thus creating favorable conditions for intelligent monitoring of unmanned substations.

Under snowy conditions, snowflakes can obstruct video surveillance systems, preventing the capture of important scenery information and drastically reducing the quality of acquired video images. This interference can also strongly affect advanced image processing techniques such as subsequent target detection and recognition. The existing methods for removing snow from video images commonly suffer from drawbacks such as unstable snow removal performance and long computational time. To address this issue, firstly, the advantages of tensors in fully capturing spatial location information within video images were leveraged. By combining a low-rank tensor decomposition model with three-dimensional total variation regularization, the snow-contaminated surveillance video was decomposed into a static background layer and a moving foreground layer. Then, based on the non-subsampled shearlet transform and mathematical morphology filtering methods, the moving foreground layer was further decomposed into a moving object layer and a snow layer. Finally, the static background layer and moving object layer were reconstructed to obtain snow-free video images. The experimental results demonstrated the effectiveness of this approach in removing snowflake interference from video images while clearly retaining scene edge information. Moreover, the proposed method outperforms existing state-of-the-art algorithms in terms of processing efficacy and operational efficiency.

Image inpainting has extensive applications in photo editing and removal. In order to address the limitations of existing deep learning-based image inpainting model, which is affected by the receptive field of convolution operators and results in distorted structure or blurred texture, a locally optimized generation model LesT-GAN was proposed. This model comprised a generator and a discriminator. The generator consisted of a locally enhanced sliding window Transformer module. This module combined the translation invariance and locality advantages of deep convolution with the Transformer’s ability to model global information. As a result, it could cover a wide range of receptive fields while optimizing local details. The discriminator part was a relative average discriminator based on mask guidance and patch. It simulated pixel propagation around the boundary of the missing region by estimating the average probability of a given real image being more realistic than a generated image. As a result, during the generator training, it could generate clearer local textures directly from real images. In comparison experiments with other advanced image inpainting methods on the Places2, CelebA-HQ, and PairsStreet datasets, LesT-GAN improved L₁ and FID by more than 10.8% and 41.36%, respectively. Experimental results demonstrated that LesT-GAN exhibited superior restoration performance across multiple scenes, and that it could be well generalized to images with higher resolution than those used during training.

In recent years, three-dimensional (3D) displays have garnered increasing attention for their superior immersive experience. However, the lack of 3D content poses a challenge to the development of 3D displays. To obtain scarce 3D content, two-dimensional (2D)-to-3D conversion has emerged as a promising and effective approach. The conversion involves adding extra depth information to 2D content. However, existing depth estimation methods cannot satisfy the requirements of 2D-to-3D conversion because of their instability. This paper presented a stereoscopic image presentation system, which was designed to transfer a monocular image to a pair of stereoscopic images for 3D displays while considering human perception. The core step of the system proposed an algorithm called depth optimization considering human perception (DOCHP), using semantic segmentation images as input and considering human perception, including attentional mechanisms and depth perception to enhance the stereoscopic effect of the stereoscopic images. The experimental results demonstrated that the stereoscopic images, which were generated through the deep map optimized by the system, provided users with a strong sense of 3D effect. This article demonstrated the necessity of incorporating human perceptual characteristics in the production of autostereoscopic images and bolstered the promotion and application of autostereoscopic images.

Fire incidents in tunnels present serious hazards due to their rapid spread in confined spaces, endangering lives and property while making rescue operations challenging. Current tunnel fire detection methods suffer from inaccuracies and sufficient data. To address the above problems, a tunnel fire detection method based on an improved student-teacher network was proposed. Firstly, the proposed method trained unsupervised learning on fire-free samples to detect fires, compensating for the lack of tunnel fire datasets. At the same time, the student network and the teacher network with the same structure were adopted to form the whole network structure, and an attention mechanism was added to residual blocks for knowledge distillation to reduce the loss of important information and filter irrelevant information. Secondly, a Mish activation function was employed to replace a Relu activation function to enhance network performance. Finally, the SPD-Conv module replaced the strided convolution and pooling layer to improve the detection accuracy in smaller fire areas. The experimental results demonstrated that the pixel-level AUC-ROC and image-level AUC-ROC of the improved student-teacher network in the self-made tunnel fire dataset reached 0.93 and 0.82, respectively. Compared with the current tunnel fire detection algorithms, the detection accuracy of the improved model was higher than other models, substantiating its effectiveness.

Online shopping has significantly enhanced people’s lives. However, the challenge for consumers when purchasing glasses online arises from the lack of opportunity to physically try on glasses and directly visualize the results of the try-on, a convenience readily available in physical stores. To tackle this issue, this paper proposed a method for virtual glasses try-ons using depth cameras. The method comprised two phases: ① 3D face model reconstruction: This phase utilized the multi-angle face point cloud collected by the depth camera. It involved a two-step process for point clouds registration. First, a coarse registration was performed using feature point clouds. Next, the overlapping parts of the point clouds were extracted for fine registration, ultimately yielding a 3D face model from the registered point cloud. ② Glasses try-on: In this phase, the face model and the glasses model were automatically aligned. The model was then transformed in response to the real-time facial pose parameters estimated based on the video stream. Afterward, the models were rendered, and the glasses model was fused into the video stream, following the occlusion culling procedure, resulting in a realistic virtual glasses try-on effect. Experimental results validated the algorithm’s excellent real-time performance and its ability to provide users with an immersive try-on experience.

In visual, augmented, or mixed reality applications, real-time acquisition of user and object poses is a prerequisite for building a highly immersive virtual environment. With the continuous development of virtual reality technology, users’ demands for the range of motion in virtual environments have been increasing. They are no longer content with limited movement within the confined space of a single room; instead, they seek to roam and interact in a larger range of environments. Most of the tracking systems used by popular AR/VR devices today are designed for room-level or even smaller range tracking. When larger range tracking is required, these systems either introduce greater error drift or require more hardware to be arranged in the room to cover a larger area (e.g. Light House), which creates huge hardware costs and a complex configuration process, making them not suitable for general and personal use. To address this, a system for optical positioning and tracking was proposed, which could achieve accurate 3D camera tracking by deploying a small number of infrared LED markers on the ceiling or floor. The proposed tracking system utilized the most basic dot and line elements to build the landmark pattern. Compared with traditional marker-based systems, individual dots do not contain any information and are identified only after they are formed into a basic graphic element with a line next to them. The straight line segments exist to increase the redundancy of the basic graph elements, thus avoiding the situation where the dots are obscured and cannot be recognized. By designing the encoding principle of the marker patterns, employing the layout repeated feature retrieval method, and implementing the corresponding points matching algorithms, the fast and accurate decoding of the landmark images was realized. Experiments have proven that the system could achieve the position accuracy at the millimeter level. In robustness experiments, the proposed method could maintain higher recognition accuracy even in the presence of challenges such as large inclination angles and marker point occlusion. These measurements show the potential of our system to cope with more extreme situations. We also count the processing time of the system, and the average latency of our method is 4.34 ms, which indicates that performing sparse graph element layout and simplifying marker point decoding effectively reduces the system computation time. The resulting tracking system possesses characteristics such as low cost, easy scalability, and resilience to occlusion, thereby meeting the demand for real-time tracking and positioning at the 100 square meter level.

In light of the development of advanced design methods and additive manufacturing technologies, the design and manufacturing of multi-scale structures have been extensively investigated. Extensive research has been conducted on multi-scale structural optimization with both variable density lattices and topologically freeform micro structures. Their additively manufactured counterparts have undergone testing, showcasing excellent mechanical properties, as reported in a vast number of publications. Recently, by introducing solid phase into the design, it is found that the resulted mechanical performance of the multi-scale structures can be further enhanced. The lattice-solid hybrid multi-scale structures possess lightweight, high-performance, and multifunctional characteristics, demonstrating great potentials in applications and scientific research. Hence, in this paper, a topology optimization method on lattice-solid hybrid structures for support-free additive manufacturing was proposed. By proposing a novel hierarchical material interpolation model, this method defined two sets of densities: the external densities defining the overall material distribution for support-free effect and the internal relative densities defining the local structural details for lattice-solid phase interpolation. PDE filter is applied to smooth density distribution and thus eliminate the check board issue. Heaviside projection is adopted to create the clear cut structural boundary, preventing hard to process grey elements. Additive manufacturing filter was applied to the external densities to ensure the support-free capability by restricting the overhang inclination angle beyond 45 degree. The effective elastic properties of the lattice-solid materials were evaluated through numerical homogenization, and polynomial fittings were performed to establish the surrogate models of the equivalent elastic matrix components on the internal relative densities. One important characteristic of the above interpolation was that the material option could automatically switch between the lattice and solid phases, thus enabling the smooth gradient-based design optimization. Then, the optimization problem is formulated similar to the density based topology optimization scheme and sensitivities of both the objective and constraint functions are derived with the adjoint method. The method of moving asymptotes are adopted for design updating the two sets of density variables. Finally, the effectiveness of the proposed method was verified through a numerical case study. The numerical result indicated that the support-free lattice-solid hybrid structure outperformed the conventional single-scale self-support topological design in terms of load-bearing capability.

Ensuring a balance between information supply and information demand among all stakeholders at each project stage is crucial for harnessing the full potential of building information modeling technology in project life cycle management. While many standards have been established at national, local, and corporate levels to standardize delivery requirements, there remains a lack of efficient model integrity checking approach to facilitate participants to deliver project information in a timely and effective manner. Currently, the conventional means of checking model information for each stage of delivery was manual and machine hard-coded. However, they suffered from high labor costs, missed judgments, and limited flexibility for updates. Considering the technical advantages and applications of the semantic web concept and related technologies in knowledge organization and management, and combining the results of existing digital delivery standards for power grid engineering, a semantic web-based BIM model integrity checking approach was proposed. It was composed of three parts: 1) establishing integrity checking ontology and model to-be-checked ontology using SKOS; 2) instantiating knowledge in the delivery standard and actual model information to the knowledge graph; 3) correlating the two through specific relationships and automating integrity checking on this basis through SPARQL. Finally, the validity of this method was demonstrated through case studies, showcasing its applicability in engineering project delivery integrity checking in other fields.

Throughout the entire lifecycle of a project, changes in requirements, participants, and operating conditions often result in frequent modifications to BIM models. The workload for model review is significant, and the foundation of this review process lies in identifying the differences between different versions of the model. For structured information in the model, a text-based comparison approach could be utilized. In terms of unstructured geometric information, a method of shape context description based on contour key points and surface description points was proposed in the paper. Firstly, this method obtained the model contour line through calculation, and then collected the model feature descriptions through uniform sampling of contour lines and uniform sampling of surface Poisson disks. Finally, a quantitative comparison was conducted on the statistical feature of the model. Compared to the current mainstream random sampling methods, the proposed method not only provided a stable description of the model with fewer sampling points, thus improving the contrast efficiency, but it was also highly sensitive to changes in the model structure. At the same time, it could resist the impact caused by curved surface subdivision or simplification. The proposed method was experimentally validated on some non-standard components of a hydropower project, demonstrating outstanding performance in terms of model differentiation comparison.

A method for migrating the Chengcheng embroidery style based on extension semantics was proposed to address the lack of innovation in Chengcheng embroidery patterns and limitations in automatic generation. The Chengcheng embroidery patterns were sorted and their semantics were extracted. This was followed by an analysis of the relevance of semantic vocabulary and design application value, thereby selecting the vocabulary with the largest extension interval. The embroidery semantics were then graphically illustrated to guide pattern design, which was evaluated in terms of semantic expression and aesthetics. Finally, the optimized pattern was input into a style transfer algorithm to migrate the Chengcheng embroidery style to the original design pattern. This process encompassed the computation of the peak signal-to-noise ratio (PSNR), an assessment of image quality, and a goodness evaluation, resulting in an innovative design scheme for the Chengcheng embroidery style. This method can transform the style characteristics of Chengcheng embroidery into semantics and apply them to innovative embroidery design, providing a means to protect and innovate cultural heritage.

By employing the parametric mathematical logic features, this study explored a rapid design method for personalized 3D-printed external fixator morphology. Through the analysis of design elements of 3D-printed external fixator morphology, three morphological design element variables were summarized and converted into parameter variables. The logical relationship was constructed between the parameter variables to obtain a personalized 3D-printed external fixator design program. Secondly, finite element strength analysis and program efficiency tests were conducted on multiple schemes generated by the parametric program. The results of the program efficiency test indicated that this program could quickly generate personalized 3D-printed external fixators and efficiently optimize the personalized morphology of the 3D-printed external fixator. According to the finite element analysis results, all the multiple 3D-printed external fixator schemes generated by this program met the requirements of mechanical strength. The modeling program written through parameterization enabled the morphological design of personalized 3D-printed external fixators. It simplified the design process of 3D-printed external fixators, allowing for rapid optimization of morphological design and improving design efficiency.

In response to the pressing issue of insufficient rescue equipment in snowy environments, a functional requirement index system for intelligent snow rescue equipment was established to achieve a quantitative evaluation of design requirements for snow rescue equipment. The AHP-TRIZ theory was explored to provide a theoretical basis for the conceptual design of snow rescue equipment. Extensive research was conducted from the perspective of product functional requirements, which involved obtaining user requirement elements through literature research and survey methods. The analytic hierarchy process (AHP) was then employed to summarize the functional requirement indicator system for snow rescue equipment from three levels: objectives, criteria, and indicators. Utilizing the 1-9 scale method, the design requirements of snow rescue equipment were subjected to quantitative evaluation, and consistency testing was completed. Additionally, by integrating the Archishuler contradiction matrix theory from the Theory of invention problem solving (TRIZ), the study identified the parameters requiring improvement and corresponding deterioration. It further explored efficient solutions for 39 engineering parameters and 40 invention principles in TRIZ, resolving conflicts between multi-level design elements and achieving the optimal configuration. Thoroughly identifying the requirements for snow rescue equipment in terms of functionality, appearance, ergonomics and technology, etc., the study successfully accomplished the conceptual design of snow rescue equipment. In response to the challenges posed by harsh environments in snow-covered areas, the subjective judgment results were quantified and processed through the integrated AHP-TRIZ method. This approach not only provided methodological guidance for the design of snow rescue equipment but also enhanced the scientific and rational nature of the design.