Fast Detection Speed Research Articles

Track Fastener Defect Detection Model Based on Improved YOLOv5s.

Defect detection of track fasteners is a prerequisite for safe and reliable railroad operation. The traditional manual visual inspection method has been unable to meet the growing demand for railroad network inspection in China. To achieve the need for accurate, fast, and intelligent detection of rail fasteners, this paper proposes a rail fastener defect detection model based on improved YOLOv5s. Firstly, the convolutional block attention module (CBAM) is added to the Neck network of the YOLOv5s model to enhance the extraction of essential features by the model and suppress the information of minor features. Secondly, a weighted bidirectional feature pyramid network (BiFPN) is introduced to realize the multi-scale feature fusion of the model. Finally, the K-means++ algorithm is used to re-cluster the dataset to obtain the anchor box suitable for the fastener dataset and improve the positioning ability of the model. The experimental results show that the improved model achieves an average mean precision (mAP) of 97.4%, a detection speed of 27.3 FPS, and a model memory occupancy of 15.5 M. Compared with the existing target detection model, the improved model has the advantages of high detection accuracy, fast detection speed, and small model memory occupation, which can provide technical support for edge deployment of rail fastener defect detection.

CourtNet: Dynamically balance the precision and recall rates in infrared small target detection

Infrared small-target detection (ISTD) is an important computer vision task. ISTD aims at separating small targets from complex background clutter. The infrared radiation decays with distance, making the targets highly dim and prone to confusion with the background clutter, which makes the detector challenging to balance the precision and recall rates. To deal with this difficulty, this paper proposes a neural-network-based ISTD method called CourtNet, which has three sub-networks: the prosecution network is designed to improve the recall rate; the defendant network is devoted to increasing the precision rate; the jury network weights their results to adaptively balance the precision and recall rates. CourtNet takes the structure of Transformers, whose feature resolution remains unchanged. Furthermore, the prosecution network utilizes a densely connected structure, which can prevent small targets from disappearing in the forward propagation. In addition, a fine-grained attention module performs attention inside patches to accurately locate the small targets. This paper implements extensive experiments on two ISTD datasets, MFIRST and SIRST, and compares CourtNet with ten other traditional and deep-learning-based methods. Experimental results show that with the fast detection speed (60.61 FPS), CourtNet achieves the best F1 score, 0.62 (in MFIRST) and 0.73 (in SIRST), among the compared methods. The code and dataset will be available at https://github.com/PengJingchao/CourtNet.

Deep Camera-Radar Fusion with an Attention Framework for Autonomous Vehicle Vision in Foggy Weather Conditions.

AVs are affected by reduced maneuverability and performance due to the degradation of sensor performances in fog. Such degradation can cause significant object detection errors in AVs' safety-critical conditions. For instance, YOLOv5 performs well under favorable weather but is affected by mis-detections and false positives due to atmospheric scattering caused by fog particles. The existing deep object detection techniques often exhibit a high degree of accuracy. Their drawback is being sluggish in object detection in fog. Object detection methods with a fast detection speed have been obtained using deep learning at the expense of accuracy. The problem of the lack of balance between detection speed and accuracy in fog persists. This paper presents an improved YOLOv5-based multi-sensor fusion network that combines radar object detection with a camera image bounding box. We transformed radar detection by mapping the radar detections into a two-dimensional image coordinate and projected the resultant radar image onto the camera image. Using the attention mechanism, we emphasized and improved the important feature representation used for object detection while reducing high-level feature information loss. We trained and tested our multi-sensor fusion network on clear and multi-fog weather datasets obtained from the CARLA simulator. Our results show that the proposed method significantly enhances the detection of small and distant objects. Our small CR-YOLOnet model best strikes a balance between accuracy and speed, with an accuracy of 0.849 at 69 fps.

Research on Forest Fire Detection Algorithm Based on Improved YOLOv5

Forest fires are one of the world’s deadliest natural disasters. Early detection of forest fires can help minimize the damage to ecosystems and forest life. In this paper, we propose an improved fire detection method YOLOv5-IFFDM for YOLOv5. Firstly, the fire and smoke detection accuracy and the network perception accuracy of small targets are improved by adding an attention mechanism to the backbone network. Secondly, the loss function is improved and the SoftPool pyramid pooling structure is used to improve the regression accuracy and detection performance of the model and the robustness of the model. In addition, a random mosaic augmentation technique is used to enhance the data to increase the generalization ability of the model, and re-clustering of flame and smoke detection a priori frames are used to improve the accuracy and speed. Finally, the parameters of the convolutional and normalization layers of the trained model are homogeneously merged to further reduce the model processing load and to improve the detection speed. Experimental results on self-built forest-fire and smoke datasets show that this algorithm has high detection accuracy and fast detection speed, with average accuracy of fire up to 90.5% and smoke up to 84.3%, and detection speed up to 75 FPS (frames per second transmission), which can meet the requirements of real-time and efficient fire detection.

State-of-the-Art Deep Learning Methods for Objects Detection in Remote Sensing Satellite Images.

Introduction: Object detection in remotely sensed satellite images is critical to socio-economic, bio-physical, and environmental monitoring, necessary for the prevention of natural disasters such as flooding and fires, socio-economic service delivery, and general urban and rural planning and management. Whereas deep learning approaches have recently gained popularity in remotely sensed image analysis, they have been unable to efficiently detect image objects due to complex landscape heterogeneity, high inter-class similarity and intra-class diversity, and difficulty in acquiring suitable training data that represents the complexities, among others. Methods: To address these challenges, this study employed multi-object detection deep learning algorithms with a transfer learning approach on remotely sensed satellite imagery captured on a heterogeneous landscape. In the study, a new dataset of diverse features with five object classes collected from Google Earth Engine in various locations in southern KwaZulu-Natal province in South Africa was used to evaluate the models. The dataset images were characterized with objects that have varying sizes and resolutions. Five (5) object detection methods based on R-CNN and YOLO architectures were investigated via experiments on our newly created dataset. Conclusions: This paper provides a comprehensive performance evaluation and analysis of the recent deep learning-based object detection methods for detecting objects in high-resolution remote sensing satellite images. The models were also evaluated on two publicly available datasets: Visdron and PASCAL VOC2007. Results showed that the highest detection accuracy of the vegetation and swimming pool instances was more than 90%, and the fastest detection speed 0.2 ms was observed in YOLOv8.

Network security risk detection method for smart microgrid monitoring system

AbstractAlthough the smart microgrid improves the power quality of the power system, it still has some security risks. So, this paper proposes a network security risk detection method of an intelligent microgrid monitoring system based on the artificial immune algorithm. By analyzing the structure of a smart microgrid monitoring system and introducing Policy Protection Detection Response (P2DR) dynamic network security model and ale static network security analysis method, the network security risk of the microgrid monitoring system is evaluated. The theoretical model is analyzed by the overall framework of the artificial immune algorithm, and its term definition is given. The mutation self is detected by self‐dynamic description and immune cell evolution process with the network structure brought into and the risk of the computing network facing different attack subjects is analyzed. This method has a high detection accuracy for more than 10 kinds of network attack modes, the highest detection accuracy is 98%, and the average time is only 0.8 s. Compared with other methods, this method has higher detection accuracy and faster detection speed. It is universal in various network attack modes and can detect the possible network security risks of the intelligent microgrid monitoring system and protect its safe and stable operation.

Remote Sensing Ship Detection and Classification Based on Improved BBAVectors Algorithm

In recent years, most of remote sensing object detection algorithms only pursue higher detection accuracy, which use two-stage anchor-based networks. However, the two-stage anchor-based network usually has many parameters, complex calculation, and low detection speed, which is not suitable for object detection tasks with real-time requirements. BBAVectors is a network based on the one-stage anchor-free type CenterNet architecture. It has a fast detection speed, but the detection accuracy is far lower than the current high-precision remote sensing object detectors. In order to improve the detection accuracy of BBAVectors so that it can be better used in real-time remote sensing ship detection and classification, we propose the BBAV-ACPP. Based on the BBAVectors architecture, we add the Coordinate Attention Module and modify the center point representation. The proposed BBAV-ACPP architecture is trained and tested on the public remote sensing ship dataset HRSC 2016. Experiments show that the proposed BBAV-ACPP architecture can improve the detection accuracy while maintaining a fast detection speed. In some remote sensing ship detection tasks with real-time requirements, BBAV-ACPP can detect ship targets quickly and accurately.

Detection and quantification of cotton trichomes by deep learning algorithm

Cotton is one of the most economically important agricultural crops with the harvest of cotton fibers. Trichomes are initiated from the plant epidermis, which is responsible for plant resistance to many abiotic and biotic stresses. Cotton fibers share similar underlying developmental molecular mechanisms with trichomes on the stem and leaf margins of cotton. Compared with fibers, it is easier to observe the trichome phenotype of cotton stem and leaf margins. However, because of the variability in coloration and the high occlusion of the trichomes, counting the density and observing the distribution of trichomes is still highly laborious. To help resolve this problem, the present research provides a solution based on deep learning in the screening of cotton trichomes, which makes it possible to monitor the phenotype of the cotton efficiently. With a fluorescent microscopy imaging procedure, we clearly observed that there are two types of trichome phenotype inGossypium hirsutum(‘TM-1′), which are singled and clustered trichomes. The clustered trichome originates from two or more adjacent epidermal cells with no branches. The TM1 trichome was used to create a collection of singled and clustered trichome pictures. Following data production, we evaluated the effectiveness of several you only look once (YOLO) model iterations in the context of identifying singled and clustered trichomes, respectively. Results show that the YOLOv5s model outperforms YOLOv3 and YOLOv4 in average accuracy, recall, accuracy, and F1 score. We utilized the clustered trichome data set to assess the performance of YOLOv5s and Mask R-CNN in order to more precisely detect the real trichome distribution with simultaneous detection of single and clustered trichomes. The obtained findings show that both models may be utilized to concurrently identify singled and clustered trichomes since the discrepancy between the detected values of the two models and the real values is quite minimal. Mask R-CNN has a high IOU value, whereas YOLOv5s has a low false-positive detection rate and fast detection speed. Our models and datasets can be used as a starting point to measure the trichome trait in other cotton species and establish a cotton trichome detection and counting system based on deep learning technology.

Recognition and localization of freshwater fish heads and tails based on lightweight neural networks

ABSTRACT The quality of fish pre-treatment processing directly affected the production competitiveness of the fish industry. The removal of heads and tails is one of the key technologies in the fish processing. This study proposed an identification method of fish head and tail, and the original YOLOV3 model was improved by replacing the backbone feature extraction network of the YOLOV3 model with the lightweight neural network MobileNetv3. Firstly, a freshwater fish image dataset was created and divided into the training, validation and test sets with the assigned ratio of 6:2:2. Next, the freshwater fish dataset was trained using the target detector YOLOV3. Finally, the average accuracy mAP (mean of Average Precision) and the average image detection time were used as the accuracy and speed indexes to evaluate the detection effect of the model. In addition, the SSD-MobileNetv3 and SSD-VGG16 were introduced into present study and they were compared with the improved algorithm. The experimental results showed that the detection speed of the YOLOV3 model with MobileNetV3 was significantly improved. The mAP of YOLOV3-MobileNetv3 model was 98.36%, the inference speed was 28.2 ms, which was 5.09%, 4.24% and 2.07% higher than the mAP of other three models (SSD-VGG16, SSD-MobileNetv3 and YOLOV3-Darknet-53), and the average detection time shortened by 86%, 9.99% and 29%, respectively. Therefore, this experimental method of head and tail of freshwater fish could achieve real-time detection and recognition of various kinds of freshwater fish, which had the great advantages of high detection accuracy and fast detection speed.

Real-time image defect detection system of cloth digital printing machine

Abstract In order to solve the surface defects such as white silk, spots and wrinkles in the process of digital printing, a surface defect detection system for printed fabrics based on accelerated robust feature algorithm was proposed. Image registration is mainly carried out through accelerated robust feature (SURF); bidirectional unique matching method is adopted to reduce mismatch points, achieve accurate image registration, and extract defect information through differential algorithm. The performance of the improved surfing algorithm is verified by using multiple images. The experimental results show that compared with the traditional template matching method, the detection accuracy of the system detection algorithm is 12% higher, and the average time is 42.81 ms shorter than the traditional template matching method. Experiments show that the improved surfing algorithm has short time and high precision. The system can meet the actual production needs. The new system can detect surface defects on printed fabrics with an accuracy of 98%. Conclusion: The algorithm has higher detection rate and faster detection speed, which can meet the needs of practical industrial applications.

Fault detection method for VSC-HVDC System with new energy sources considering fault current limiter

Voltage source converter based high voltage direct current transmission (VSC-HVDC) is widely adopted in the power system with new energy sources, but with the problem of fast rise of fault current. Magnetic saturation fault current limiter (MSFCL) suppresses the rise of short-circuit current by increasing the inductance value, which has the advantages of high speed and no additional control. However, the change of inductance will affect the detection of VSC-HVDC system fault, may leading to missed or incorrect removal of the fault and causing harm. Based on this problem, this paper proposes a fault detection method used in VSC-HVDC system considering MSFCL, which has fast detection speed, high sensitivity and reliability. The proposed method is robust to short-circuit resistance variation and can locate faults external and internal. Firstly, we establish the short-circuit equivalent circuit of VSC-HVDC system. Then we establish the differential equation and solve it, to obtain the analytical expression of voltage across MSFCL. Furthermore, we proposed a short circuit fault detection method based on the rise rate of the voltage across MSFCL. Finally we establish a four port VSC-HVDC system simulation model to prove that the method is effective. The results show that compared with traditional methods, the proposed method has faster detection speed. The proposed method is independent of the power flow direction, can identify the short circuit fault with large short-circuit resistance of 500 Ω and can discriminate the faulty branch from the healthy ones accurately.

Real-Time Detection of Apple Leaf Diseases in Natural Scenes Based on YOLOv5

Aiming at the problem of accurately locating and identifying multi-scale and differently shaped apple leaf diseases from a complex background in natural scenes, this study proposed an apple leaf disease detection method based on an improved YOLOv5s model. Firstly, the model utilized the bidirectional feature pyramid network (BiFPN) to achieve multi-scale feature fusion efficiently. Then, the transformer and convolutional block attention module (CBAM) attention mechanisms were added to reduce the interference from invalid background information, improving disease characteristics’ expression ability and increasing the accuracy and recall of the model. Experimental results showed that the proposed BTC-YOLOv5s model (with a model size of 15.8M) can effectively detect four types of apple leaf diseases in natural scenes, with 84.3% mean average precision (mAP). With an octa-core CPU, the model could process 8.7 leaf images per second on average. Compared with classic detection models of SSD, Faster R-CNN, YOLOv4-tiny, and YOLOx, the mAP of the proposed model was increased by 12.74%, 48.84%, 24.44%, and 4.2%, respectively, and offered higher detection accuracy and faster detection speed. Furthermore, the proposed model demonstrated strong robustness and mAP exceeding 80% under strong noise conditions, such as exposure to bright lights, dim lights, and fuzzy images. In conclusion, the new BTC-YOLOv5s was found to be lightweight, accurate, and efficient, making it suitable for application on mobile devices. The proposed method could provide technical support for early intervention and treatment of apple leaf diseases.

Aptamer-based nanointerferometer enables amplification-free ultrasensitive detection and differentiation of SARS-CoV-2 variants

The state-of-the-art SARS-CoV-2 detection methods include qRT-PCR and antibody-based lateral flow assay (LFA) point-of-care tests. Despite the high sensitivity and selectivity, qRT-PCR is slow, expensive and needs well-trained operators. On the other extreme, LFA suffers from low sensitivity albeit its fast detection speed, low detection cost and ease of use. Therefore, the continuing COVID-19 pandemic calls for a SARS-CoV-2 detection method that is rapid, convenient and cost-effective without compromise in sensitivity. Here we provide a proof-of-principle demonstration of an optimized aptamer-based nanointerferometer that enables rapid and amplification-free detection of SARS-CoV-2 spike protein-coated pseudovirus directly from human saliva with the limit of detection (LOD) of about 400 copies per mL. This LOD is on par with that of qRT-PCR, making it 1000 to 100,000-fold more sensitive than commercial LFA tests. Using various combinations of negative selections during the screens for the aptamer targeting the receptor binding domain of the spike protein of SARS-CoV-2, we isolated two aptamers that can distinguish the Omicron and Delta variants. Integrating these two aptamers with LFA strips or the nanointerferometer sensors allows both detection and differentiation of the Omicron and Delta variants which has the potential to realize rapid triage of patients infected different SARS-CoV-2 variants.

A Lightweight YOLOv4-EDAM Model for Accurate and Real-time Detection of Foreign Objects Suspended on Power Lines

Foreign objects suspended on transmission and distribution lines like nests, kites, balloons, and trash, etc. may lead discharge faults and greatly affect the safety of power grid. How to perform accurate and timely detection of these foreign objects is an urgent problem. This paper proposes a lightweight YOLOv4 model with embedded dual attention mechanism (YOLOv4-EDAM) to detect foreign objects from visible images, using MobileNetV2 embedded with the squeeze and excitation networks (SENet) to replace the CSPDarkNet53 feature extraction network, the depthwise separable convolutions (DSC) to replace the standard convolutions in SPP and PANet module, and embedding the convolutional block attention module (CBAM) into SPP and PANet module to improve the detection accuracy. Firstly, an image dataset was constructed using inspection images and public datasets, and expanded by Poisson blending and some data augmentation methods. The denoising convolutional neural network (DnCNN) was applied for image preprocessing. Next, the lightweight YOLOv4-EDAM model was trained combining Mosaic data enhancement, cosine annealing and label smoothing skills. Several detection cases were carried out and the experimental results show that the proposed model has a high accuracy with the mean average precision (mAP) of 96.71%, and a fast detection speed with the frames per second (FPS) of 45, whose overall performance is better than other object detection models. This study offers a reference for power line inspection and provides a possible way to deploy edge computing devices on unmanned aerial vehicles.

Tomato recognition and location algorithm based on improved YOLOv5

In order to meet the requirements of target detection and location for intelligent tomato picking, a recognition and location algorithm based on improved YOLOv5s is proposed in this paper. The CARAFE module structure was used to optimize the upsampling algorithm of YOLOv5s, which increased the receptive field of the network while maintaining the lightweight. EIoU and Quality Focal Loss were introduced to optimize the loss function of the network, which solved the problem of reduced accuracy caused by uneven samples, and at the same time accelerated the training convergence speed and improved the regression accuracy. The improved model is denoted as YOLOv5s-CQE. Compared with YOLOv5s, the mAP_0.5 and mAP_0.5:0.95 of YOLOv5s-CQE was increased by 1.67 and 3.43 percentages points, respectively. The recognition accuracy of YOLOv5s-CQE model in the test set was 99.77%, which had increased by 2.40 percentages points compared with before improvement, and was 4.10, 6.47 and 3.03 percentages points higher than that of the lightweight networks YOLOv4-tiny, YOLOv5-Lite-e and YOLOv5-Lite-s, respectively. Compared with YOLOv7 and Faster RCNN, the recognition accuracy rate was increased by 1.69, 3.97 percentages points respectively. In order to improve the accuracy of positioning, distortion removal and ROI clipping were carried out on the obtained images. The accuracy of location was tested by laboratory positioning test and field picking test. The results showed the total average errors decreased by 6.65 mm compared with those before distortion removal, and the field positioning accuracy was improved by 6.67 percentages points. The experimental results showed that the algorithm in this paper had the advantages of high precision, fast detection speed and strong robustness, which provided a theoretical basis for intelligent tomato picking.

Automatic Detection and Classification of Dead Nematode-Infested Pine Wood in Stages Based on YOLO v4 and GoogLeNet

Pine wood nematode disease has harmed forests in several countries, and can be reduced by locating and clearing infested pine trees from forests. The target detection model of deep learning was utilized to monitor a pine nematode-infested wood. The detecting effect was good, but limited by low-resolution photos with poor accuracy and speed. Our work presents a staged detection and classification approach for a dead nematode-infested pine wood based using You Only Look Once version 4 (YOLO v4) and Google Inception version 1 Net (GoogLeNet), employing high-resolution images acquired by helicopter. Experiments showed that the detection accuracy of the staged detection and classification method and the method using only the YOLO v4 model were comparable for a dead nematode-infested pine wood when the amount of data was sufficient, but when the amount of data was limited the detection accuracy of the former was higher than that of the latter. The staged detection and classification method retained the fast training and detection speed of the one-stage target detection model, further improving the detection accuracy with limited data volume, and was more flexible in achieving accurate classification, meeting the needs of forest areas for pine nematode disease epidemic prevention and control.

Hierarchical fusion detection algorithm for sleep spindle detection.

Sleep spindles are a vital sign implying that human beings have entered the second stage of sleep. In addition, they can effectively reflect a person's learning and memory ability, and clinical research has shown that their quantity and density are crucial markers of brain function. The "gold standard" of spindle detection is based on expert experience; however, the detection cost is high, and the detection time is long. Additionally, the accuracy of detection is influenced by subjectivity. To improve detection accuracy and speed, reduce the cost, and improve efficiency, this paper proposes a layered spindle detection algorithm. The first layer used the Morlet wavelet and RMS method to detect spindles, and the second layer employed an improved k-means algorithm to improve spindle detection efficiency. The fusion algorithm was compared with other spindle detection algorithms to prove its effectiveness. The hierarchical fusion spindle detection algorithm showed good performance stability, and the fluctuation range of detection accuracy was minimal. The average value of precision was 91.6%, at least five percentage points higher than other methods. The average value of recall could reach 89.1%, and the average value of specificity was close to 95%. The mean values of accuracy and F1-score in the subject sample data were 90.4 and 90.3%, respectively. Compared with other methods, the method proposed in this paper achieved significant improvement in terms of precision, recall, specificity, accuracy, and F1-score. A spindle detection method with high steady-state accuracy and fast detection speed is proposed, which combines the Morlet wavelet with window RMS and an improved k-means algorithm. This method provides a powerful tool for the automatic detection of spindles and improves the efficiency of spindle detection. Through simulation experiments, the sampled data were analyzed and verified to prove the feasibility and effectiveness of this method.

A Hierarchical Accident Recognition Method for Highway Traffic Systems

With the rapid development of highway traffic systems, the accident rate of highway remains high. Once an accident occurs, it will cause serious casualties. In this paper, we proposed a highway accident identification method for real-time accident detection through high-speed monitoring. The proposed method focuses on high detecting speed and high recall for car accident. The designed system consists of four parts: first, yolov3 is used to detect the car. Second, identified car frames are put the identified car frame into the Mosse Tracker for tracking. Third, speed and future center point are estimated according to its size and pixel movement. If the speed exceeds a certain threshold or the future center points among other trackers are too close, the object frame will be put into a trained SVM model. Finally, the flow vector estimation calculation will be used to determine whether there is an accident. The system has fast detection speed, high recall and strong real-time performance. Through this system, we can send warning information to upstream car drivers and traffic management personnel to improve the efficiency of accident handling and reduce the rate of highway traffic accidents and casualties.

Multiple face mask wearer detection based on YOLOv3 approach

<div align="left"><a name="_Hlk108683337"></a><span lang="EN-US">The coronavirus disease 2019 (COVID-19) is a highly infectious disease caused by the SARS-CoV-2 coronavirus. In breaking the transmission chain of SARS-CoV-2, the government has made it compulsory for the people to wear a mask in public places to prevent COVID-19 transmission. Hence, an automated face mask detection is crucial to facilitate the monitoring process in ensuring people to wear a face mask in public. This project aims to develop an automated face and face mask detection for multiple people by applying deep learning-based object detection algorithm you only look once version 3 (YOLOv3). YOLOv3 object detection algorithm was concatenated with different backbones including ResNet-50 and Darknet-53 to develop the face and face mask detection model. Datasets were collected from online resources including Kaggle and Github and the images were filtered and labelled accordingly. The models were trained on 4393 images and evaluated based on precision, recall, mean average precision and the detection time. In conclusion, DarkNet53_YOLOv3 was chosen as the better model compared to ResNet50_YOLOv3 model with its good performance on accuracy with a mAP of 95.94% and a fast detection speed with a detection time of 50 seconds on 776 images. </span></div>

RGB-D-Based Stair Detection and Estimation Using Deep Learning.

Stairs are common vertical traffic structures in buildings, and stair detection tasks are important in environmental perception for autonomous mobile robots. Most existing algorithms have difficulty combining the visual information from binocular sensors effectively and ensuring reliable detection at night and in the case of extremely fuzzy visual clues. To solve these problems, we propose a stair detection network with red-green-blue (RGB) and depth inputs. Specifically, we design a selective module, which can make the network learn the complementary relationship between the RGB feature maps and the depth feature maps and fuse the features effectively in different scenes. In addition, we propose several postprocessing algorithms, including a stair line clustering algorithm and a coordinate transformation algorithm, to obtain the stair geometric parameters. Experiments show that our method has better performance than existing the state-of-the-art deep learning method, and the accuracy, recall, and runtime are improved by 5.64%, 7.97%, and 3.81 ms, respectively. The improved indexes show the effectiveness of the multimodal inputs and the selective module. The estimation values of stair geometric parameters have root mean square errors within 15 mm when ascending stairs and 25 mm when descending stairs. Our method also has extremely fast detection speed, which can meet the requirements of most real-time applications.