CN114764799A - Material detection method based on Guided adsorbing - Google Patents
Material detection method based on Guided adsorbing Download PDFInfo
- Publication number
- CN114764799A CN114764799A CN202210490399.0A CN202210490399A CN114764799A CN 114764799 A CN114764799 A CN 114764799A CN 202210490399 A CN202210490399 A CN 202210490399A CN 114764799 A CN114764799 A CN 114764799A
- Authority
- CN
- China
- Prior art keywords
- construction
- anchor point
- anchor
- capital
- capital construction
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 54
- 239000000463 material Substances 0.000 title claims abstract description 24
- 238000010276 construction Methods 0.000 claims abstract description 59
- 238000000034 method Methods 0.000 claims abstract description 35
- 239000004035 construction material Substances 0.000 claims abstract description 22
- 238000004873 anchoring Methods 0.000 claims abstract description 20
- 238000013135 deep learning Methods 0.000 claims abstract description 13
- 238000011176 pooling Methods 0.000 claims abstract description 4
- 238000012549 training Methods 0.000 claims description 16
- 238000009826 distribution Methods 0.000 claims description 14
- 238000012360 testing method Methods 0.000 claims description 8
- 238000012544 monitoring process Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 8
- 230000004927 fusion Effects 0.000 abstract description 2
- 230000008901 benefit Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000004566 building material Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011425 standardization method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a method for detecting key materials for capital construction site construction based on Guided Anchoring, and relates to the technical field of detection of key materials for capital construction site construction. The method comprises the steps of utilizing a deep learning target detection model to generate anchor points for a construction material sample picture of a capital construction scene, wherein the generated anchor points are used for extracting region-of-interest pooling, and classifying and frame regression are carried out through a subsequent network structure. According to the invention, a characteristic pyramid structure is added to the fast RCNN target detection model taking ResNet50 as a frame, so that the fusion of shallow geometrical information and deep semantic information of an image is realized, and the detection effect of the model on a long-distance target is improved; the part of artificially setting the anchor points by the fast RCNN is optimized to generate the anchor points based on the input characteristics in a self-adaptive manner, so that the detection effect of the model on the construction materials in different positions and proportions in the complex environment is improved.
Description
Technical Field
The invention relates to the technical field of detection of key materials for capital construction site construction, in particular to a method for detecting key materials for capital construction site construction based on Guided Anchoring.
Background
The monitoring of the construction materials is the premise and guarantee of the engineering quality, the damage or the loss of the construction materials seriously influences the development of construction projects, and even potential safety hazards are left in the future. With the popularization of video monitoring technology and the development of computer vision algorithms, deep learning is gradually adopted in the industry to realize automatic monitoring and management of construction materials. However, the construction material has the characteristics of different shapes and sizes, and the conventional target detection algorithm usually adopts a sliding window mode to generate an anchor point frame, so that the effect on the identification of the construction material is poor.
Disclosure of Invention
Technical problem to be solved
Aiming at the defects of the prior art, the invention provides a method for detecting key materials in capital construction site construction based on Guided Anchoring, which solves the problem that the conventional target detection algorithm usually adopts a sliding window mode to generate an anchor point frame and has poor effect on the identification of construction materials.
(II) technical scheme
In order to achieve the purpose, the invention is realized by the following technical scheme: a method for detecting key materials of construction site construction of capital construction based on Guided Anchoring utilizes a deep learning target detection model to generate anchor points for construction material sample pictures of a capital construction scene, the generated anchor points are used for extracting region pooling of interest, and classification and frame regression are carried out through a subsequent network structure;
the establishment of the deep learning target detection model comprises the following steps:
s1, collecting construction material sample pictures of the capital construction scene, and making a corresponding sample label file for each picture;
s2, using the combination of fast R-CNN and FPN structure as the basic detection frame, and adopting a Guided Anchoring method to generate an anchor point frame, and establishing a deep learning target detection model;
s3, randomly dividing all the obtained capital construction scene construction key material sample pictures and corresponding sample labels thereof into a training set and a testing set;
S4, training the deep learning target detection model by using a training set to obtain a primarily trained detection model of key materials for construction in the capital construction site;
and S5, debugging the performance of the primarily trained infrastructure site construction key material detection model by using the test set, adjusting the training parameters and the detection confidence coefficient threshold according to the test result, and optimizing and solidifying the infrastructure site construction material detection model.
Preferably, the sample tag file conforms to the JSON tag file standard of the COCO data set.
Preferably, the construction scene construction key material sample picture is a picture acquired by taking construction materials of a construction operation site as a target object under various construction scenes and acquiring the left-right deviation of the monitoring camera facing the target object by 15 degrees and the shooting distance of 5-25 meters.
Preferably, the process of generating the detection anchor point by the Guided Anchoring method includes: the anchor point position prediction branch, the anchor point shape prediction branch and the generation of sparse anchor points are completed according to the local characteristics of the characteristic diagram;
the deep learning target detection model compares the output results of the anchor point position prediction branch and the anchor point shape prediction branch with a set threshold value, firstly obtains the central position of a target possibly existing on a feature map, and then predicts the most possible anchor point shape according to local features near the central position.
Preferably, the anchor point position prediction branch generates a probability map with the same size as the input feature map, the probability of the (I, j) position on the probability map is p (I, j | I), and the numerical value represents the probability that the point is the center of the anchor point;
the anchor shape prediction branch produces a two-channel output result corresponding to the relative height and relative width of the anchor, respectively.
Preferably, the Guided Anchoring method decouples the joint distribution of anchor point parameters into two independent conditional distributions, and performs information calibration by combining semantic information of an input feature map to output more accurate features for detection;
the anchor point parameters comprise anchor point positions and anchor point shapes;
the joint distribution of the anchor points, the conditional distribution of the anchor point positions and the anchor point shapes meet the following formula:
p(x，y，w，h)＝p(x，y|I)p(w，h|x，y，I)
where (x, y) represents the location of the anchor, (w, h) represents the shape of the anchor, and I represents the input picture feature.
(III) advantageous effects
Firstly, marking acquired image data of the construction material on the construction site, and then learning deep semantic information of the characteristics of the construction material on the construction site by using a targeted detection model based on Guided Anchoring; and detecting on the image of the infrastructure site by using the trained model, predicting the position of the construction material in the image, predicting the detection confidence coefficient of the corresponding position, and finally removing the overlapped detection frames according to the set overlapped threshold value to finish the detection.
The method can realize automatic detection of the construction materials on the infrastructure site, has the advantages of high accuracy, good stability, strong anti-interference capability, high universality and the like, has good robustness, and can be applied to an intelligent supervision system on the infrastructure site.
The invention has the beneficial effects that:
1) compared with the traditional detection method for the construction materials on the construction site, the method has the advantages of high accuracy, good robustness and universality to various construction environments;
2) according to the method, a characteristic Pyramid structure (FPN) is added to a Faster RCNN target detection model taking ResNet50 as a frame, so that the fusion of shallow geometrical information and deep semantic information of an image is realized, and the detection effect of the model on a remote target is improved; the part of the FasterRCNN which is artificially provided with the anchor points is optimized to generate the anchor points based on the input characteristics in a self-adaptive manner, so that the detection effect of the model on the construction materials at different positions and proportions in the complex environment is improved;
3) the method realizes high detection precision on the premise of high efficiency and has stronger anti-interference capability.
Drawings
FIG. 1 is a schematic diagram of the overall network architecture of the present invention;
FIG. 2 is a schematic diagram of the Guided Anchoring structure of 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. All other embodiments, which can be obtained by a person skilled in the art without making any creative effort based on the embodiments in the present invention, belong to the protection scope of the present invention.
As shown in FIGS. 1-2, a method for detecting key materials in capital construction site construction based on Guided Anchoring is implemented as follows:
and establishing a building material image library of a capital construction site, and manually marking the training set. The marking mode meets the JSON label file standard of the COCO format and comprises the category of the target to be detected and the position of a real target frame.
And establishing a deep learning target detection model, wherein the backbone network is ResNet50+ FPN (characteristic pyramid network), and adaptively generating an anchor point by using a Guided Anchoring mode. The network structure of the object detection network model is shown in fig. 1.
Specifically, the picture is sent to the network after being preprocessed. After the features are extracted in stages using ResNet50, the features of each stage are blended using an FPN structure.
The method comprises the steps that the traditional idea of artificially setting an anchor point is improved, two new training branches are introduced to replace the traditional anchor point generating process, a position prediction branch and a shape prediction branch are used for helping a frame to complete the generation of a sparse anchor point according to local features of a feature map, the frame compares the output results of the two branches with a set threshold value to obtain the central position of a target possibly existing on the feature map, and then predicts the most possible anchor point shape according to the local features of the features near the central position. Specifically, the joint distribution of anchor point parameters is decoupled into two independent conditional distributions, and the semantic information of the input feature map is combined to calibrate the feature information and output more accurate features for detection. The joint distribution of the anchor points, the conditional distribution of the anchor point positions and the anchor point shapes meet the following formula:
p(x，y，w，h)＝p(x，y|I)p(w，h|x，y，I)
where (x, y) represents the location of the anchor point, (w, h) represents the shape of the anchor point, and I represents the input picture feature.
The position prediction branch generates a probability map with the same size as the input feature map, the probability of the (I, j) position on the probability map is p (I, j | I), and the numerical value represents the probability that the point is the center of the anchor point; the shape prediction branch produces a two-channel output corresponding to the relative height and relative width of the anchor point, respectively. According to the method, only one anchor point with dynamic change is predicted at each position instead of anchor point matrixes with dense distribution, the complex anchor point shape distribution caused by different forms and scales of the construction materials in the picture is well fitted in the capital construction scene, and the generated anchor point has higher recall rate than that of the conventional scheme. The anchor points generated by the anchor point self-adaption method are used for subsequently extracting region-of-interest pooling, and classification and frame regression are carried out through the subsequent network structure which is the same as that of the fast-RCNN.
The total number of experimental pictures was 3000. The number of pictures used for training was 2000, and the remaining 1000 were used as a test set. And performing data enhancement before the training picture enters the model training, and adopting a random turning and color channel standardization method. And (4) uniformly zooming the data-enhanced pictures to 1333 multiplied by 800, and adopting ResNet50 model parameters pre-trained on ImageNet. The parameter updating mode is SGD, the initial learning rate is 0.01, the momentum term is 0.9, the weight attenuation coefficient is 1 multiplied by 10 < -4 >, the batch training size is 4, and the training iteration times are 6000. The training uses 300 iterations to start slowly and uses a learning rate stage reduction mode to reduce the learning rate by 10 times at times 4000 and 5000 iterations.
Compared with the traditional model, the method has the advantage that the detection effect is obviously improved. Table 1 shows the comparison result between the method and the detection result of the Faster-RCNN detection network which takes ResNet50 as a backbone network. The detection precision refers to the proportion of effective detection frames obtained by running on the test set to the total number of target frames. Wherein, the effective detection frame means that the coincidence degree of the detection frame and the marking frame exceeds 0.5.
TABLE 1 comparison of results
Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (6)
1. A method for detecting key materials for capital construction site construction based on Guided Anchoring is characterized in that: generating anchor points for the construction material sample pictures of the infrastructure scene by using a deep learning target detection model, wherein the generated anchor points are used for extracting the pooling of the interested region, and performing classification and frame regression through a subsequent network structure;
the establishment of the deep learning target detection model comprises the following steps:
s1, collecting construction material sample pictures of the capital construction scene, and making a corresponding sample label file for each picture;
s2, taking the combination of Faster R-CNN and FPN structure as a basic detection frame, generating an anchor point frame by adopting a guided anchoring method, and establishing a deep learning target detection model;
s3, randomly dividing all the obtained capital construction scene construction key material sample pictures and corresponding sample labels thereof into a training set and a testing set;
s4, training the deep learning target detection model by using a training set to obtain a primarily trained detection model of key materials for construction in the capital construction site;
and S5, debugging the performance of the primarily trained detection model of the key materials for construction in the construction site by using the test set, adjusting the training parameters and the detection confidence degree threshold according to the test result, and optimizing and solidifying the detection model of the materials for construction in the construction site.
2. The method for detecting key materials for capital construction site construction based on Guided Anchoring as claimed in claim 1, wherein the method comprises the following steps: the sample tag file conforms to the JSON tag file standard of the COCO data set.
3. The method for detecting the key materials for the capital construction site construction based on Guided Anchoring according to claim 1, wherein the method comprises the following steps: the construction key material sample picture of the capital construction scene is a picture acquired by taking construction materials of a capital construction operation site as a target object under various capital construction scenes and carrying out left-right deviation of 15 degrees and shooting distance of 5-25 meters on the target object through a monitoring camera.
4. The method for detecting the key materials for the capital construction site construction based on Guided Anchoring according to claim 1, wherein the method comprises the following steps: the process of generating the detection anchor point by the Guided Anchoring method comprises the following steps: the anchor point position prediction branch, the anchor point shape prediction branch and the sparse anchor point generation are completed according to the local characteristics of the characteristic diagram;
the deep learning target detection model compares the output results of the anchor point position prediction branch and the anchor point shape prediction branch with a set threshold value, firstly obtains the central position of a target possibly existing on a feature map, and then predicts the most possible anchor point shape according to local features near the central position.
5. The method for detecting key materials for capital construction site construction based on Guided Anchoring as claimed in claim 4, wherein the method comprises the following steps: the anchor point position prediction branch generates a probability map with the same size as the input feature map, the probability of the (I, j) position on the probability map is p (I, j | I), and the numerical value represents the probability that the point is the center of the anchor point;
the anchor shape prediction branch produces a two-channel output result corresponding to the relative height and relative width of the anchor point, respectively.
6. The method for detecting the key materials for the capital construction site construction based on Guided Anchoring according to claim 1, wherein the method comprises the following steps: the Guided Anchoring method decouples the joint distribution of anchor point parameters into two independent conditional distributions, and combines the semantic information of the input feature map to calibrate information and output more accurate features for detection;
the anchor point parameters comprise anchor point positions and anchor point shapes;
the joint distribution of the anchor points, the conditional distribution of the anchor point positions and the anchor point shapes meet the following formula:
p(x，y，w，h)＝p(x，y|I)p(w，h|x，y，I)
where (x, y) represents the location of the anchor, (w, h) represents the shape of the anchor, and I represents the input picture feature.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210490399.0A CN114764799A (en) | 2022-05-07 | 2022-05-07 | Material detection method based on Guided adsorbing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210490399.0A CN114764799A (en) | 2022-05-07 | 2022-05-07 | Material detection method based on Guided adsorbing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114764799A true CN114764799A (en) | 2022-07-19 |
Family
ID=82365257
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210490399.0A Pending CN114764799A (en) | 2022-05-07 | 2022-05-07 | Material detection method based on Guided adsorbing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114764799A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2020100705A4 (en) * | 2020-05-05 | 2020-06-18 | Chang, Jiaying Miss | A helmet detection method with lightweight backbone based on yolov3 network |
| CN113903002A (en) * | 2021-10-12 | 2022-01-07 | 广东电网有限责任公司广州供电局 | Tower crane below abnormal intrusion detection method based on tower crane below personnel detection model |
| CN113902958A (en) * | 2021-10-12 | 2022-01-07 | 广东电网有限责任公司广州供电局 | Anchor point self-adaption based infrastructure field personnel detection method |
-
2022
- 2022-05-07 CN CN202210490399.0A patent/CN114764799A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2020100705A4 (en) * | 2020-05-05 | 2020-06-18 | Chang, Jiaying Miss | A helmet detection method with lightweight backbone based on yolov3 network |
| CN113903002A (en) * | 2021-10-12 | 2022-01-07 | 广东电网有限责任公司广州供电局 | Tower crane below abnormal intrusion detection method based on tower crane below personnel detection model |
| CN113902958A (en) * | 2021-10-12 | 2022-01-07 | 广东电网有限责任公司广州供电局 | Anchor point self-adaption based infrastructure field personnel detection method |
Non-Patent Citations (1)
| Title |
|---|
| J. WANG: "Region Proposal by Guided Anchoring", 《2019 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR)》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110363122B (en) | Cross-domain target detection method based on multi-layer feature alignment | |
| US20220245945A1 (en) | Video anomaly detection method based on human-machine cooperation | |
| CN110751209B (en) | Intelligent typhoon intensity determination method integrating depth image classification and retrieval | |
| US11367195B2 (en) | Image segmentation method, image segmentation apparatus, image segmentation device | |
| CN114998673B (en) | Dam defect time sequence image description method based on local self-attention mechanism | |
| CN110796580B (en) | Intelligent traffic system management method and related products | |
| CN109949209B (en) | Rope detection and removal method based on deep learning | |
| CN109919073B (en) | Pedestrian re-identification method with illumination robustness | |
| CN111145222A (en) | Fire detection method combining smoke movement trend and textural features | |
| CN111008608B (en) | Night vehicle detection method based on deep learning | |
| CN114170627A (en) | Pedestrian detection method based on improved Faster RCNN | |
| CN112580569B (en) | Vehicle re-identification method and device based on multidimensional features | |
| CN107871315B (en) | Video image motion detection method and device | |
| CN113284093A (en) | Satellite image cloud detection method based on improved D-LinkNet | |
| CN110942026B (en) | Deep learning-based capsule robot drain pipe disease detection method and system | |
| CN114764799A (en) | Material detection method based on Guided adsorbing | |
| CN115359094B (en) | Moving target detection method based on deep learning | |
| CN115661932A (en) | Fishing behavior detection method | |
| CN113450321B (en) | Single-stage target detection method based on edge detection | |
| CN113902958A (en) | Anchor point self-adaption based infrastructure field personnel detection method | |
| CN114120061A (en) | Small target defect detection method and system for power inspection scene | |
| CN113378598A (en) | Dynamic bar code detection method based on deep learning | |
| CN111401225A (en) | Crowd abnormal behavior detection method based on improved logistic regression classification | |
| CN114743222A (en) | Adaptive sampling-based traffic assistant in-place detection scheme | |
| CN115565068B (en) | Full-automatic detection method for breakage of high-rise building glass curtain wall based on light-weight deep convolutional neural network |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220719 |
|
| RJ01 | Rejection of invention patent application after publication |