Faster Region Convolutional Neural Network Research Articles

A Quaternion Two‐Stream R‐CNN Network for Pixel‐Level Color Image Splicing Localization

Recently, Zhou et al. designed a two-stream faster Region-Convolutional neural networks (R-CNN) model RGB-N for color image splicing localization in CVPR2018. However, the RGB-N locates spliced regions only at block-level and ignores the entirety and inherent correlation of three channels. Therefore, an improved quaternion two-stream R-CNN model is proposed to solve these drawbacks: a mask branch combining fully convolutional network and condition random field is added for locating spliced regions at pixel-level; quaternion representation of color images is used to process color spliced images in a holistical way. In addition, feature pyramid network based on quaternion residual network is considered to extract multi-scale features for color spliced images; attention region proposal network is combined with attention mechanism and is designed to pay more attention to the spliced regions; a high-pass filter designed for image splicing detection specifically is adopted to replace steganalysis rich model filter in the RGB-N to obtain noise input for the noise stream. Experimental results on a new synthetic dataset and three standard forgery datasets demonstrate that the proposed method is superior to the existing methods in the abilities of localization, generalization, and robustness.

Application of local fully Convolutional Neural Network combined with YOLO v5 algorithm in small target detection of remote sensing image.

This exploration primarily aims to jointly apply the local FCN (fully convolution neural network) and YOLO-v5 (You Only Look Once-v5) to the detection of small targets in remote sensing images. Firstly, the application effects of R-CNN (Region-Convolutional Neural Network), FRCN (Fast Region-Convolutional Neural Network), and R-FCN (Region-Based-Fully Convolutional Network) in image feature extraction are analyzed after introducing the relevant region proposal network. Secondly, YOLO-v5 algorithm is established on the basis of YOLO algorithm. Besides, the multi-scale anchor mechanism of Faster R-CNN is utilized to improve the detection ability of YOLO-v5 algorithm for small targets in the image in the process of image detection, and realize the high adaptability of YOLO-v5 algorithm to different sizes of images. Finally, the proposed detection method YOLO-v5 algorithm + R-FCN is compared with other algorithms in NWPU VHR-10 data set and Vaihingen data set. The experimental results show that the YOLO-v5 + R-FCN detection method has the optimal detection ability among many algorithms, especially for small targets in remote sensing images such as tennis courts, vehicles, and storage tanks. Moreover, the YOLO-v5 + R-FCN detection method can achieve high recall rates for different types of small targets. Furthermore, due to the deeper network architecture, the YOL v5 + R-FCN detection method has a stronger ability to extract the characteristics of image targets in the detection of remote sensing images. Meanwhile, it can achieve more accurate feature recognition and detection performance for the densely arranged target images in remote sensing images. This research can provide reference for the application of remote sensing technology in China, and promote the application of satellites for target detection tasks in related fields.

Quantitative loosening detection of threaded fasteners using vision-based deep learning and geometric imaging theory

With the development of signal processing and deep learning, vision-based methods for loosening detection have progressed in recent years. However, existing visual inspection methods are mainly used to measure the loosening angle, which requires prior knowledge of the initial tightening states, and is only applicable to hexagon bolts and nuts with clear edges. This paper proposes a method to quantitatively calculate the length of the exposed bolt for detecting loosening using vision-based deep learning and geometric imaging theory. The newly proposed method consists of three modules: the region of interest (RoI) location, keypoint detection, and length calculation modules. The RoI location module applies a faster regional convolutional neural network (Faster-RCNN)-based deep learning algorithm to locate the exposed bolt, whereas the keypoint detection module employs a cascaded pyramid network (CPN)-based deep learning algorithm to detect five keypoints on the exposed bolt. An accurate expression is derived from the geometric imaging theory, and the coordinates of the five keypoints and camera parameters are imported into the expression to calculate the length of the exposed bolt. In the experiment, the resolution of input image was 640 × 640 pixels, and the test speed of the proposed method was approximately 200 milliseconds per image. Experimental results showed an average measurement error of 0.61 mm using our method, outperforming other measurement methods and the state-of-the-art networks of human pose estimation. Moreover, the robustness of the proposed method was validated by varying the shooting conditions, bolt sizes and colours. Finally, some factors influencing detection and measurement accuracies are discussed, and prospects for further research are proposed.

Development and application of a detection platform for colorectal cancer tumor sprouting pathological characteristics based on artificial intelligence

ObjectiveTumor sprouting can reflect independent risk factors for tumor malignancy and a poor clinical prognosis. However, there are significant differences and difficulties associated with manually identifying tumor sprouting. This study used the Faster region convolutional neural network (RCNN) model to build a colorectal cancer tumor sprouting artificial intelligence recognition framework based on pathological sections to automatically identify the budding area to assist in the clinical diagnosis and treatment of colorectal cancer. MethodsWe retrospectively collected 100 surgical pathological sections of colorectal cancer from January 2019 to October 2019. The pathologists used LabelImg software to identify tumor buds and to count their numbers. Finally, 1,000 images were screened, and the total number of tumor buds was approximately 3,226; the images were randomly divided into a training set and a test set at a ratio of 6:4. After the images in the training set were manually identified, the identified buds in the 600 images were used to train the Faster RCNN identification model. After the establishment of the artificial intelligence identification detection platform, 400 images in the test set were used to test the identification detection system to identify and predict the area and number of tumor buds. Finally, by comparing the results of the Faster RCNN system and the identification information of pathologists, the performance of the artificial intelligence automatic detection platform was evaluated to determine the area and number of tumor sprouting in the pathological sections of the colorectal cancers to achieve an auxiliary diagnosis and to suggest appropriate treatment. The selected performance indicators include accuracy, precision, specificity, etc. ROC (receiver operator characteristic) and AUC (area under the curve) were used to quantify the performance of the system to automatically identify tumor budding areas and numbers. ResultsThe AUC of the receiver operating characteristic curve of the artificial intelligence detection and identification system was 0.96, the image diagnosis accuracy rate was 0.89, the precision was 0.855, the sensitivity was 0.94, the specificity was 0.83, and the negative predictive value was 0.933. After 400 test sets, pathological image verification showed that there were 356 images with the same positive budding area count, and the difference between the positive area count and the manual detection count in the remaining images was less than 3. The detection system based on tumor budding recognition in pathological sections is comparable to that of pathologists’ accuracy; however, it took significantly less time (0.03±0.01)s for the pathologist (13±5)s to diagnose the sections with the assistance of the AI model. ConclusionThis system can accurately and quickly identify the tumor sprouting area in the pathological sections of colorectal cancer and count their numbers, which greatly improves the diagnostic efficacy, and effectively avoids the need for confirmation by different pathologists. The use of the AI reduces the burden of pathologists in reading sections and it has a certain clinical diagnosis and treatment value.

A new oil spill detection algorithm based on Dempster-Shafer evidence theory

Features of oil spills and look-alikes in polarimetric synthetic aperture radar (SAR) images always play an important role in oil spill detection. Many oil spill detection algorithms have been implemented based on these features. Although environmental factors such as wind speed are important to distinguish oil spills and look-alikes, some oil spill detection algorithms do not consider the environmental factors. To distinguish oil spills and look-alikes more accurately based on environmental factors and image features, a new oil spill detection algorithm based on Dempster-Shafer evidence theory was proposed. The process of oil spill detection taking account of environmental factors was modeled using the subjective Bayesian model. The Faster-region convolutional neural networks (RCNN) model was used for oil spill detection based on the convolution features. The detection results of the two models were fused at decision level using Dempster-Shafer evidence theory. The establishment and test of the proposed algorithm were completed based on our oil spill and look-alike sample database that contains 1 798 image samples and environmental information records related to the image samples. The analysis and evaluation of the proposed algorithm shows a good ability to detect oil spills at a higher detection rate, with an identification rate greater than 75% and a false alarm rate lower than 19% from experiments. A total of 12 oil spill SAR images were collected for the validation and evaluation of the proposed algorithm. The evaluation result shows that the proposed algorithm has a good performance on detecting oil spills with an overall detection rate greater than 70%.

Deep Learning-Based CT Imaging in Diagnosing Myeloma and Its Prognosis Evaluation.

Imaging examination plays an important role in the early diagnosis of myeloma. The study focused on the segmentation effects of deep learning-based models on CT images for myeloma, and the influence of different chemotherapy treatments on the prognosis of patients. Specifically, 186 patients with suspected myeloma were the research subjects. The U-Net model was adjusted to segment the CT images, and then, the Faster region convolutional neural network (RCNN) model was used to label the lesions. Patients were divided into bortezomib group (group 1, n = 128) and non-bortezomib group (group 2, n = 58). The biochemical indexes, blood routine indexes, and skeletal muscle of the two groups were compared before and after chemotherapy. The results showed that the improved U-Net model demonstrated good segmentation results, the Faster RCNN model can realize the labeling of the lesion area in the CT image, and the classification accuracy rate was as high as 99%. Compared with group 1, group 2 showed enlarged psoas major and erector spinae muscle after treatment and decreased bone marrow plasma cells content, blood M protein, urine 24 h light chain, pBNP, ß-2 microglobulin (β2MG), ALP, and white blood cell (WBC) levels (P < 0.05). In conclusion, deep learning is suggested in the segmentation and classification of CT images for myeloma, which can lift the detection accuracy. Two different chemotherapy regimens both improve the prognosis of patients, but the effects of non-bortezomib chemotherapy are better.

Prediction of penetration based on infrared thermal and visual images during pulsed GTAW process

Aiming at improving the recognition accuracy and robustness of the penetration state recognition model, a Dual-input Faster R-CNN (region-convolutional neural network) model with the input of original infrared thermal (IR) and visual (CCD) image was established. For avoiding the negative effects on recognition caused by arc flicker inside CCD image and irrelevant information inside background, synchronous feature extraction, convolutional descriptor selection and recognition based on synthetic features were designed in model. Since the industrial personal computer running the model is usually not equipped with powerful GPU and enough memory space, and sometimes it needs to train a specific model according to the new data set, the model is required to have high recognition accuracy, low recognition time, short training time and small occupation space. By sharing RPN (region proposal network) and ROI (region of interest) Pooling Layer inside Faster R-CNN and introducing Label-integrated Layer, the above requirements can be greatly met. The recognition accuracy of convolutional descriptor selection assisted Dual-input Faster RCNN reached more than 95%, while the recognition time for each IR&CCD-image data pair was less than 270 ms.

Faster Region-Convolutional Neural network oriented feature learning with optimal trained Recurrent Neural Network for bone age assessment for pediatrics

This paper tactics to develop the novel Tanner-Whitehouse 3 (TW3)-based automated Bone Age Assessment (BAA) model for children with the assistance of Faster Region- Convolutional Neural Network (Faster R-CNN) and Optimal trained Recurrent Neural Network (RNN). The proposed model covers three main stages: (a) Segmentation, (b) Faster R-CNN-based feature learning, and (c) optimal classification. Further, the segmentation of those regions is performed by adaptive Otsu thresholding. A Faster R-CNN is built to learn the features from the segmented regions. Once the features are extracted from the pooling layers of Faster R-CNN, it is subjected to the RNN. As a modification to RNN, the training weight is optimized by the Average Fitness-based Sun Flower Optimization (AF-SFO), and the optimized network predicts the age of the bone. The experimental evaluation of the proposed model over a set of images collected manually and publically shows its superior performance when compared to the state-of-the-art techniques.

Region of interest-based predictive algorithm for subretinal hemorrhage detection using faster R-CNN.

Macular edema (ME) is an essential sort of macular issue caused due to the storing of fluid underneath the macula. Age-related Macular Degeneration (AMD) and diabetic macular edema (DME) are the two customary visual contaminations that can lead to fragmentary or complete vision loss. This paper proposes a deep learning-based predictive algorithm that can be used to detect the presence of a Subretinal hemorrhage. Region Convolutional Neural Network (R-CNN) and faster R-CNN are used to develop the predictive algorithm that can improve the classification accuracy. This method initially detects the presence of Subretinal hemorrhage, and it then segments the Region of Interest (ROI) by a semantic segmentation process. The segmented ROI is applied to a predictive algorithm which is derived from the Fast Region Convolutional Neural Network algorithm, that can categorize the Subretinal hemorrhage as responsive or non-responsive. The dataset, provided by a medical institution, comprised of optical coherence tomography (OCT) images of both pre- and post-treatment images, was used for training the proposed Faster Region Convolutional Neural Network (Faster R-CNN). We also used the Kaggle dataset for performance comparison with the traditional methods that are derived from the convolutional neural network (CNN) algorithm. The evaluation results using the Kaggle dataset and the hospital images provide an average sensitivity, selectivity, and accuracy of 85.3%, 89.64%, and 93.48% respectively. Further, the proposed method provides a time complexity in testing as 2.64s, which is less than the traditional schemes like CNN, R-CNN, and Fast R-CNN.

A Mountain Summit Recognition Method Based on Improved Faster R-CNN

Mountain summits are vital topographic feature points, which are essential for understanding landform processes and their impacts on the environment and ecosystem. Traditional summit detection methods operate on handcrafted features extracted from digital elevation model (DEM) data and apply parametric detection algorithms to locate mountain summits. However, these methods may no longer be effective to achieve desirable recognition results in small summits and suffer from the objective criterion lacking problem. Thus, to address these problems, we propose an improved Faster region-convolutional neural network (R-CNN) to accurately detect the mountain summits from DEM data. Based on Faster R-CNN, the improved network adopts a residual convolution block to replace the traditional part and adds a feature pyramid network (FPN) to fuse the features with adjacent layers to better address the mountain summit detection task. The residual convolution is employed to capture the deep correlation between visual and physical morphological features. The FPN is utilized to integrate the location and semantic information in the extracted feature maps to effectively represent the mountain summit area. The experimental results demonstrate that the proposed network could achieve the highest recall and precision without manually designed summit features and accurately identify small summits.

Recognition Method of Digital Meter Readings in Substation Based on Connected Domain Analysis Algorithm

Aiming at the problem that the number and decimal point of digital instruments in substations are prone to misdetection and missed detection, a method of digital meter readings in a substation based on connected domain analysis algorithm is proposed. This method uses Faster R-CNN (Faster Region Convolutional Neural Network) as a positioning network to localize the dial area, and after acquiring the partial image, it enhances the useful information of the digital area. YOLOv4 (You Only Look Once) convolutional neural network is used as the detector to detect the digital area. The purpose is to distinguish the numbers and obtain the digital area that may contain a decimal point or no decimal point at the tail. Combined with the connected domain analysis algorithm, the difference between the number of connected domain categories and the area ratio of the digital area is analyzed, and the judgment of the decimal point is realized. The method reduces the problem of mutual interference among categories when detecting YOLOv4. The experimental results show that the method improves the detection accuracy of the algorithm.

Virtual generation of pavement crack images based on improved deep convolutional generative adversarial network

To solve the problems associated with a small sample size during intelligent road detection, a virtual image set generation method for asphalt pavement cracks is proposed based on improved deep convolutional generative adversarial networks (DCGANs). First, a small set of sample crack images is collected and used as the basic image set to perform filtering, gamma transformation, and other processes, whereby crack feature recognition is enhanced. Second, a variational autoencoder (VAE) is used to encode real crack images. The latent variable values obtained from the VAE are provided as input to the DCGAN model generator, and the model hyperparameters are optimized. Subsequently, the adaptive moment estimation (Adam) optimizer is used to reoptimize the model and thereby improve the model convergence speed and generalization ability. The proposed method has the advantages of both VAE and DCGAN. Finally, a pavement crack classification detection model based on faster region convolutional neural network (Faster R-CNN) is used to evaluate the reliability of the generated crack images. The results show that the augmented dataset of the proposed method with the detection model has an average precision of 90.32%, which is higher than that of the conventional method evaluated using the same test dataset. The proposed method generates virtual crack images that are moderately identical to real ones, thereby solving the problem of insufficient image datasets of cracks in specific road sections. The method also provides data assurance for the intelligentization of pavement crack detection and the reduction of pavement maintenance costs.

Deep learning based real-time tourist spots detection and recognition mechanism.

More and more information on tourist spots is being represented as pictures rather than text. Consequently, tourists who are interested in a specific attraction shown in pictures may have no idea how to perform a text search to get more information about the interesting tourist spots. In the view of this problem and to enhance the competitiveness of the tourism market, this research proposes an innovative tourist spot identification mechanism, which is based on deep learning-based object detection technology, for real-time detection and identification of tourist spots by taking pictures on location or retrieving images from the Internet. This research establishes a tourist spot recognition system, which is a You Only Look Once version 3 model built in Tensorflow AI framework, and is used to identify tourist attractions by taking pictures with a smartphone's camera. To verify the possibility, a set of tourist spots in Hsinchu City, Taiwan is taken as an example. Currently, the tourist spot recognition system of this research can identify 28 tourist spots in Hsinchu. In addition to the attraction recognition feature, tourists can further use this tourist spot recognition system to obtain more information about 77 tourist spots from the Hsinchu City Government Information Open Data Platform, and then make dynamic travel itinerary planning and Google MAP navigation. Compared with other deep learning models using Faster region-convolutional neural networks or Single-Shot Multibox Detector algorithms for the same data set, the recognition time by the models using You Only Look Once version 3, Faster region-convolutional neural networks, and Single-Shot Multibox Detector algorithms are respectively 4.5, 5, and 9 s, and the mean average precision for each when IoU = 0.6 is 88.63%, 85%, and 43.19%, respectively. The performance experimental results of this research show the model using the You Only Look Once version 3 algorithm is more efficient and precise than the models using the Faster region-convolutional neural networks or the Single-Shot Multibox Detector algorithms, where You Only Look Once version 3 and Single-Shot Multibox Detector are one-stage learning architectures with efficient features, and Faster region-convolutional neural networks is a two-stage learning architecture with precise features.

Aorta Detection with Fetal Echocardiography Images Using Faster Regional Convolutional Neural Network (R-CNNs)

The fetal heart structure has an important role in analyzing the location of abnormalities in the heart. The aorta is one of the fetal heart structures, which has an essential part in exploring how the fetal heart is structured. To see the fetal heart structure can be seen with the help of an echocardiography tool in the form of ultrasound to see ultrasound images of the fetal heart. In ultrasound image data, detection is challenging because of its low image features, shadows, and contrast levels. So that is the first to do it yourself in one of the points of the culture in the culture in the aorta. The approach in this study uses deep learning in cases using Faster Regional Convolutional Neural Network (R-CNNs) with the R-CNNs mask method. The proposed approach has been applied to 151 ultrasound images of the fetal heart for the aortic region. The evaluation results were tested by evaluating metrics on the detection object with an mAP value of 83.71%.

An IoT‐based human detection system for complex industrial environment with deep learning architectures and transfer learning

Artificial intelligence (AI), combined with the Internet of Things (IoT), plays a beneficial role in various fields, including intelligent surveillance applications. With IoT and 5G advancement, intelligent sensors, and devices in the surveillance environment collect large amounts of data in the form of videos and images. These collected data require intelligent information processing solutions, help analyze the recorded videos and images to detect and identify various objects in the scene, particularly humans. In this study, an automated human detection system is presented for a complex industrial environment, in which people are monitored/detected from a top view perspective. A top view is usually preferred because it can provide sufficient coverage and enough visibility of a scene. This study demonstrates the applications, efficiency, and effectiveness of deep learning architectures, that is, Faster Region Convolutional Neural Network (Faster R-CNN), Single Shot MultiBox Detector (SSD), and You Only Look Once (YOLOv3), with transfer learning. Experimental results reveal that with additional training and transfer learning, the performance of all detection architectures is significantly improved. The detection results are also compared using the same data set. The deep learning architectures achieve promising results with maximum true-positive rate of 93%, 94%, and 94% for Faster-RCNN, SSD, and YOLOv3, respectively. Furthermore, a detailed study is performed on output results that highlight challenges and probable future trends.

A Comparison of vision-based and CNN-based detector for fish monitoring in complex environment

Aquaculture farming can help soften the environmental impact of overfishing by fulfilling seafood demands with farmed fishes. However, to maintain big scale farms can be challenging, even with the help of underwater cameras affixed in farm cages, because there are hours’ worth of footages to sift through, which can be a laborious task if performed manually. Vision-based system therefore could be deployed to filter useful information from video footages, automatically. This work proposed to solve the above mentioned problems by deploying the; 1) Extended UTAR Aquaculture Farm Fish Monitoring System Framework (UFFMS), being the handcrafted method, and 2) Faster Region Convolutional Neural Network (Faster RCNN), being the CNN-based method, for fish detection. These two methods could extract information about fishes from video footages. Experimental results show that for well-lit footages, Faster RCNN performs better, compared to the extended-UFFMS. However, accuracy of Faster RCNN drops drastically for non-well-lit footages, at an average of 28.57%, despite still having perfect precision scores.

Real-time nondestructive fish behavior detecting in mixed polyculture system using deep-learning and low-cost devices

Fish behavior has attracted increasing attention in global aquaculture because it provides important information about productivity and fish quality. The use of images to detect fish behavior has shown potential in aquaculture behavioral studies by providing higher spatial resolution, efficiency, and accuracy than conventional approaches such as manual measurement. In addition, it allows for more quantitative data analysis than do other methods. To date, conventional image processing approaches to monitor fish behavior have been based primarily on appearance, morphology, and color information. This approach is complex and/or time-consuming and limits the practicality of such methods in aquaculture. To address these problems, we present herein a noninvasive, rapid, low-cost procedure based on an underwater imaging system and a deep learning framework to detect fish behavior with high accuracy in a mixed polyculture system. The specific objectives of this study are (1) to design a low-cost underwater imaging system that can describe and quantify fish behavior via visual images, and (2) to develop a lightweight deep learning structure to rapidly and accurately detect fish behavior under various conditions. Toward this end, images of fish are first captured via a low-cost imaging system, following which they are preprocessed to reduce noise and enhance data information. Finally, an improved You Only Look Once version 3 Lite (YOLOv3-Lite) network with a novel backbone structure is used to improve the pooling block and loss function and thereby better recognize fish behavior. The proposed method was tested on a real dataset and produced a Precision of 0.897, a Recall of 0.884, an intersection over union of 0.892, and 240 frames per second. Furthermore, when compared with faster region-convolutional neural network, YOLO, YOLOv2, YOLOv3, and single shot multi-Box detector, the performance of each evaluation metric of the proposed method was improved by 10%–20%. This comprehensive analysis indicates that the proposed method provides state-of-the-art performance and may be used in fish farms.

An improved faster-RCNN model for handwritten character recognition

Existing techniques for hand-written digit recognition (HDR) rely heavily on the hand-coded key points and requires prior knowledge. Training an efficient HDR network with these preconditions is a complicated task. Recently, work on HDR is mainly focused on deep learning (DL) approaches and has exhibited remarkable results. However, effective detection and classification of numerals is still a challenging task due to people’s varying writing styles and the presence of blurring, distortion, light and size variations in the input sample. To cope with these limitations, we present an effective and efficient HDR system, introducing a customized faster regional convolutional neural network (Faster-RCNN). This approach comprises three main steps. Initially, we develop annotations to obtain the region of interest. Then, an improved Faster-RCNN is employed in which DenseNet-41 is introduced to compute the deep features. Finally, the regressor and classification layer is used to localize and classify the digits into ten classes. The performance of the proposed method is analyzed on the standard MNIST database, which is diverse in terms of changes in lighting conditions, chrominance, shape and size of digits, and the occurrence of blurring and noise effects, etc. Additionally, we have also evaluated our technique over a cross-dataset scenario to prove its efficacy. Experimental evaluations demonstrate that the approach is more competent and able to accurately detect and classify numerals than other recent methods.

Bolt damage identification based on orientation-aware center point estimation network

With the development of deep learning, object detection algorithms based on horizontal box are widely used in the field of damage identification. However, damages can be in any direction and position, and they are not necessarily horizontal or vertical. This article proposes a bolt damage identification network, namely, orientation-aware center point estimation network, which models a damage as a center point of its rotated bounding box. The proposed orientation-aware center point estimation network uses deep layer aggregation network to search center points and regress to all other damage properties, such as size and angle. A loss function is designed to improve the optimization efficiency of network. Orientation-aware center point estimation network is applied to bolt damage detection, and comparison with the well-known Faster Region-Convolutional Neural Network (a benchmark using horizontal bounding box) demonstrates the accuracy of the proposed method. Finally, videos were utilized to verify the capability of the proposed orientation-aware center point estimation network in real-time detection of bolt damages.

Target Detection-Based Tree Recognition in a Spruce Forest Area with a High Tree Density—Implications for Estimating Tree Numbers

Here, unmanned aerial vehicle (UAV) remote sensing and machine vision were used to automatically, accurately, and efficiently count Tianshan spruce and improve the efficiency of scientific forest management, focusing on a typical Tianshan spruce forest on Tianshan Mountain, middle Asia. First, the UAV in the sampling area was cropped from the image, and a target-labeling tool was used. The Tianshan spruce trees were annotated to construct a data set, and four models were used to identify and verify them in three different areas (low, medium, and high canopy closures). Finally, the combined number of trees was calculated. The average accuracy of the detection frame, mean accuracy and precision (mAP), was used to determine the target detection accuracy. The Faster Region Convolutional Neural Network (Faster-RCNN) model achieved the highest accuracies (96.36%, 96.32%, and 95.54% under low, medium, and high canopy closures, respectively) and the highest mAP (85%). Canopy closure affected the detection and recognition accuracy; YOLOv3, YOLOv4, and Faster-RCNN all showed varying spruce recognition accuracies at different densities. The accuracy of the Faster-RCNN model decreased by at least 0.82%. Combining UAV remote sensing with target detection networks can identify and quantify statistics regarding Tianshan spruce. This solves the shortcomings of traditional monitoring methods and is significant for understanding and monitoring forest ecosystems.