CN114387625A - Ecological biological identification method based on FPN algorithm - Google Patents
Ecological biological identification method based on FPN algorithm Download PDFInfo
- Publication number
- CN114387625A CN114387625A CN202210059996.8A CN202210059996A CN114387625A CN 114387625 A CN114387625 A CN 114387625A CN 202210059996 A CN202210059996 A CN 202210059996A CN 114387625 A CN114387625 A CN 114387625A
- Authority
- CN
- China
- Prior art keywords
- ecological
- image data
- biological
- fpn
- algorithm
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses an ecological organism identification method based on an FPN algorithm, which comprises the following steps: collecting ecological biological characteristics, collecting and classifying the collected ecological characteristics, and establishing a distributed ecological characteristic database; initiating an ecological biological identification request, and acquiring biological image data in an ecological environment according to the request; analyzing and processing the acquired biological image data, and storing the analyzed and processed biological image data; carrying out target detection on the acquired biological image data through an FPN algorithm; and comparing and identifying the obtained biological image data required to be identified with the characteristic data in the distributed ecological characteristic database. According to the invention, by setting the FPN algorithm, the high-resolution of the low-layer features and the high-semantic information of the high-layer features can be simultaneously utilized, the prediction effect is achieved by fusing the features of different layers, and the prediction is independently carried out on each fused feature layer, so that the small object detection performance is greatly improved.
Description
Technical Field
The invention relates to the technical field of biological recognition, in particular to an ecological biological recognition method based on an FPN algorithm.
Background
The aquatic organism community and the water environment have a complex and complicated mutual relationship and play an important role in water quality change. Different types of aquatic organisms have different adaptability to water body pollution, and some types are only suitable for living in clean water and are called as clean water organisms (or oligozoophorous organisms). Some aquatic organisms live in sewage and are called as sewage organisms. The survival marks of the aquatic organisms indicate the water quality change degree, so the organisms become indexes of water pollution, the water pollution condition can be evaluated through the investigation of the aquatic organisms, a plurality of aquatic organisms are sensitive to water poisoning substances, and the water quality pollution degree can also be judged through the toxicity experiment results of the aquatic organisms;
however, many existing aquatic organisms are small in size, and the existing biological identification method is poor in small object detection performance and low in identification accuracy.
Disclosure of Invention
Based on the technical problems in the background art, the invention provides an ecological biological identification method based on an FPN algorithm.
The invention provides an ecological organism identification method based on an FPN algorithm, which comprises the following steps:
s1, collecting ecological biological characteristics, collecting and classifying the collected ecological characteristics, and establishing a distributed ecological characteristic database;
s2, initiating an ecological biological identification request, and acquiring biological image data in an ecological environment according to the request;
s3, analyzing the collected biological image data and storing the analyzed biological image data;
s4, carrying out target detection on the acquired biological image data through an FPN algorithm;
s41, firstly, selecting a picture to be processed, and then preprocessing the picture;
s42, sending the processed pictures into a pre-trained feature network, namely constructing a so-called bottom-up network;
s43, constructing a corresponding top-down network;
s44, respectively performing RPN operation on the last three layers of the top-down network, namely dividing a convolution of 3x3 into two paths, and respectively connecting a convolution of 1x1 for classification and regression operation;
s45, respectively inputting the candidate ROIs obtained in the previous step to the upper surfaces of the last three layers of the top-down network, and respectively carrying out ROI Pool operation;
s46, connecting two 1024 layers of full-connection network layers on the basis of the previous step, dividing into two branches, and connecting the corresponding classification layer and regression layer;
s5 compares the obtained biological image data to be identified with the feature data in the distributed ecological feature database.
Preferably, the FPN in the step S4 includes three parts, a bottom-up path, a Top-down path and lateral connections.
Preferably, the bottom-up path: calculating different resolution characteristics in a hierarchical mode, extracting 5 hierarchical characteristics of { C1, C2, C3, C4 and C5} by adopting ResNet as a backbone, and taking four stages of { C2, C3, C4 and C5} to form a characteristic pyramid, wherein the resolution of the image is sampled to be {4,8,16 and 32 }.
Preferably, the Top-down path: starting from C5, the feature map is up-sampled by 2 times by a nearest neighbor method to obtain C5 ', C4 adjusts the channel number by 1-1 convolution to obtain C4', and the resolution of C5 'and C4' are the same, and the C4 'and the C4' can be directly added element by element, so that the feature fusion of C3 and C2 is realized iteratively, and Top-down gradually enhances small target information.
Preferably, the molar connection: each summed signature is obtained and the paper is again processed by 3x3 convolution to obtain the final signature { P2, P3, P4, P5 }.
Preferably, the ROI Pool operation in step S45, the roiploling selection: different scales of ROI, using different feature layers as input of ROI posing layer, large scale ROI can select P5 layer; the feature layer of the small-scale ROI may be selected as P4.
Preferably, the step S2 acquires biological image data, and performs preprocessing on the biological image data by using an image normalization method and an image enhancement method.
Preferably, the step S5 compares the identified biological image data and sends the compared biological image data to the distributed ecological characteristic database for storage.
According to the ecological organism identification method based on the FPN algorithm, the FPN algorithm is arranged, high-resolution of low-layer features and high-semantic information of high-layer features can be simultaneously utilized, the prediction effect is achieved by fusing the features of different layers, the prediction is independently carried out on each fused feature layer, and the small object detection performance is greatly improved.
Drawings
FIG. 1 is a schematic structural diagram of an ecological biological identification method based on FPN algorithm according to the present invention;
fig. 2 is a schematic structural diagram of the ecological biological identification method based on the FPN algorithm according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments.
Referring to fig. 1-2, the ecological biological identification method based on the FPN algorithm includes the following steps:
s1, collecting ecological biological characteristics, collecting and classifying the collected ecological characteristics, and establishing a distributed ecological characteristic database;
s2, initiating an ecological biological identification request, and acquiring biological image data in an ecological environment according to the request;
s3, analyzing the collected biological image data and storing the analyzed biological image data;
s4, carrying out target detection on the acquired biological image data through an FPN algorithm;
s41, firstly, selecting a picture to be processed, and then preprocessing the picture;
s42, sending the processed pictures into a pre-trained feature network, namely constructing a so-called bottom-up network;
s43, constructing a corresponding top-down network;
s44, respectively performing RPN operation on the last three layers of the top-down network, namely dividing a convolution of 3x3 into two paths, and respectively connecting a convolution of 1x1 for classification and regression operation;
s45, respectively inputting the candidate ROIs obtained in the previous step to the upper surfaces of the last three layers of the top-down network, and respectively carrying out ROI Pool operation;
s46, connecting two 1024 layers of full-connection network layers on the basis of the previous step, dividing into two branches, and connecting the corresponding classification layer and regression layer;
s5 compares the obtained biological image data to be identified with the feature data in the distributed ecological feature database.
In the present invention, the FPN in step S4 includes three parts, bottom-up path, Top-down path and lateral connections.
In the invention, bottom-up path: calculating different resolution characteristics in a hierarchical mode, extracting 5 hierarchical characteristics of { C1, C2, C3, C4 and C5} by adopting ResNet as a backbone, and taking four stages of { C2, C3, C4 and C5} to form a characteristic pyramid, wherein the resolution of the image is sampled to be {4,8,16 and 32 }.
In the present invention, Top-down path: starting from C5, the feature map is up-sampled by 2 times by a nearest neighbor method to obtain C5 ', C4 adjusts the channel number by 1-1 convolution to obtain C4', and the resolution of C5 'and C4' are the same, and the C4 'and the C4' can be directly added element by element, so that the feature fusion of C3 and C2 is realized iteratively, and Top-down gradually enhances small target information.
In the present invention, the later connection: each summed signature is obtained and the paper is again processed by 3x3 convolution to obtain the final signature { P2, P3, P4, P5 }.
In the present invention, the ROI Pool operation and roiploling selection in step S45: different scales of ROI, using different feature layers as input of ROI posing layer, large scale ROI can select P5 layer; the feature layer of the small-scale ROI may be selected as P4.
In the present invention, step S2 acquires biological image data, and preprocesses the biological image data using an image normalization method and an image enhancement method.
In the present invention, step S5 compares the recognized biometric image data and sends the compared biometric image data to the distributed biometric feature database for storage.
The invention comprises the following steps: collecting ecological biological characteristics, collecting and classifying the collected ecological characteristics, and establishing a distributed ecological characteristic database; initiating an ecological biological identification request, and acquiring biological image data in an ecological environment according to the request; analyzing and processing the acquired biological image data, and storing the analyzed and processed biological image data; carrying out target detection on the acquired biological image data through an FPN algorithm; firstly, selecting a picture to be processed, and then carrying out preprocessing operation on the picture; then, the processed pictures are sent into a pre-trained feature network, namely, a so-called bottom-up network is constructed; then, constructing a corresponding top-down network; next, performing RPN operation on the last three layers of the top-down network respectively, namely dividing a convolution of 3x3 into two paths, and connecting a convolution of 1x1 for classification and regression operation respectively; secondly, respectively inputting the candidate ROI obtained in the last step to the upper parts of the last three layers of the top-down network, and respectively carrying out ROI Pool operation; finally, connecting two 1024 layers of fully-connected network layers on the basis of the previous step, dividing the fully-connected network layers into two branches, and connecting corresponding classification layers and regression layers; and comparing and identifying the obtained biological image data required to be identified with the characteristic data in the distributed ecological characteristic database.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (8)
1. The ecological biological identification method based on the FPN algorithm is characterized by comprising the following steps:
s1, collecting ecological biological characteristics, collecting and classifying the collected ecological characteristics, and establishing a distributed ecological characteristic database;
s2, initiating an ecological biological identification request, and acquiring biological image data in an ecological environment according to the request;
s3, analyzing the collected biological image data and storing the analyzed biological image data;
s4, carrying out target detection on the acquired biological image data through an FPN algorithm;
s41, firstly, selecting a picture to be processed, and then preprocessing the picture;
s42, sending the processed pictures into a pre-trained feature network, namely constructing a so-called bottom-up network;
s43, constructing a corresponding top-down network;
s44, respectively performing RPN operation on the last three layers of the top-down network, namely dividing a convolution of 3x3 into two paths, and respectively connecting a convolution of 1x1 for classification and regression operation;
s45, respectively inputting the candidate ROIs obtained in the previous step to the upper surfaces of the last three layers of the top-down network, and respectively carrying out ROI Pool operation;
s46, connecting two 1024 layers of full-connection network layers on the basis of the previous step, dividing into two branches, and connecting the corresponding classification layer and regression layer;
s5 compares the obtained biological image data to be identified with the feature data in the distributed ecological feature database.
2. The FPN algorithm-based ecological biometric recognition method of claim 1, wherein the FPN in step S4 includes bottom-up path, Top-down path and relational connections.
3. The FPN algorithm-based ecological biometric recognition method according to claim 2, wherein the bottom-up path: calculating different resolution characteristics in a hierarchical mode, extracting 5 hierarchical characteristics of { C1, C2, C3, C4 and C5} by adopting ResNet as a backbone, and taking four stages of { C2, C3, C4 and C5} to form a characteristic pyramid, wherein the resolution of the image is sampled to be {4,8,16 and 32 }.
4. The FPN algorithm-based ecological biometric recognition method according to claim 2, wherein the Top-down path: starting from C5, the feature map is up-sampled by 2 times by a nearest neighbor method to obtain C5 ', C4 adjusts the channel number by 1-1 convolution to obtain C4', and the resolution of C5 'and C4' are the same, and the C4 'and the C4' can be directly added element by element, so that the feature fusion of C3 and C2 is realized iteratively, and Top-down gradually enhances small target information.
5. The FPN algorithm-based ecological biometric method according to claim 2, wherein said hierarchical connection: each summed signature is obtained and the paper is again processed by 3x3 convolution to obtain the final signature { P2, P3, P4, P5 }.
6. The FPN algorithm-based ecological biometric recognition method according to claim 1, wherein the ROI Pool operation, roiploling selection in step S45: different scales of ROI, using different feature layers as input of ROI posing layer, large scale ROI can select P5 layer; the feature layer of the small-scale ROI may be selected as P4.
7. The FPN algorithm-based ecological biometric recognition method according to claim 1, wherein said step S2 is to collect biometric image data, and to pre-process the biometric image data by using an image normalization method and an image enhancement method.
8. The FPN algorithm-based ecological biometric identification method according to claim 1, wherein the step S5 compares the identified biometric image data and sends the compared biometric image data to a distributed ecological feature database for storage.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210059996.8A CN114387625A (en) | 2022-01-19 | 2022-01-19 | Ecological biological identification method based on FPN algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210059996.8A CN114387625A (en) | 2022-01-19 | 2022-01-19 | Ecological biological identification method based on FPN algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114387625A true CN114387625A (en) | 2022-04-22 |
Family
ID=81203315
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210059996.8A Pending CN114387625A (en) | 2022-01-19 | 2022-01-19 | Ecological biological identification method based on FPN algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114387625A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114743223A (en) * | 2022-05-19 | 2022-07-12 | 澜途集思生态科技集团有限公司 | Ecological organism recognition method based on Fitness-NMS algorithm |
-
2022
- 2022-01-19 CN CN202210059996.8A patent/CN114387625A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114743223A (en) * | 2022-05-19 | 2022-07-12 | 澜途集思生态科技集团有限公司 | Ecological organism recognition method based on Fitness-NMS algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | A comprehensive review of computer-aided whole-slide image analysis: from datasets to feature extraction, segmentation, classification and detection approaches | |
| KR102108050B1 (en) | Method for classifying breast cancer histology images through incremental boosting convolution networks and apparatus thereof | |
| Ye et al. | A targeted change-detection procedure by combining change vector analysis and post-classification approach | |
| CN106570505A (en) | Method for analyzing histopathologic image and system thereof | |
| CN111985536A (en) | Gastroscope pathological image classification method based on weak supervised learning | |
| JP7422235B2 (en) | Non-tumor segmentation to aid tumor detection and analysis | |
| Ghassemi et al. | Learning and adapting robust features for satellite image segmentation on heterogeneous data sets | |
| US7606779B2 (en) | Methods and system for data aggregation of physical samples | |
| CN113674252A (en) | Histopathology image diagnosis system based on graph neural network | |
| CN113034462A (en) | Method and system for processing gastric cancer pathological section image based on graph convolution | |
| CN114693971A (en) | Classification prediction model generation method, classification prediction method, system and platform | |
| CN114429649B (en) | Target image identification method and device | |
| CN114387625A (en) | Ecological biological identification method based on FPN algorithm | |
| CN114782948A (en) | Global interpretation method and system for cervical liquid-based cytology smear | |
| Sharma et al. | Isolated handwritten digit recognition using adaptive unsupervised incremental learning technique | |
| Koresh et al. | Classification of noiseless corneal image using capsule networks | |
| CN116740418A (en) | Target detection method based on graph reconstruction network | |
| CN113128518B (en) | Sift mismatch detection method based on twin convolution network and feature mixing | |
| CN114445356A (en) | Multi-resolution-based full-field pathological section image tumor rapid positioning method | |
| CN112464802A (en) | Automatic identification method and device for slide sample information and computer equipment | |
| CN116524255A (en) | Wheat scab spore identification method based on Yolov5-ECA-ASFF | |
| Sertel et al. | Computer-aided prognosis of neuroblastoma: classification of stromal development on whole-slide images | |
| Thapa et al. | Object vs pixel-based flood/drought detection in paddy fields using deep learning | |
| CN114399793A (en) | Ecological biological identification method based on SPP-Net algorithm | |
| Lu et al. | Snowvision: Segmenting, identifying, and discovering stamped curve patterns from fragments of pottery |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |