Design and implementation of the intelligent system for automatically monitoring waterbirds in Quanzhou Bar Wetland

doi:10.21203/rs.3.rs-2482123/v1

Download PDF

Research Article

Design and implementation of the intelligent system for automatically monitoring waterbirds in Quanzhou Bar Wetland

https://doi.org/10.21203/rs.3.rs-2482123/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

The abundance and quantity of waterbird species are often used to evaluate the ecological status of wetlands because most waterbirds are sensitive to the environment. Traditional methods of detecting waterbirds are not only time-consuming but also inaccurate. Some investigations may even be at risk of the natural environment, E.g., bad weather or wild animal attacks. To address this issue, we designed an intelligent waterbird automatic identification system based on Model-View-Viewmodel (MVVM) framework which can support high effectively, safe and long-time monitoring the native wetland waterbirds. To implement the system, we trained a waterbird identification model (WIM) using the YOLOv5 algorithm and deployed it on a back-end for real-time detections, species identifications, and recording counts at different time intervals. It was integrated into a WebGIS-based application, which can be helpful for user to observe the spatial distributions of waterbirds of different species and analyzing their changing laws. We employed a PostgreSQL Database to manage geospatial data and designed the corresponding data access APIs. In addition, a tool for uploading and labeling images online was implemented for pre-training, retraining, and updating the WIM. In our current system, the image Database plays a vital role and it is designed to be auto-update, which means that once our users finished uploading a new image, the pretrained WIM on the back-end will be updated automatically. Although the system is still on its initial testing phase, some results show that it works well. The identifying rate and recall of native waterbird can reach 81.00% and 99.00%, respectively. The ongoing system is able to meet the basic requirements of detecting native wetland waterbird species and record the maximum number of observations of different species within a certain time interval specified by users. And it will provide more information about for managers to better understand the waterbirds and the wetland environment.

Waterbird detection

wetland

YOLOv5

information system

image classification

There are invaluable ecological resources in Quanzhou Bay Estuary Wetland (QBEW) such as mangroves and hence it is a habitat for various waterbirds (Lin et al., 2021; Juan, et al., 2011).QBEW plays a vital role in maintaining ecological balance, since it can help maintain biodiversity, regulate runoff, improve water quality, regulate microclimate, supply food and industrial materials, and provide tourism resources. However, with the rapid economic growth, QBEW faces severe environmental problems, such as water and air pollution, overfishing, invasive alien species, and urban land encroachment. These issues seriously endanger the habitat of native waterbirds, which are sensitive to changes in the environment (Wang et al., 2022; Pöysä et al., 2021). Thus, the species and population of waterbird often regard as environmental indicators (Green and Elmberg, 2014; Beatty et al., 2014; Zhang et al., 2018; Luo et al., 2019; Fan et al., 2021). There are plenty of waterbirds wintering in QBEW every year, including 10 kinds of national first-level key protected birds such as Egretta eulophotes, Platalea minor, Pelecanus crispus, Tringa guttifer and Ciconia boyciana, and 25 kinds of national second-level key protected birds such as Cygnus columbianus, Egretta sacra, and Halcyon smyrnensis. Heron is a dominant species in QBEW area and its numbers have increased significantly following the implementation of environmental restoration policies which implemented between 2016 and 2019 (Zhang et al., 2021). Among them, the most abundant waterbird is the Egretta garzetta, accounting for 65.82% of the total. Ardea cinerea accounted for 15.46% while Ardea alba accounted for 13.91%. Besides, Alpina and Larus ridibundus have also been observed in large groups widely distributed in QBEW (Chen et al., 2017). Although lots of sparrows are observed in the peripheral farming areas and vill/ages, most of waterbirds live in mudflat mangrove areas. Many waterbirds habitats have been disturbed or destroyed due to human activities, and waterbirds diversity and population have declined, which may indicate a weakening of mangrove conservation capacity in QBEW.

Although the QBEW Nature Reserve carried out lots of measures on the ecological restoration of mangroves and waterbird habitats, and even deployed many monitoring videos. It is still difficult tell the total number of different species of waterbirds, and it is hard to get a dynamic and comprehensive picture of changes in waterbirds. As an indicator of biodiversity, waterbirds are always deserving of being surveyed for biodiversity conservation. Recording the species and their population is crucial to understanding wetland environments. Managers can use the recorded data to see the distributions of different species of waterbirds, changes in the population of each species ,and analyze the relationship between waterbirds and their habitats. Therefore, to better protect the waterbirds and wetland resources in QBEW, we need to design and implement an intelligent waterbird automatic identification system that can automatically detect, identify, and monitor waterbirds in real-time.

With the rapidly development of deep learning-based image and video recognition algorithms, AlexNet, ResNet, Unet, and FCN algorithm showed great performance on objects identifications (Krizhevsky et al., 2012; Ronneberger et al., 2015; Long et al., 2015; Villa et al., 2018). These deep learning algorithms obtain huge success in many fields such as medical image segmentation, animal identification, face identification and remote sensing image recognition. R-CNN (Region-Convolutional Neural Network, Liu et al., 2020) series, SSD (Single Shot MultiBox Detector) (Liu et al., 2016), and YOLO (You Only Look Once) series are three widely used (Girshick et al., 2015) deep learning algorithms in the real-time applications of objects identification. Fast R-CNN (Faster R-CNN, Morera et al., 2020) is a new version of R-CNN which can recognize target objects faster than R-CNN with high accuracy, but the model training process is still very time-consuming, and it needs more labeled samples (Yan et al., 2021) and computation capability (Yan et al., 2021) compared with YOLO series algorithms (Wang et al., 2022; Yang et al., 2020; Chou et al., 2022; Lei et al., 2022).The SSD series algorithms are also time consuming relative to YOLO series algorithms. YOLOv5 is the latest version of the current YOLO series algorithms, which is a developed version of YOLO4. Although the recognition accuracy of both models are similar, YOLOv5 is much less time-consuming than YOLOv4, and its model size is almost only 10% of YOLOv4. Among these algorithms, YOLOv5 has the lightest model size, and it needs less computer resource. Moreover, the detection speed of YOLOv5 is fastest, up to 140 FPS. As Table 1 listed, YOLOv5 has obvious advantages and can meet the demand of real-time detecting in QBEW. Considering the current situation of QBEW and the trade-off between recognition accuracy and speed, we choose YOLOv5, and based on this, we designed and built an intelligent system for automatic monitoring of water birds.

Table 1

The comparison of deep learning models
Model	Advantages
Model	Flexible	Light model size	Low training time	High accuracy	Fast detecting speed	Fewer training sets
YOLOv5	✔	✔	✔	✔	✔	✔
YOLOv4	✔			✔		✔
SSD				✔
Faster R-CNN				✔

2.1 Study area

QBEW is located on the southeast coast of Fujian Province, China and has a subtropical marine monsoon climate. The latitude of our study area is from 24°47'23.2″ to 24°59'48.6″N, and the longitude is from 118°37'40.8″ to 118°42'40.0E.. Luoyang River and Jinjiang River flow through QBEW. QBEW has a warm and humid climate, with an average annual temperature of 19.9 ℃, a precipitation of 1095.4 mm and a relative humidity of 78%. QBEW are abundant in biological resources. The main plant in this area is mangrove, including Aegiceras corniculatum, Kandelia candel, and Avicennia marina (Li et al,. 2015). The area of mangrove increased from 17.1hm² to nearly 466.7hm².Among them, the artificial mangrove forest of Luoyang Estuary became the largest and best-growing continuous mangrove forest in the southeast coast of China. Besides,there are more than 213 species of waterbirds recorded in QBEW, including Egretta eulophotes, Platalea minor, Pelecanus crispus. However, waterbirds populations have declined due to the impact of human activities.

2.2 Structure of our waterbird monitoring system

The framework of our system is based on the Model-View-View-Model (MVVM). It deals with the Massive View Controller problem that the MVC model cannot solve. The problem indicates that when the Controller objects exceed 150 lines, the Controller will be inefficient in dealing with objects. (Aljamea et al., 2018). As shown in Fig. 1, the front-end provides services including waterbirds image recognition, waterbirds monitoring and model training. Users can upload their images or videos and then get the corresponding identification results, and they can also specify the service to detect the waterbirds of different video streams. Besides, our system provides a service for online model training, which makes it easy for non-experts to train a new model by setting different parameters and training sets. Online model training module can help auto build or rebuild model by specifying training sets. The training set not only includes the original waterbird images but also their corresponding label files in XML format. The XML label files of training sets contain information about the location and species of waterbirds in the images (Haifeng, D. et al., 2020).

When the system run the identification function, the back end started to identify images or videos, and then record the results and log into the back end Database. Before training a new waterbird identification model, users can first set the hyperparameters or choose loss function, activation function, optimization, and batch size, learning rate etc. Once a new model was built, its performance will be evaluated and users can retrain the model if they were not satisfied with the evaluation results.

2.3 Data storage and access

To improve the recognition accuracy of monitoring waterbirds, it is necessary to build an image database of native waterbirds. Most waterbirds images in our current database are from the Bird Watching Association of Fujian (FBWA).The open data set CUB-200-2011 is also included for training WIM (Ren et al., 2020). Besides, the database can recored the user-specified training sets, user-specified training sets, images to be recognized and their corresponding results. As shown in Fig. 2, we implemented a tool for interactive labeling, which supports users to frame and label the waterbirds in an image.

2.4 Design of the storage access interface of waterbird identification system

The storage access interface we designed includes Users, Dataconfigs, Training Module, and Identification Module. Before logging into the system, users need to select their identity, visitor or administrator, since the system provides different operational services to the two roles. Visitors or administrators can call the basic functions, such as waterbird identification, real-time waterbird monitoring and model training through the interfaces. In addition, administrators can also update the database and even revise models. When start running the identification procedure, the identification module will retrieve data from the database and then locate and predict the waterbird species. Users need to train their model first, in which they should create training sets including training images and XML files, and then store the basic information into the Dataconfigs. The “src_img_url” and "res_img_url" are the input and output directories of model traing and predicting, respectively. Parameters setting and training set building in the model training module are integrated into the front-end, and once posted the back-end server starts to train the model and record corresponding information into the Dataconfigs.

Our implemented prototype system support dynamic monitoring and supervisory management, which can help protect and maintain the biodiversity of wetland water birds. Real-time waterbird identification results can be automatically obtained and recorded. As shown in Fig. 4, users can choose a camera on the left control panel, and then the real-time identification output video will be shown on the right pop-up windows. Users can also zoom in or out of the video frame and change the focus dynamically through the remote video control panel. Once the camera capture pictures contains waterbird and the back end algorithm successfully detects waterbirds, it will record their species and numbers with time and location into the database. With many cameras deployed in the QBEW, most of native waterbirds can be observed. However, since full-coverd real-time waterbirds identifications require lots of GPU resources, currently our system can only do real-time counting of a single selected camera, but with the optimization of algorithms and hardware, it is believed that real-time monitoring of cameras covering the entire protected area can be realized in the near future. And then the composition of waterbirds populations in an area change significantly, managers can analyze the waterbirds information recorded in the Database, getting a real-time picture of the area and take appropriate measures.

To evaluate the performance of the WIM, we employed several metrics, Glou, objectness, Classification, Precision, Recall, and the mean Average Precision (mAP) (Du, 2018; Huang et al., 2018; Hou et al., 2019; Zhang et al., 2022), to evaluate the model performance. GIou is a modification of IOU to measure the extent to which the target object is inside the box (Rezatofighi et al.,2019). Objectness is a classification loss of foreground and background prediction of positive and negative samples. Classification is used to calculate the classification loss of positive samples. GIou loss, Objectness loss and Classification loss are used to update the gradients so that the model can better learn the features of the training set. These three values reflect the process of model training, when they are close to 0, it means the model training is completed.

Precision can evaluate the algorithm's positioning and target detection capabilities accurately and reasonably. The Recall is the probability of being predicted to be positive in an actual positive sample (Kalaivani et al.,2022). mAP was used to assess the performance of the algorithm, which was appropriate for the multi-label image classification task in this study. [email protected] is the average accuracy of the model when the overlap ratio between the "predicted borders" and the "real borders" is 50% (Li et al., 2022). Precision, Recall and mAP reflect the accuracy of the model recognition, when they are closer to 1, it means that the model has high recognition accuracy. Results (Fig. 5) show the evaluation of our WIM system performance in QEBW, with a precision of over 80%, a recall rate and mAP of over 90%. It can be seen that the loss of GIOU, Objectness and Classification are under 0.02, which means that the model has fully learned the features of the training set. These metrics indicate that the WIM can be well used to monitor waterbirds.

We designed and implemented a system that can help users to obtain the real-time population numbers of different species of waterbirds, thus reflecting the current status of wetlands. It is useful for reducing field survey work and avoid potential risks such as weather changes, wildlife attacks and natural disasters. In generally. the advantages of our system are concluded as follows: 1) It can automatically recognize the species of native waterbirds and record its numbers. 2) Users could obtain real-time waterbirds recognition results if they specified and manipulate the current surveillance camera. 3) It supports uploading files of pictures or videos by users for identifictions and the results will be released immediately. 4) The online training module we designed will facilities the parameters settings of YOLOv5 and can improve the update efficiency of the WIM. In addition, some data analysis modules will be integrated into the system to summarize or visualize the monitoring results, which will be helpful to the scientific assessments of the ecological quality of wetlands. Besides, we will continuously update the image database by adding labeled image samples for continuously improve performance of the WIM. One of our visions is to integrate some spatiotemporal processes modeling algorithms into the current system. For example, spatiotemporal weighted regression to explore the non-stationarity relationships between the numbers of different waterbird species and multiple environmental factors variables.

There are also several disadvantages of our current system. First, the monitor hardware ,such as cameras and servers, need to be upgraded. Low quality of video leads to low recognition rate because of the captured blurred images of waterbirds. Although we trained blurred surveillance pictures, the training set essentially become low-quality, which will limit the performances of our system. In addition, the current real-time detections still have some delay about few seconds. More efforts are needed in both hardware and software in the future work.

data availability statement:

The figures (1-5) in the paper are all made by us.

Authors’ Contributions:

Dehui Wang developed the system and wrote the paper. Xiang Que developed the system and provided ideas. Jinfu Liu, Ruohai Chen, and Yiming Li collectted data. Yingqian Zhang revised the paper

Competing interests: There are no financial and non-financial competing interests.

Fundings:

Natural Science Foundation of Fujian Province of China under Grant 2021J05030 and 2021L3003. Central Guidance Local Science and Technology Development Project of China under Grant 2020L3006. Provincial Science and Technology Innovation Key Projects in Fujian Province of China under Grant 2021G02007. Special Foundation for Science and Technology Innovation of Fujian Agriculture and Forestry University under Grant CXZX2020149C and KCX21F33A.

Ethical Statement:

All authors have read, understood, and have complied as applicable with the statement on "Ethical responsibilities of Authors.

Aljamea, M., & Alkandari, M. (2018). MMVMi: A validation model for MVC and MVVM design patterns in iOS applications. IAENG Int. J. Comput. Sci, 45(3), 377-389.
August, T., Harvey, M., Lightfoot, P., Kilbey, D., Papadopoulos, T., & Jepson, P. (2015). Emerging technologies for biological recording. Biological Journal of the Linnean Society, 115(3), 731-749.
Beatty, W. S., Webb, E. B., Kesler, D. C., Raedeke, A. H., Naylor, L. W., & Humburg, D. D. (2014). Landscape effects on mallard habitat selection at multiple spatial scales during the non-breeding period. Landscape ecology, 29(6), 989-1000.
Chen, R.H., Lin, W.D., Huang, L, Ji, J.F (2017). Analysis of bird community diversity in the mangrove wetlands of Quanzhou Bay estuary. Journal of Quanzhou Normal University,35(2):1320.DOI:10.16125/j.cnki.1009-8224.2017.02.003.
Chang, Y. H., & Zhang, Y. Y. (2022). Deep Learning for Clothing Style Recognition Using YOLOv5. Micromachines, 13(10), 1678.
Chou, P. Y., Lin, C. H., & Kao, W. C. (2022). A Novel Plug-in Module for Fine-Grained Visual Classification. arXiv preprint arXiv:2202.03822.
Du, J. Understanding of Object Detection Based on CNN Family and YOLO. In Proceedings of the 2nd International Conference on Machine Vision and Information Technology (CMVIT), Hong Kong, China, 23–25 February 2018.
Fan, J., Wang, X., Wu, W., Chen, W., Ma, Q., & Ma, Z. (2021). Function of restored wetlands for waterbird conservation in the Yellow Sea coast. Science of the Total Environment, 756, 144061.
Girshick, R. (2015). Fast r-cnn. In Proceedings of the IEEE international conference on computer vision (pp. 1440-1448).
Green, A.J., Elmberg, J., 2014. Ecosystem services provided by waterbirds. Biol. Rev. 89, 105–122.
H. Rezatofighi, N. Tsoi, J. Gwak, A. Sadeghian, I. Reid and S. Savarese (2019). "Generalized Intersection Over Union: A Metric and a Loss for Bounding Box Regression," 2019 CVPR, pp. 658-666, doi: 10.1109/CVPR.2019.00075.
Haifeng, D., & Siqi, H. (2020, September). Natural scene text detection based on YOLO V2 network model. In journal of physics: conference series, 1634 (1), 12-13.
Hou, Q.; Cheng, M.-M.; Hu, X.; Borji, A.; Tu, Z.; Torr, P.H.S. (2019). Deeply Supervised Salient Object Detection with Short Connections. IEEE Trans. Pattern Anal. Mach. Intell., 41, 815–828.
Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017). Densely connected convolutional networks. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 4700-4708).
Huang, R.; Pedoeem, J.; Chen, C. YOLO-LITE: A Real-Time Object Detection Algorithm Optimized for Non-GPU Computers. In Proceedings of the IEEE International Conference on Big Data (Big Data), Seattle, WA, USA, 10–13 December 2018; pp. 2503–2510.
Juan, Z., Bin, C., Wei-Wei, Y., & Hao, H. (2011). Study on coastal wetland habitat quality evaluation in Quanzhou Bay, Fujian, China. Acta Ecologica Sinica, 31(5), 264-270.
Kalaivani, V., & Chanthiya, P. (2022). A novel custom optimized convolutional neural network for a satellite image by using forest fire detection. Earth Science Informatics, 1-11.
Kentaro, N. (2021). Browser-Based Manipulation of Virtual Objects Through MVVM Architecture with Data Binding. In 2021CANDAR (pp. 134-140). IEEE.
Kim, J. A., Sung, J. Y., & Park, S. H. (2020, November). Comparison of Faster-RCNN, YOLO, and SSD for real-time vehicle type recognition. In 2020 IEEE international conference on consumer electronics-Asia (ICCE-Asia) (pp. 1-4). IEEE.
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017). Imagenet classification with deep convolutional neural networks. Communications of the ACM, 60(6), 84-90.
Lasseck, M. (2018). Audio-based Bird Species Identification with Deep Convolutional Neural Networks. CLEF (working notes), 2125.
Lei, F., Tang, F., & Li, S. (2022). Underwater Target Detection Algorithm Based on Improved YOLOv5. Journal of Marine Science and Engineering, 10(3), 310.
Li, B., & Li, J. (2022). Methods for landslide detection based on lightweight YOLOv4 convolutional neural network. Earth Science Informatics, 1-11.
Lin, J, Hong, Y, Lin, Z.W., Que, X., Liu, J.F., & Lian, H.F. (2021).Spatiotemporal dynamics and its driving mechanism of the Quanzhou Bay Estuary Wetland, Fujian Province of eastern China. Journal of Beijing Forestry University, 43(6),75-82.
Li, W.; Cui, L.; Zhang, M.; Wang, Y.; Zhang, Y.; Lei, Y.; Zhao, X (2015). Effect of mangrove restoration on crab burrow density in Luoyangjiang Estuary, China. For. Ecosyst. 2, 263–27
Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C. Y., & Berg, A. C. (2016, October). Ssd: Single shot multibox detector. In European conference on computer vision (pp. 21-37). Springer, Cham.
Liu, Y., Lu, B., Peng, J., & Zhang, Z. (2020). Research on the use of YOLOv5 object detection algorithm in mask wearing recognition. World Scientific Research Journal, 6(11), 276-284.
Li, Y., Zhou, H., & Zhao, Y. (2022, May). The Fine-Grained Recognition of Bird Images Based on Joint Semantic Components and ResNet. In 2022 5th International Conference on Artificial Intelligence and Big Data (ICAIBD) (pp. 406-410). IEEE.
Long, J., Shelhamer, E., & Darrell, T. (2015). Fully convolutional networks for semantic segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 3431-3440).
Luo, K., Wu, Z., Bai, H., & Wang, Z. (2019). Bird diversity and waterbird habitat preferences in relation to wetland restoration at Dianchi Lake, south-west China. Avian Research, 10(1), 1-12.
Morera, Á., Sánchez, Á., Moreno, A. B., Sappa, Á. D., & Vélez, J. F. (2020). SSD vs. YOLO for detection of outdoor urban advertising panels under multiple variabilities. Sensors, 20(16), 4587.
Niemi, J., & Tanttu, J. T. (2020). Deep learning–based automatic bird identification system for offshore wind farms. Wind Energy, 23(6), 1394-1407.
Orimoloye, I. R., Mazinyo, S. P., Kalumba, A. M., Nel, W., Adigun, A. I., & Ololade, O. O. (2019). Wetland shift monitoring using remote sensing and GIS techniques: landscape dynamics and its implications on Isimangaliso Wetland Park, South Africa. Earth Science Informatics, 12(4), 553-563.
Pöysä, H., & Linkola, P. (2021). Extending temporal baseline increases understanding of biodiversity change in European boreal waterbird communities. Biological Conservation, 257, 109139.
Ren, P., Wang, L., Fang, W., Song, S., & Djahel, S. (2020). A novel squeeze YOLO-based real-time people counting approach. International Journal of Bio-Inspired Computation, 16(2), 94-101.
Ronneberger, O., Fischer, P., & Brox, T. (2015, October). U-net: Convolutional networks for biomedical image segmentation. In International Conference on Medical image computing and computer-assisted intervention (pp. 234-241). Springer, Cham.
Sindhwani, N., Verma, S., Bajaj, T., & Anand, R. (2021). Comparative analysis of intelligent driving and safety assistance systems using YOLO and SSD model of deep learning. International Journal of Information System Modeling and Design (IJISMD), 12(1), 131-146.
Villa, M., Dardenne, G., Nasan, M., Letissier, H., Hamitouche, C., & Stindel, E. (2018). FCN-based approach for the automatic segmentation of bone surfaces in ultrasound images. International journal of computer assisted radiology and surgery, 13(11), 1707-1716.
Wang, C., Zhou, Y., Zhang, H., Li, Y., Liu, H., & Dong, B. (2022). Study on the rare waterbird habitat networks of a new UNESCO World Natural Heritage site based on scenario simulation. Science of The Total Environment, 843, 157058.
Wang, C., Zhou, Y., Zhang, H., Li, Y., Liu, H., & Dong, B. (2022). Study on the rare waterbird habitat networks of a new UNESCO World Natural Heritage site based on scenario simulation. Science of The Total Environment, 843, 157058.
Wang, M., & Deng, W. (2021). Deep face recognition: A survey. Neurocomputing, 429, 215-244.
Wang, Z., Jin, L., Wang, S., & Xu, H. (2022). Apple stem/calyx real-time recognition using YOLO-v5 algorithm for fruit automatic loading system. Postharvest Biology and Technology, 185, 111808.
Wu, W., Liu, H., Li, L., Long, Y., Wang, X., Wang, Z., ... & Chang, Y. (2021). Application of local fully Convolutional Neural Network combined with YOLO v5 algorithm in small target detection of remote sensing image. PloS one, 16(10), e0259283.
Yan, B., Fan, P., Lei, X., Liu, Z., & Yang, F. (2021). A real-time apple targets detection method for picking robot based on improved YOLOv5. Remote Sensing, 13(9), 1619.
Yang, G., Feng, W., Jin, J., Lei, Q., Li, X., Gui, G., & Wang, W. (2020, December). Face mask recognition system with YOLOV5 based on image recognition. In 2020 IEEE 6th International Conference on Computer and Communications (ICCC) (pp. 1398-1404). IEEE.
Zhang, W., Li, X., Yu, L., & Si, Y. (2018). Multi-scale habitat selection by two declining East Asian waterbird species at their core spring stopover area. Ecological Indicators, 87, 127-135.
Zhang, Y.M., Huang, Z.J., Li, Y., Chen, Z.H., Wang, W.Q (2021). Integrated sea-land restoration of coastal zones based on heron habitat needs. Wetland Science & Management,17(1):56-60.
Zhang, Y., Guo, Z., Wu, J., Tian, Y., Tang, H., & Guo, X. (2022). Real-Time Vehicle Detection Based on Improved YOLO v5. Sustainability, 14(19), 12274.

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Design and implementation of the intelligent system for automatically monitoring waterbirds in Quanzhou Bar Wetland

Status:

Version 1

Abstract

Figures

1. Introduction

2. Design Of The Waterbird Monitoring System

2.1 Study area

2.2 Structure of our waterbird monitoring system

2.3 Data storage and access

2.4 Design of the storage access interface of waterbird identification system

3. Results And Discussion

4. Conclusions

Declarations

References

Additional Declarations

Status:

Version 1