Down-sampling Operations Research Articles

Progressive Transformer Machine for Natural Character Reenactment

Character reenactment aims to control a target person’s full-head movement by a driving monocular sequence that is made up of the driving character video. Current algorithms utilize convolution neural networks in generative adversarial networks, which extract historical and geometric information to iteratively generate video frames. However, convolution neural networks can merely capture local information with limited receptive fields and ignore global dependencies that play a crucial role in face synthesis, leading to generating unnatural video frames. In this work, we design a progressive transformer module that introduces multi-head self-attention with convolution refinement to simultaneously capture global-local dependencies. Specifically, we utilize the non-lapping window-based multi-head self-attention mechanism with hierarchical architecture to obtain the larger receptive fields at low-resolution feature map and thus extract global information. To better model local dependencies, we introduce the convolution operation to further refine the attentional weight in the multi-head self-attention mechanism. Finally, we use several stacked progressive transformer modules with the down-sampling operation to encode information of appearance information of previously generated frames and parameterized 3D face information of the current frame. Similarly, we use several stacked progressive transformer modules with the up-sampling operation to iteratively generate video frames. In this way, it can capture global-local information to facilitate generating video frames that are globally natural while preserving sharp outlines and rich detail information. Extensive experiments on several standard benchmarks suggest that the proposed method outperforms current leading algorithms.

Wavelet integrated attention network with multi-resolution frequency learning for mixed-type wafer defect recognition

Wafer defect recognition has been an important measure in developing the manufacturing process. In real-life manufacturing, however, defects can be complicated, and their wafer maps are often accompanied by noise. This encourages us to build a noise-robust framework with outstanding performance and sound interpretability on defect recognition. Therefore, this paper chooses discrete wavelet transform for its clear physical meaning and ability of frequency learning. Moreover, it outperforms traditional down-sampling operations in the preservation of fringe information. Based on this, we propose a multiresolution wavelet integrated attention network (MRWA-Net). Specifically, we design a learnable discrete wavelet transform layer (DWT-Layer), which expands convolutional neural network’s (CNN’s) feature learning space to the wavelet domain. This helps the framework procure hidden information from different frequency components and their location information. Furthermore, we utilize different levels of wavelet transform to interpret the images with different resolutions, thus learning features from different perspectives. Additionally, we insert a frequency-location attention module (FLA) to select the useful frequency-location information captured by DWT-Layer. The proposed approach is evaluated on a dataset with 38015 subjects and 38 types of defects and reaches 98.84% accuracy. To demonstrate the noise-robustness of our framework, we further compare it with other state-of-the-art methods on wafer maps with different ratios of additional noise. The results show that our framework excels other methods under all noise ratios and exhibits more notable excellence on data accompanied by a higher ratio of noise. Finally, we present visualizing analysis to demonstrate that the proposed DWT-Layer can learn from different frequency bands and retrieve information with multiple resolutions.

Defense Against Adversarial Attacks with Efficient Frequency-Adaptive Compression and Reconstruction

The increasing use of deep neural networks exposes themselves to adversarial attacks in the real world drawn from closed-set and open-set, which poses great threats to their application in safety-critical systems. Since adversarial attacks tend to mislead an original model by adding small perturbations into clean images, an intuitive idea of defensing adversarial attacks is eliminating perturbations as much as possible to mitigate attacking effects. However, such elimination-based strategies unfortunately fail to achieve satisfactory robustness. Aiming to investigate the intrinsic reasons for this phenomenon, systematic experiments are carried out in this paper to indicate that even a 20% residual perturbation can still preserve and exhibit attacking effects as strong as a full one. Our study also indicates that there are strong correlations between perturbations and legitimate images. Thus, breaking the correlation across multiple bands is more effective in mitigating attacking effects. Based on these findings, this paper proposes an efficient defense strategy called “Frequency-Adaptive Compression and rEconstruction (FACE)” to improve the robustness of the model to adversarial attacks. Specifically, low-frequency bands containing semantic information are compressed by a down-sampling operation, while the channel width of high-frequency bands is squeezed and further compressed by adding noise before the Tanh activating function. Meanwhile, attachment spaces of perturbations are also squeezed to the extent as much as possible. Finally, a clean output is obtained by upsampling together with expanded reconstruction. Experiments are extensively conducted on widely used datasets to demonstrate the effectiveness of the proposed method. For closed-set attacks, FACE outperforms the STOA elimination-based methods on ImageNet, achieving a 27.9% improvement. For the MNIST open-set attacks, it not only reduces the success rate of targeted attack by a large margin (from 100% to 24.7%), but also mitigates attacking effects with an FPR-95 value of 0.3.

Inner crossover fusion network with pixel-wise sampling for dynamic scene deblurring

Despite dynamic scene deblurring methods based on multi-scale or scale-recursive or hierarchical multi-patch strategy have made great progress, they still have some shortcomings. In the multi-scale or scale-recursive networks, bicubic down-sampling operation may cause some important high frequency information loss, e.g., strong edges. In addition, these networks are limited to using the same receptive field information while neglecting the cross-scale and cross-receptive field relationships inside the subnetwork, which may lead to information drop-out. To alleviate the above problems, we propose an inner crossover fusion network with pixel-wise sampling, called ICFNet, for dynamic scene deblurring. More specifically, we first design a pixel-wise sampling operation to replace bicubic down-sampling in the multi-scale or scale-recursive networks, so as to avoid the loss of the strong edge information. In addition, different from multi-scale and multi-patch architecture, we design a novel inner crossover fusion architecture to make full use of blur information with cross-scale and cross-receptive field in the network. To maximize the advantages of the ICFNet architecture, we design three region-enhanced residual blocks with different receptive fields for learning multiple receptive field features. Finally, extensive experimental results on different datasets have demonstrated that the proposed network can obtain better deblurring effect.

Context-aware fusion of transformers and CNNs for medical image segmentation

PurposeLocalization and screening of the target tissue is a main prerequisite of numerous medical procedures, including capsule endoscopy, colonoscopy and histology. Convolutional Neural Networks (CNNs), by stacking convolutional, down-sampling, and up-sampling operators in an encoder-decoder fashion, were the de-facto standard and has shown great promise in recent years. The main deficiency of these models is their local convolutional operators, which degrade accuracy, especially for targets with long-range dependencies. While CNNs excel at local feature extraction, Transformers are known for their ability to capture long-range dependencies. Also, CNN-Transformer models employ complex attention mechanisms for fusion that could increase model complexity and the potential for overfitting and underfitting in many datasets. MethodsIn this paper, we propose an efficient context-aware CNN-Transformer fusion mechanism based on Semi-supervised Spatial and Global Attention mechanism (SSG-Att). Our model is designed to combine the strengths of both models and overcome their limitations. High-level features that are extracted from the two parallel branches are combined and fused using the proposed SSG-Att mechanism. A hybrid loss function is also employed, which is better adapted to the introduced fusion system. ResultsWe evaluated the performance of our proposed model on the Kvasir-SEG, a polyp segmentation and detection dataset, and the Gland segmentation dataset. The experimental results confirmed that the improvements yield a top-performing yet efficient deep fused CNN-Transformer architecture. The proposed model outperformed the best-reported accuracies, achieving improved dice scores of 92.11 ± 1.10 % and 91.16 ± 0.81 on the Kvasir-SEG and GlaS datasets, respectively. ConclusionWe concluded that the proposed context-aware fusion mechanism has the potential to be used in screening and localization applications in a more reliable and accurate operation compared to other state-of-the-art methods.

Two Channel Filter Banks on Arbitrary Graphs With Positive Semi Definite Variation Operators

We propose novel two-channel filter banks for signals on graphs. Our designs can be applied to arbitrary graphs, given a positive semi definite variation operator, while using arbitrary vertex partitions for downsampling. The proposed generalized filter banks (GFBs) also satisfy several desirable properties including perfect reconstruction and critical sampling, while having efficient implementations. Our results generalize previous approaches that were only valid for the normalized Laplacian of bipartite graphs. Our approach is based on novel graph Fourier transforms (GFTs) given by the generalized eigenvectors of the variation operator. These GFTs are orthogonal in an alternative inner product space which depends on the downsampling and variation operators. Our key theoretical contribution is showing that the spectral folding property of the normalized Laplacian of bipartite graphs, at the core of bipartite filter bank theory, can be generalized for the proposed GFT if the inner product matrix is chosen properly. In addition, we study vertex domain and spectral domain properties of GFBs and illustrate their probabilistic interpretation using Gaussian graphical models. While GFBs can be defined given any choice of a vertex partition for downsampling, we propose an algorithm to optimize these partitions with a criterion that favors balanced partitions with large graph cuts, which are shown to lead to efficient and stable GFB implementations. Our numerical experiments show that partition-optimized GFBs can be implemented efficiently on 3D point clouds with hundreds of thousands of points (nodes), while also improving the color signal representation quality over competing state-of-the-art approaches.

A dual branch and fine-grained enhancement network for pancreatic tumor segmentation in contrast enhanced CT images

Segmentation of pancreatic tumors in CT images is important for clinical diagnosis and treatment, but it faces challenges of small size, low contrast, and large position difference. To address these issues, the abnormal pancreas is first segmented based on a dual branch coding network (DB-Net) using a coarse-to-fine segmentation strategy. In the encoder part, one branch extracts the semantic features of the pancreas and its surroundings, and the other branch captures the complex pancreas through wide-channel convolution and few down-sampling operations. An aggregation layer is used to fuse the different feature maps obtained by the two branches, and a U-Net decoder is used to segment the abnormal pancreas in CT images with pancreatic tumors. DB-Net is further trained to obtain the accurate pancreatic segmentation. Then, pancreatic tumors are segmented in the pancreas based on the fine-grained enhancement network (FE-Net). The FE-Net integrates a contrast enhancement block with a reverse attention block to extract detailed features and excavate effective information from the feature maps of the encoder and decoder to segment pancreatic tumors. In order to segment the tumor more accurately, the pancreatic tumor is segmented in the cropped pancreas. Experiments on 116 contrast-enhanced abdominal CT volumes of pancreatic cancer and 42 contrast-enhanced abdominal CT volumes of normal pancreas verify the effectiveness of the proposed framework in pancreatic tumor segmentation by using the two-fold cross-validation strategy. Compared to state-of-the-art deep learning segmentation network, the proposed method can achieve better segmentation of pancreas and pancreatic tumors.

High-Speed Semantic Segmentation Based on Dual-Path Feature Fusion Codec Structure

To address the problem of large parameters and slow segmentation speed in high-precision image semantic segmentation, a semantic segmentation deep learning-based model is proposed. By encoding both semantic and spatial paths, the model is able to fuse different feature information and compensate the disadvantage that it is difficult to combine spatial and semantic information, so that the feature map can be convolved efficiently. Furthermore, this model decoder merges high-level semantic information and low-level spatial information, to effectively compensate for the loss of feature information via the down-sampling operation during encoding. The experimental results on the Cityscapes and Camvid datasets show that the parameter of the model is only 3.91×10⁶, with mean intersection over union of 67.7% and 65.8% on the two datasets respectively. Also, the segmentation speeds are 111 FPS and 86 FPS, respectively. Compared with some similar models, the proposed model has fewer parameters and higher accuracy, and its segmentation speed significantly exceeds the minimum frames per second (24 FPS) required for real-time semantic segmentation.

Deep Progressive Fusion Stereo Network

Stereo matching depth estimation for rectified image pairs is of great importance to many compute vision tasks, specifically in autonomous driving. With the flourishing of convolution neural networks, responsible depth estimation of stereo matching with artificial intelligence is the most severe challenge for autonomous driving in recent years. Previous research on end-to-end trainable stereo matching networks has usually used cascading convolution blocks with down-sampling or pooling operations to extract the unary features required for matching cost construction. Such approaches lack a reconstruction stage for increasing feature map pixel-wise alignment and strength, factors which play an important role in representing the similarity between stereo image pairs. To address this issue, in this paper, we propose the progressive fusion stereo matching network (PFSM-Net). We exploit an encoder-decoder feature extraction network architecture for multi-stage and -scale dynamic feature extraction. Moreover, we propose a group-wise concatenation method to construct the cost volume, which provides a more efficient cost volume for cost aggregation. Furthermore, we propose the use of multi-scale cost aggregation networks with a progressive fusion strategy. The aggregated cost volume is progressively fused with the multi-stage and -scale cost volume as the size of the cost volume increases. Multi-stage and -scale outputs are supervised with and learned in a coarse-to-fine manner. Experimental results demonstrate that our method outperforms previous methods on the SceneFlow, KITTI 2012, and KITTI 2015 datasets.

GCHA-Net: Global context and hybrid attention network for automatic liver segmentation

Liver segmentation is a critical step in liver cancer diagnosis and surgical planning. The U-Net's architecture is one of the most efficient deep networks for medical image segmentation. However, the continuous downsampling operators in U-Net causes the loss of spatial information. To solve these problems, we propose a global context and hybrid attention network, called GCHA-Net, to adaptive capture the structural and detailed features. To capture the global features, a global attention module (GAM) is designed to model the channel and positional dimensions of the interdependencies. To capture the local features, a feature aggregation module (FAM) is designed, where a local attention module (LAM) is proposed to capture the spatial information. LAM can make our model focus on the local liver regions and suppress irrelevant information. The experimental results on the dataset LiTS2017 show that the dice per case (DPC) value and dice global (DG) value of liver were 96.5% and 96.9%, respectively. Compared with the state-of-the-art models, our model has superior performance in liver segmentation. Meanwhile, we test the experiment results on the 3Dircadb dataset, and it shows our model can obtain the highest accuracy compared with the closely related models. From these results, it can been seen that the proposed model can effectively capture the global context information and build the correlation between different convolutional layers. The code is available at the website: https://github.com/HuaxiangLiu/GCAU-Net.

A Big Coal Block Alarm Detection Method for Scraper Conveyor Based on YOLO-BS

With the aim of solving the problem of coal congestion caused by big coal blocks in underground mine scraper conveyors, in this paper we proposed the use of a YOLO-BS (YOLO-Big Size) algorithm to detect the abnormal phenomenon of coal blocks on scraper conveyors. Given the scale of the big coal block targets, the YOLO-BS algorithm replaces the last layer of the YOLOv4 algorithm feature extraction backbone network with the transform module. The YOLO-BS algorithm also deletes the redundant branch which detects small targets in the PAnet module, which reduces the overall number of parameters in the YOLO-BS algorithm. As the up-sampling and down-sampling operations in the PAnet module of the YOLO algorithm can easily cause feature loss, YOLO-BS improves the problem of feature loss and enhances the convergence performance of the model by adding the SimAM space and channel attention mechanism. In addition, to solve the problem of sample imbalance in big coal block data, in this paper, it was shown that the YOLO-BS algorithm selects focal loss as the loss function. In view of the problem that the same lump coal in different locations on the scraper conveyor led to different congestion rates, we conducted research and proposed a formula to calculate the congestion rate. Finally, we collected 12,000 image datasets of coal blocks on the underground scraper conveyor in Daliuta Coal Mine, China, and verified the performance of the method proposed in this paper. The results show that the processing speed of the proposed method can reach 80 fps, and the correct alarm rate can reach 93%. This method meets the real-time and accuracy requirements for the detection of abnormal phenomena in scraper conveyors.

Enhanced mechanisms of pooling and channel attention for deep learning feature maps.

The pooling function is vital for deep neural networks (DNNs). The operation is to generalize the representation of feature maps and progressively cut down the spatial size of feature maps to optimize the computing consumption of the network. Furthermore, the function is also the basis for the computer vision attention mechanism. However, as a matter of fact, pooling is a down-sampling operation, which makes the feature-map representation approximately to small translations with the summary statistic of adjacent pixels. As a result, the function inevitably leads to information loss more or less. In this article, we propose a fused max-average pooling (FMAPooling) operation as well as an improved channel attention mechanism (FMAttn) by utilizing the two pooling functions to enhance the feature representation for DNNs. Basically, the methods are to enhance multiple-level features extracted by max pooling and average pooling respectively. The effectiveness of the proposals is verified with VGG, ResNet, and MobileNetV2 architectures on CIFAR10/100 and ImageNet100. According to the experimental results, the FMAPooling brings up to 1.63% accuracy improvement compared with the baseline model; the FMAttn achieves up to 2.21% accuracy improvement compared with the previous channel attention mechanism. Furthermore, the proposals are extensible and could be embedded into various DNN models easily, or take the place of certain structures of DNNs. The computation burden introduced by the proposals is negligible.

Enhancing resolution of micro-CT images of reservoir rocks using super resolution

Current hardware configuration of micro-CT detectors puts a lower limit on voxel size that can be acquired while maintaining a sufficiently large field of view. This limits the degree to which rock pores can be resolved in a micro-CT image and thus restricting the application envelope of Digital Rock technology. Super resolution techniques can refine voxel size while maintaining a sufficiently large field of view using pairs of low- and high-resolution images for training. However, for interpretation of quality of Digital Rock results, image quality is not determined by voxel size alone but by the degree to which a feature such as pore throat is resolved, which depends on both the physical size of the feature and voxel size. Furthermore, artificially down-sampling finer voxel size images to obtain images of coarser voxel size for training deep learning networks is not sufficient to capture the mapping between images acquired at different resolutions. This is especially true for reservoir rocks because the noise and artifacts introduced during imaging and reconstruction are more complex than that captured by simple down-sampling operation. We overcome these two limitations, by (1) using the ratio of pore throat size and voxel size (N) to group training dataset instead of voxel size and (2) using pairs of registered micro-CT images acquired using a state-of-the-art detector instead of synthetically down-sampled images. We show that combination of these two techniques produced images with better sharpness and contrast and enabled us to refine voxel size significantly beyond what is possible using the current imaging technology while maintaining the field of view.

Deep Cascade Residual Networks (DCRNs): Optimizing an Encoder–Decoder Convolutional Neural Network for Low-Dose CT Imaging

To suppress noise and artifacts caused by the reduced radiation exposure in low-dose computed tomography, several deep learning (DL)-based image restoration methods have been proposed over the past few years. Many of these popular DL-based methods adopt an encoder–decoder framework, for instance, the residual encoder–decoder convolutional neural network. However, this popular framework may suffer from information loss for continual downsampling operations. In this article, deep cascaded residual networks (DCRNs) are proposed to optimize the popular encoder–decoder network. First, cross up- and downsampling operations as well as attention extraction are substitutes for the strict “downsampling and then upping” principle. What is more, four hybrid loss functions, namely, mean absolute error, edge loss, perceptual loss and adversarial loss, are engaged to achieve better visual effects and suppress noise. The experiments are conducted on three individual clinical CT datasets: dental CT data collected with a scanner manufactured by Zhongke Tianyue Company (ZTC), data from the American Association of Physicists in Medicine (AAPM) Challenge, and data collected with a commercial CT scanner from United Imaging Healthcare (UIH). The experimental results indicate the effective noise reduction and detail preservation capabilities of the proposed methods under different radiation dose-reduction strategies.

AIE-YOLO: Auxiliary Information Enhanced YOLO for Small Object Detection.

Small object detection is one of the key challenges in the current computer vision field due to the low amount of information carried and the information loss caused by feature extraction. You Only Look Once v5 (YOLOv5) adopts the Path Aggregation Network to alleviate the problem of information loss, but it cannot restore the information that has been lost. To this end, an auxiliary information-enhanced YOLO is proposed to improve the sensitivity and detection performance of YOLOv5 to small objects. Firstly, a context enhancement module containing a receptive field size of 21×21 is proposed, which captures the global and local information of the image by fusing multi-scale receptive fields, and introduces an attention branch to enhance the expressive ability of key features and suppress background noise. To further enhance the feature expression ability of small objects, we introduce the high- and low-frequency information decomposed by wavelet transform into PANet to participate in multi-scale feature fusion, so as to solve the problem that the features of small objects gradually disappear after multiple downsampling and pooling operations. Experiments on the challenging dataset Tsinghua-Tencent 100 K show that the mean average precision of the proposed model is 9.5% higher than that of the original YOLOv5 while maintaining the real-time speed, which is better than the mainstream object detection models.

UCR-Net: U-shaped context residual network for medical image segmentation

Medical image segmentation prerequisite for numerous clinical needs is a critical step in biomedical image analysis. The U-Net framework is one of the most popular deep networks in this field. However, U-Net’s successive pooling and downsampling operations result in some loss of spatial information. In this paper, we propose a U-shaped context residual network, called UCR-Net, to capture more context and high-level information for medical image segmentation. The proposed UCR-Net is an encoder–decoder framework comprising a feature encoder module and a feature decoder module. The feature decoder module contains four newly proposed context attention exploration(CAE) modules, a newly proposed global and spatial attention (GSA) module, and four decoder blocks. We use the proposed CAE module to capture more multi-scale context features from the encoder. The proposed GSA module further explores global context features and semantically enhanced deep-level features. The proposed UCR-Net can recover more high-level semantic features and fuse context attention information from CAE and global and spatial attention information from GSA module. Experiments on the retinal vessel, femoropopliteal artery stent, and polyp datasets demonstrate that the proposed UCR-Net performs favorably against the original U-Net and other advanced methods.

Progressive downsampling and adaptive guidance networks for dynamic scene deblurring

The existing learning-based dynamic scene deblurring methods have made good progress to some extent. However, these methods are usually based on multiscale strategy, which has the following shortcomings: (1) The bilinear downsampling operation will cause some loss of important high-frequency information, e.g., strong edges, which also further affects the network learning a better deblurring mapping. (2) Existing methods only use a single activation function, which limits the ability of the network model to fit data and causes the network performance to be easily saturated. Therefore, we propose an end-to-end progressive downsampling and adaptive guidance network called PDAG-Net for solving above problems. The proposed PDAG-Net can retain more strong edges and other high-frequency information of a blurry image so as to make the network learn a more effective deblurring mapping between the input and label images. In the proposed PDAG-Net, we implement a multiscale blended activation residual block called MSBA-ResBlock for learning the nonlinear characteristics of dynamic scene blur, which can also alleviate the performance saturation problem caused by a single activation function and improve multiscale feature extraction ability. Finally, we propose a multisupervision strategy for obtaining more robust and effective features and making the network possess more stable trainging and faster convergence. Extensive experimental results on a public dataset indicate that the proposed network outperforms the state-of-the-art image deblurring methods.

A high-resolution feature difference attention network for the application of building change detection

Deep learning based change detection has brought a significant improvement in the accuracy and efficiency when compared with conventional machine learning methods. However, the issues of the lack of differential information and the diversity of the scale features of artificial objects are crucial barriers to the application of building change detection algorithms. A novel deep learning based approach named the high-resolution feature difference attention network (HDANet) is proposed in this work to solve these issues. HDANet can handle the change characteristics well, due to the Siamese network structure. To tackle the loss of the spatial features of buildings caused by the multiple successive down-sampling operations in the current change detection algorithms using fully convolutional networks (FCNs), a multi-resolution parallel structure is introduced in HDANet, and the image information with different resolutions is comprehensively employed, without any spatial information loss. Moreover, an innovative difference attention module is elaborated for the enhancement of the sensitivity to difference information, to keep the building change information. The experimental results obtained on building change detection datasets confirm that HDANet can improve the differential feature representation for change detection, and the performance of the building change detection is also superior to that of the other advanced change detection methods.

An Application of Pixel Interval Down-Sampling (PID) for Dense Tiny Microorganism Counting on Environmental Microorganism Images

This paper proposes a novel pixel interval down-sampling network (PID-Net) for dense tiny object (yeast cells) counting tasks with higher accuracy. The PID-Net is an end-to-end convolutional neural network (CNN) model with an encoder–decoder architecture. The pixel interval down-sampling operations are concatenated with max-pooling operations to combine the sparse and dense features. This addresses the limitation of contour conglutination of dense objects while counting. The evaluation was conducted using classical segmentation metrics (the Dice, Jaccard and Hausdorff distance) as well as counting metrics. The experimental results show that the proposed PID-Net had the best performance and potential for dense tiny object counting tasks, which achieved 96.97% counting accuracy on the dataset with 2448 yeast cell images. By comparing with the state-of-the-art approaches, such as Attention U-Net, Swin U-Net and Trans U-Net, the proposed PID-Net can segment dense tiny objects with clearer boundaries and fewer incorrect debris, which shows the great potential of PID-Net in the task of accurate counting.

Statistical System of Cultural Heritage Tourism Information Based on Image Feature Extraction Technology

This paper presents an in-depth study of a statistical system using image feature extraction technology for cultural heritage tourism information. Based on the specificity of the tourism industry, it combines the technical architecture, framework of image feature extraction, GIS basic design, database design, and other related theories; realizes visual mapping through relevant tourism statistics in GIS; and uses the spatial analysis capability of ArcGIS software to complete the statistical scheme of cultural heritage tourism information data in the context of Internet big data. In this paper, the image preprocessing process is accelerated by the parallel optimization of the filtering operation and down-sampling operation in the relevant image preprocessing algorithm. The preprocessed images can effectively improve the recognition effect of the image feature extraction and classification stage. Based on cultural heritage tourism information mining results, the design and development of the tourism information statistics platform are carried out. The platform is built and developed using GIS secondary development technology and HTML + CSS front-end development technology API, Echarts, and other image feature components. The platform enables the massive and disorganized tourism text information to be displayed in a way that is easier to understand in both maps and text. This paper’s research on critical technologies provides technical support for image feature extraction of tourism statistics, which has specific practical value. Implementing the cultural heritage tourism platform offers a visual operating environment for tourism, enabling cultural heritage tourism businesses to tap into user preferences so that users can effectively access the travel options that interest them, which is conducive to promoting the development of travel recommendation platforms.