Feature Fusion Algorithm Research Articles

DMs-MAFM+EfficientNet: a hybrid model for predicting dysthyroid optic neuropathy.

Thyroid-associated ophthalmopathy (TAO) is a very common autoimmune orbital disease. Approximately 4%–8% of TAO patients will deteriorate and develop the most severe dysthyroid optic neuropathy (DON). According to the current data provided by clinical experts, there is still a certain proportion of suspected DON patients who cannot be diagnosed, and the clinical evaluation has low sensitivity and specificity. There is an urgent need for an efficient and accurate method to assist physicians in identifying DON. This study proposes a hybrid deep learning model to accurately identify suspected DON patients using computed tomography (CT). The hybrid model is mainly composed of the double multiscale and multi attention fusion module (DMs-MAFM) and a deep convolutional neural network. The DMs-MAFM is the feature extraction module proposed in this study, and it contains a multiscale feature fusion algorithm and improved channel attention and spatial attention, which can capture the features of tiny objects in the images. Multiscale feature fusion is combined with an attention mechanism to form a multilevel feature extraction module. The multiscale fusion algorithm can aggregate different receptive field features, and then fully obtain the channel and spatial correlation of the feature map through the multiscale channel attention aggregation module and spatial attention module, respectively. According to the experimental results, the hybrid model proposed in this study can accurately identify suspected DON patients, with Accuracy reaching 96%, Specificity reaching 99.5%, Sensitivity reaching 94%, Precision reaching 98.9% and F1-score reaching 96.4%. According to the evaluation by experts, the hybrid model proposed in this study has some enlightening significance for the diagnosis and prediction of clinically suspect DON.Graphical abstract

MFF: Multi-modal feature fusion for zero-shot learning

Generative Zero-Shot Learning (ZSL) methods generally generate pseudo-samples/features based on the semantic description information of unseen classes, thereby transforming ZSL tasks into traditional supervised learning tasks. Under this learning paradigm, the quality of pseudo-samples/features guided by the classes’ semantic description information is the key to the success of the model. However, the semantic description information used in the existing generative methods is mainly the low-dimensional representation (e.g., attributes) of classes, which leads to the low quality of the generated pseudo-samples/features and may aggravate the problem of domain shift. To alleviate this problem, we introduce the visual principal component feature, which is extracted by a principal component analysis network, to make up for the deficiency of using only semantic description information and propose a novel Variational Auto-Encoder (VAE) and Generative Adversarial Network (GAN) based generative method for ZSL, which we call Multi-modal Feature Fusion algorithm (MFF). In MFF, the input of different modal information enables VAE better fit the original data distribution and the proposed alignment loss ensures the consistency of the generated visual features and the corresponding semantic features. With the help of high-quality pseudo-samples/features, the ZSL model can make more accurate predictions for unseen classes. Extensive experiments on five public datasets demonstrate that our proposed algorithm outperforms several state-of-the-art methods under both ZSL and generalized ZSL settings.

Face Recognition Attendance System Method Based on Fusion of LBP and HOG

As one of the hot topics in the field of computer vision research, face recognition technology has received significant attention due to its potentiality for a wide range of applications in government as well as commercial purposes. In practical applications, although several existing face recognition methods have achieved good performances in specific scenes, they easily suffer from a sharp decline in recognition rate if affected by different conditions of light, expression, posture and occlusion. Among many factors, influences of complex illuminations on face recognition are particularly significant. To further improve the performance of the existing local binary pattern (LBP) operator, neighbourhood weighted average LBP (NWALBP) is first proposed for fully considering the strong correlations between pixel pairs in the neighbourhood, which extends the traditional LBP uni-layer neighbourhood template window to the bi-layer neighbourhood template window and calculates the weighted average of bi-layer neighbourhood pixels in each direction. Then, inspired by centre symmetric LBP (CS-LBP), centre symmetric NWALBP (CS-NWALBP) is further proposed, which can effectively reduce computation complexity by only comparing the weighted average values of the neighbourhood pixels that are symmetric about the centre pixel. Finally, by combining the merit of histogram of oriented gradient (HOG), a feature fusion algorithm named CS-NWALBP+HOG is suggested. Several experiments have eventually demonstrated that our proposed algorithms have more robust performance under complex illumination conditions if compared with many other latest algorithms.

Feature layer fusion of linear features and empirical mode decomposition of human EMG signal

To explore the influence of the fusion of different features on recognition, this paper took the electromyogram (EMG) signals of rectus femoris under different motions (walk, step, ramp, squat, and sitting) as signals, linear features (time-domain features (variance (VAR) and root mean square (RMS)), frequency-domain features (mean frequency (MF) and mean power frequency (MPF)), and nonlinear features (EMD) of the signals were extracted. Two feature fusion algorithms, the series splicing method and complex vector method, were designed, which were verified by a double hidden layer error back propagation (BP) neural network. Results show that with the increase of the types and complexity of feature fusions, the recognition rate of the EMG signal to actions is gradually improved. When the EMG signal is used in the series splicing method, the recognition rate of time-domain ​+ ​frequency-domain ​+ ​empirical mode decomposition (TD ​+ ​FD ​+ ​EMD) splicing is the highest, and the average recognition rate is 92.32%. And this value is raised to 96.1% by using the complex vector method, and the variance of the BP system is also reduced.

Cultural Tourism Industry Feature Extraction Based on Multiscale Feature Fusion Algorithm and Artificial Intelligence

The cultural tourism industry combines the similarities of the cultural industry and tourism industry, which can be the most effective way to meet people's spiritual and cultural needs as well as their leisure needs simultaneously, and has a vast development potential. However, there are numerous and dispersed areas where the market value of cultural tourism resources is clustered, and frequently, each city has the clustering area with the highest market value concentration of cultural tourism resources. This feature of the spatial distribution of the market value of cultural tourism resources is significant for promoting the development of cultural tourism as a whole and constructing the industry's overall structure. It has broad application potential for extracting and differentiating cultural tourism industry characteristics. Texture feature extraction is typically performed using dual-tree complex wavelet transform (DT-CWT) and Gabor wavelet. In this paper, we propose a multiscale DT-CWT and Gabor-based method for identifying the cultural tourism industry. The method first decomposes the images of cultural tourism into multiscale space using a Gaussian pyramid, then extracts the multiscale features of the images using DT-CWT and Gabor, and lastly achieves feature fusion. Using a support vector machine (SVM) classifier to achieve classification, the effectiveness of the feature extraction method is determined. The experimental findings demonstrate that the method proposed in this paper can achieve a high rate of recognition.

PSuc-FFSEA: Predicting Lysine Succinylation Sites in Proteins Based on Feature Fusion and Stacking Ensemble Algorithm.

Being a new type of widespread protein post-translational modifications discovered in recent years, succinylation plays a key role in protein conformational regulation and cellular function regulation. Numerous studies have shown that succinylation modifications are closely associated with the development of many diseases. In order to gain insight into the mechanism of succinylation, it is vital to identify lysine succinylation sites. However, experimental identification of succinylation sites is time-consuming and laborious, and traditional identification tools are unable to meet the rapid growth of datasets. Therefore, to solve this problem, we developed a new predictor named pSuc-FFSEA, which can predict succinylation sites in protein sequences by feature fusion and stacking ensemble algorithm. Specifically, the sequence information and physicochemical properties were first extracted using EBGW, One-Hot, continuous bag-of-words, chaos game representation, and AAF_DWT. Following that, feature selection was performed, which applied LASSO to select the optimal subset of features for the classifier, and then, stacking ensemble classifier was designed using two-layer stacking ensemble, selecting three classifiers, SVM, broad learning system and LightGBM classifier, as the base classifiers of the first layer, using logistic regression classifier as the meta classifier of the second layer. In order to further improve the model prediction accuracy and reduce the computational effort, bayesian optimization algorithm and grid search algorithm were utilized to optimize the hyperparameters of the classifier. Finally, the results of rigorous 10-fold cross-validation indicated our predictor showed excellent robustness and performed better than the previous prediction tools, which achieved an average prediction accuracy of 0.7773 ± 0.0120. Besides, for the convenience of the most experimental scientists, a user-friendly and comprehensive web-server for pSuc-FFSEA has been established at https://bio.cangmang.xyz/pSuc-FFSEA, by which one can easily obtain the expected data and results without going through the complicated mathematics.

Electromagnetic Modulation Signal Classification Using Dual-Modal Feature Fusion CNN.

AMC (automatic modulation classification) plays a vital role in spectrum monitoring and electromagnetic abnormal signal detection. Up to now, few studies have focused on the complementarity between features of different modalities and the importance of the feature fusion mechanism in the AMC method. This paper proposes a dual-modal feature fusion convolutional neural network (DMFF-CNN) for AMC to use the complementarity between different modal features fully. DMFF-CNN uses the gram angular field (GAF) image coding and intelligence quotient (IQ) data combined with CNN. Firstly, the original signal is converted into images by GAF, and the GAF images are used as the input of ResNet50. Secondly, it is converted into IQ data and as the complex value network (CV-CNN) input to extract features. Furthermore, a dual-modal feature fusion mechanism (DMFF) is proposed to fuse the dual-modal features extracted by GAF-ResNet50 and CV-CNN. The fusion feature is used as the input of DMFF-CNN for model training to achieve AMC of multi-type signals. In the evaluation stage, the advantages of the DMFF mechanism proposed in this paper and the accuracy improvement compared with other feature fusion algorithms are discussed. The experiment shows that our method performs better than others, including some state-of-the-art methods, and has superior robustness at a low signal-to-noise ratio (SNR), and the average classification accuracy of the dataset signals reaches 92.1%. The DMFF-CNN proposed in this paper provides a new path for the AMC field.

Automatic Multichannel Electrocardiogram Record Classification Using XGBoost Fusion Model.

There is an increasing demand for automatic classification of standard 12-lead electrocardiogram signals in the medical field. Considering that different channels and temporal segments of a feature map extracted from the 12-lead electrocardiogram record contribute differently to cardiac arrhythmia detection, and to the classification performance, we propose a 12-lead electrocardiogram signal automatic classification model based on model fusion (CBi-DF-XGBoost) to focus on representative features along both the spatial and temporal axes. The algorithm extracts local features through a convolutional neural network and then extracts temporal features through bi-directional long short-term memory. Finally, eXtreme Gradient Boosting (XGBoost) is used to fuse the 12-lead models and domain-specific features to obtain the classification results. The 5-fold cross-validation results show that in classifying nine categories of electrocardiogram signals, the macro-average accuracy of the fusion model is 0.968, the macro-average recall rate is 0.814, the macro-average precision is 0.857, the macro-average F1 score is 0.825, and the micro-average area under the curve is 0.919. Similar experiments with some common network structures and other advanced electrocardiogram classification algorithms show that the proposed model performs favourably against other counterparts in F1 score. We also conducted ablation studies to verify the effect of the complementary information from the 12 leads and the auxiliary information of domain-specific features on the classification performance of the model. We demonstrated the feasibility and effectiveness of the XGBoost-based fusion model to classify 12-lead electrocardiogram records into nine common heart rhythms. These findings may have clinical importance for the early diagnosis of arrhythmia and incite further research. In addition, the proposed multichannel feature fusion algorithm can be applied to other similar physiological signal analyses and processing.

Application of Transfer Learning and Feature Fusion Algorithms to Improve the Identification and Prediction Efficiency of Premature Ovarian Failure.

Ultrasound imaging technology has the advantages of noninvasiveness, real-time, low price, and easy operation. It is one of the most used diagnostic tools for early detection and classification of premature ovarian failure. Although the rapid development of computer-aided diagnosis has provided a great help to the ultrasound diagnosis of premature ovarian failure, it still has many limitations and shortcomings, so this paper adopts transfer learning and feature fusion algorithms to improve the identification and prediction efficiency of premature ovarian failure. In this study, the POF group and the control group both adopted a unified scale. From the four aspects of sociological characteristics, past medical history, environmental factors, and living habits, a dedicated person asked and filled out the scale face to face. All patients participating in the experiment underwent ultrasound examinations. In this paper, the bottom-level feature fusion method is used to improve classification performance. The experiment uses 100 epochs. After each epoch training is completed, we used all the data and labels of the target domain to test. All experiments were performed five times, and the result is the average of five experiments. All the results of baseline and direct classification without migration use the average of five experimental results as the result. Migrating the features extracted by the InceptionV3 network has the best performance for predicting premature ovarian failure. Its classification accuracy is as high as 85.13%, and the F1 value is 0.78. The results show that the migration learning and feature fusion algorithms used in this paper can provide reliable predictive analysis and decision support for doctors in the diagnosis of premature ovarian failure.

Hyperspectral remote sensing image classification based on spectral-spatial feature fusion and PSO algorithm

Hyperspectral images(HSI) have rich spectral information and spatial information. In the classification of hyperspectral images, the combination of spectral information and spatial information has become an effective means to obtain good classification results. Specifically, firstly, PCA algorithm is introduced to extract the spectral information of the image. Secondly, a bilateral filter is introduced for each band to extract the spatial information of the image. Thirdly, the image is classified by using SVM to get the final classification result. Considering that the performance of SVM depends on the choice of parameters, particle swarm optimization algorithm (PSO) is combined with the proposed method to find the optimal penalty parameters and kernel parameters. We find that the classification accuracy is effectively improved by using PCA, bilateral filtering and optimizing SVM-PSO parameters on this basis. Experimental results based on one real hyperspectral dataset show that the proposed algorithm can achieve higher classification accuracy in a shorter time than other algorithms.

Lane Departure Warning System Based on New Feature Fusion Algorithm and Departure Judgment Rule

Lane Departure Warning System Based on New Feature Fusion Algorithm and Departure Judgment Rule

Magnetic Anomaly Detection Method Based on Feature Fusion and Isolation Forest Algorithm

Under the background of Gaussian colored environmental magnetic noise, for the problem that a supervised learning algorithm has high requirements on samples and poor practicability, a magnetic anomaly detection method based on feature fusion and isolation forest (IForest) algorithm is proposed. The method uses different feature algorithms to extract the statistical features, time-frequency features and fractal features of the signal, reduces the dimensionality of the features by principal component analysis (PCA) and generates feature fusion tensors. Finally the IForest algorithm is used to achieve target detection. The simulation and experimental results show that the method has a higher detection rate under different SNR of Gaussian color noise, which is approximately 5%-18% higher than that of the traditional feature detection algorithm. This method can train an effective detection model with only a small number of negative samples. Compared with the fully connected neural network (FCN) model trained with unbalanced samples, the detection rate increases by approximately 5%-12%, and it takes less time.

A Paddy Field Segmentation Method Combining Attention Mechanism and Adaptive Feature Fusion

HighlightsA paddy field segmentation network model, SA-DeepLabv3+, for hilly areas in southern China is proposed.SA-DeepLabv3+ uses attention mechanism and adaptive spatial feature fusion algorithm.The deep model was trained end-to-end using a few annotated image datasets.SA-DeepLabv3+ has good segmentation results for the four common types of paddy fields.Abstract. The hilly areas of southern China have complex feature distributions, and the boundaries of cultivated land are irregular. This kind of environment causes several image segmentation problems such as low accuracy of UAV low-altitude remote sensing arable land image interpretation. This paper proposes a SA-DeepLabv3+ paddy field segmentation model that combines attention mechanism and adaptive spatial feature fusion algorithm based on DeepLabv3+ semantic segmentation model. The sampling areas include Yangwan Village, Duchang County, Jiangxi Province, China, Keli Village, Xinjian County, Nanchang City and Chengxin Farm, Nanchang City. A high-resolution image dataset was found by capturing low-altitude remote sensing images of paddy fields with DJI UAVs, and the dataset images are categorised into four types of paddy fields. By analysing the image features of the dataset, the feature map is remapped by incorporating scSE attention mechanism module in the backbone network of the DeepLabv3+ encoder to enhance the paddy field feature representation. The ASFF algorithm is applied in the decoder, which obtains an adaptive weight coefficient to improve the problem of inadequate fusion of multi-scale features in the upsampling. By comparing the segmentation results of models with different ASPP dilation rates, the optimal set of dilation rates is determined to improve the model's ability to extract fine features of paddy fields. Finally, the performance of this model for paddy fields segmentation is verified by comparison experiments of different models. The data show that the mean values of PA, MIoU, Recall, and F1 score of the SA-DeepLabv3+ paddy field segmentation model are 0.955, 0.875, 0.865, and 0.908, respectively, which are improved by 0.019, 0.032, 0.008, and 0.005, and also has a higher segmentation performance compared with the typical segmentation models of UNet, SegNet, and PSPNet. The results show that the SA-DeepLabv3+ paddy field segmentation model with high accuracy and robustness, which provides an important basis for further acquiring high-precision paddy field boundary positioning information and constructing high-precision maps of several paddy fields in larger areas, and plays a positive role in promoting efficient and accurate information management of paddy fields. Keywords: Attention mechanism, ASFF, DeepLabv3+ network, Paddy field.

A multi-feature fusion decoding study for unilateral upper-limb fine motor imagery.

To address the fact that the classical motor imagination paradigm has no noticeable effect on the rehabilitation training of upper limbs in patients after stroke and the corresponding feature extraction algorithm is limited to a single domain, this paper describes the design of a unilateral upper-limb fine motor imagination paradigm and the collection of data from 20 healthy people. It presents a feature extraction algorithm for multi-domain fusion and compares the common spatial pattern (CSP), improved multiscale permutation entropy (IMPE) and multi-domain fusion features of all participants through the use of decision tree, linear discriminant analysis, naive Bayes, a support vector machine, k-nearest neighbor and ensemble classification precision algorithms in the ensemble classifier. For the same subject, the average classification accuracy improvement of the same classifier for multi-domain feature extraction relative to CSP feature results went up by 1.52%. The average classification accuracy improvement of the same classifier went up by 32.87% relative to the IMPE feature classification results. This study's unilateral fine motor imagery paradigm and multi-domain feature fusion algorithm provide new ideas for upper limb rehabilitation after stroke.

Clustering Feature Constraint Multiscale Attention Network for Shadow Extraction From Remote Sensing Images

Shadow extraction is an important and challenging task in remote sensing image analysis because the presence of shadows not only reduces radiation information but also affects the interpretation of remote sensing images. In this article, a clustering feature constraint multiscale attention network for shadow extraction from remote sensing images is proposed. First, in addition to the pixel-level description of the traditional neural network, our method focuses on the clustering relationships between pixel pairs to obtain the pixel group features of shadows. The feature extraction capability of the network is improved with a reweighting mechanism at the pixel level and pixel group features. Second, we employ a feature fusion algorithm by considering contextual information to improve the network’s attention toward shadow areas and enhance the nonlinear expression ability during the encoding and decoding layers. Furthermore, considering the most prominent multiscale features of shadows in remote sensing images, a deep multiscale feature aggregation structure is established to better fit the multiscale feature expression of shadows. Finally, we construct a shadow extraction dataset to verify the proposed approach. We compare our method with the results of state-of-the-art deep learning models. The results show that the intersection over union (IOU) of our method is improved by 0.85%–9.51% and that the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F1$ </tex-math></inline-formula> -score is improved by 0.73–6.48. In addition, the test results for images with different resolutions prove that the proposed approach is more robust than the other methods.

Motor imagery EEG classification algorithm based on CNN-LSTM feature fusion network

Motor imagery brain-computer interface (MI-BCI) provides a novel way for human-computer interaction. Traditional neural networks often use serial structure to extract spatial features when dealing with motor imagery EEG signal classification, ignoring temporal information and a large amount of available information in the middle layer, resulting in poor classification performance of MI-BCI. A neural network feature fusion algorithm is proposed by combining the convolutional neural network (CNN) and the long-short-term memory network (LSTM). Specifically, the CNN and LSTM are connected in parallel. The CNN extracts spatial features, the LSTM extracts temporal features and the flatten layer added after the convolutional layer extracts the middle layer features. Then all the features are merged in the fully connected layer to improve the accuracy of classification. The average accuracy and Kappa value of all subjects were 87.68% and 0.8245, respectively. The result shows that the feature fusion neural network proposed in this paper can effectively improve the accuracy of motor imagery EEG, and provides new ideas for the study of feature extraction and classification of motor imagery brain-computer interfaces.

Multi-modal biometric fusion based continuous user authentication for E-proctoring using hybrid LCNN-Salp swarm optimization.

In Covid 19, pandemic remote proctoring of the employee or human being is evolved as a big challenge for the information retrieval process. On the other side, memory-based system access authentication is becoming outdated and less preferred for live applications, especially where data security and customer privacy are crucial. Multi-modal authentication has outperformed the unimodal process with high accuracy and improved security in the user authentication field. Multi-modal biometric verification includes user attributes such as keystrokes, iris, speech, face, etc. For real-time execution of multi-modal biometric fusion-based live tracking for compatible applications. The study proposes an efficient continuous biometric user authentication system for a new challenge of pandemic time, a live online authentication of the evaluation process (CBUA-OE). The proposed CBUA-OE system can address the challenges associated with live proctoring and is also compatible with real-time implementation, deployment of authentication systems. The modified wolf optimization algorithm and CUBA-OE's optimal feature fusion algorithm give an edge over the other contemporary methods and make it more robust. In modern forms of authentication, the classification stage affects the overall outcome of the system, and the model's performance is also a factor of varying quality of datasets. In contrast, a hybrid LCNN-Salp swarm optimization-based classifier is more efficient and consistent in continuous user authentication. Here the performance of the proposed hybrid LCNN-Salp swarm optimization classifier is analyzed with different standard datasets. The results are compared with the existing state-of-art classifiers regarding the accuracy, precision, recall, and F-measure. This projected work is novel in terms of usability factors and scalability to live tracking systems.

Fine-Grained Semantic Image Synthesis with Object-Attention Generative Adversarial Network

Semantic image synthesis is a new rising and challenging vision problem accompanied by the recent promising advances in generative adversarial networks. The existing semantic image synthesis methods only consider the global information provided by the semantic segmentation mask, such as class label, global layout, and location, so the generative models cannot capture the rich local fine-grained information of the images (e.g., object structure, contour, and texture). To address this issue, we adopt a multi-scale feature fusion algorithm to refine the generated images by learning the fine-grained information of the local objects. We propose OA-GAN, a novel object-attention generative adversarial network that allows attention-driven, multi-fusion refinement for fine-grained semantic image synthesis. Specifically, the proposed model first generates multi-scale global image features and local object features, respectively, then the local object features are fused into the global image features to improve the correlation between the local and the global. In the process of feature fusion, the global image features and the local object features are fused through the channel-spatial-wise fusion block to learn ‘what’ and ‘where’ to attend in the channel and spatial axes, respectively. The fused features are used to construct correlation filters to obtain feature response maps to determine the locations, contours, and textures of the objects. Extensive quantitative and qualitative experiments on COCO-Stuff, ADE20K and Cityscapes datasets demonstrate that our OA-GAN significantly outperforms the state-of-the-art methods.

Deep Learning-Based Magnetic Resonance Imaging Features in Diagnosis of Perianal Abscess and Fistula Formation.

There was an investigation of the diagnostic and prognostic effect of magnetic resonance imaging (MRI) based on multimodal feature fusion algorithm for impotence of perianal abscess. In this study, the second to fifth convolution blocks of the visual geometric group network were applied to extract the depth features in the way of transfer learning, and a multimode feature fusion algorithm was constructed. The whole network was trained by maximizing the energy proportion of the feature layers, which was compared with the fully convolutional neural network (FCN) algorithm. Then, this algorithm was adopted to the imaging diagnosis of 50 patients with anorectal diseases admitted to our hospital, and it was found that the similarity coefficient (85.37%), accuracy (80.02%), and recall rate (79.38%) of the improved deep learning algorithm were higher markedly than those of the FCN algorithm (70.18%, 67.82%, and 66.92%) (P < 0.05). As the number of convolutional layers increased, the segmentation accuracy of the convolutional neural network (CNN) algorithm was also improved. The detection rate of the observation group (84%) rose hugely compared with the control group (64%), and the difference was statistically obvious (P < 0.05). Besides, the detection accuracy of abscess location (84%), impotent tract location (80%), and internal orifice location (92%) in patients from the observation group was higher substantially than the accuracy of abscess location (60%), impotent tract location (68%), and internal orifice location (72%) from the control group (P < 0.05). In conclusion, the performance of the multimodal feature fusion algorithm was better, and the MRI image feature analysis based on this algorithm had a higher diagnostic accuracy, which had a positive effect on improving the detection rate, detection accuracy, and disease classification.

Synthetic region screening and adaptive feature fusion for constructing a flexible object detection database

A large amount of training data (including samples and labels) are required to achieve the task of object detection. However, the construction of databases generally has several problems. The samples are not balanced. In some scenarios, very few or even no samples can be collected; manual collection and labeling are expensive. Obtaining samples from public object detection data sets is simple, but the categories that can be obtained are limited and cannot meet the needs of other detection tasks. Therefore, we propose a method to automatically construct a flexible object detection database and complete automatic labeling. Our main advantage lies in the proposed synthetic region screening and adaptive feature fusion algorithm, which generates sample images and corresponding annotation files that are similar to the real collection. Our algorithm has a stronger and more realistic synthesis ability for three-dimensional objects than other data synthesis methods. Our method improves the system’s object detection accuracy in the following ways: (1) it synthesizes a wider range of images for certain categories with few samples in the data set to solve the problem of sample imbalance; (2) it adds new categories to the public data set to meet the needs of rapid deployment of special object detection tasks; and (3) it fuses different foregrounds and backgrounds to enrich the diversity of database samples. To effectively evaluate the proposed method, we conducted experiments separately, and the experimental results proved that our method is superior to other existing data synthesis and data enhancement methods. Combining the generated samples with the real data set, the object detection accuracy value increased from 9.47% to 58.92%. Extending the objects of the self-collected data set to the public data set and automatically generating high-quality annotations, the detection accuracy of extended objects reached more than 50%.