CN113256560A - Heading machine nose area intrusion detection method based on YOLOv5 - Google Patents
Heading machine nose area intrusion detection method based on YOLOv5 Download PDFInfo
- Publication number
- CN113256560A CN113256560A CN202110399552.4A CN202110399552A CN113256560A CN 113256560 A CN113256560 A CN 113256560A CN 202110399552 A CN202110399552 A CN 202110399552A CN 113256560 A CN113256560 A CN 113256560A
- Authority
- CN
- China
- Prior art keywords
- worker
- intrusion detection
- heading machine
- yolov5
- nose area
- 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
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- 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]
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
-
- 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
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
Abstract
The invention discloses a heading machine nose area intrusion detection method based on YOLOv5, which relates to the field of artificial intelligence and comprises the following steps: (1) acquiring images of a worker and a 'gun head'; (2) YoloV5 test person and "gun head"; (3) performing border crossing detection according to the prediction frame; (4) and sending out early warning. The invention adopts a YOLOv5 algorithm to construct an intrusion detection model of a heading machine nose area, provides a multi-dimensional intrusion detection method, judges whether a worker invades the heading machine nose area or not by calculating the IOU ratio of the worker and a 'nose' prediction frame and the distance of the intersection point of diagonal lines of the two prediction frames, and carries out grading early warning on the worker. According to the method, the invasion behavior of the worker is judged from two dimensions, so that the detection accuracy is improved; the method is realized by artificial intelligence deployment, so that the intrusion detection of the heading machine nose area can be realized quickly and conveniently, and a large amount of manpower and material resources are saved.
Description
Technical Field
The invention relates to a heading machine nose area intrusion detection method based on YOLOv5, and belongs to the field of artificial intelligence.
Background
At present, the safety of a coal mine always threatens the health and life safety of mining personnel. On one hand, the underground natural conditions of the coal mine are severe, and the coal mine is threatened by natural disasters such as gas, a roof, water permeability and the like in the mining process; on the other hand, many underground coal mine personnel inevitably cause various errors and violations in the operation process. The above two points are the main aspects threatening the health and life safety of mining personnel, where errors and non-specifications of human operation are the main problems. This just needs managers to carry out the video screenshot to wrong operation and nonstandard action and proves, because the video is too many, needs a large amount of manpowers, and efficiency is also very low, can not guarantee workman's life safety in time moreover.
Therefore, the novel intrusion detection method for the heading machine nose area is provided, mainly aiming at intrusion detection, the system can detect the intrusion of the heading machine nose area, store corresponding records, give a prompt when detecting the intrusion, quickly detect under the condition of not influencing the normal operation of workers, and better improve the safety guarantee of the workers.
Disclosure of Invention
Aiming at the problems, the invention provides a rapid and simple intrusion detection method for the heading machine gun head area.
In order to achieve the purpose, the invention adopts the technical scheme that: a heading machine nose area intrusion detection method based on YOLOv5 comprises the following steps:
(1) acquiring images of a worker and a 'gun head';
(2) yoolov 5 detects workers and "gun heads";
(3) carrying out intrusion detection according to the prediction frame;
(4) and sending out early warning.
In step (1), in order to better apply the present invention to a practical downhole environment, we acquire an image data set of workers and "shots" from the real downhole environment.
In the step (2), we divide it into the following aspects:
training: the underground environment is severe, and interference is inevitably generated in the identification process. Firstly, the data set collected by us comprises two parts of fuzzy and clear, wherein a fuzzy image accounts for 20 percent, and a clear image accounts for 80 percent; then, random sampling is carried out from clear and fuzzy images in a random sampling mode, and the clear and fuzzy images are divided into independent and non-repetitive training sets and test sets according to a certain proportion, wherein the training sets account for 70% and the test sets account for 30%. Finally, the training set is input into the YoloV5 algorithm for training. The step ensures that the underground environment is simulated in the training process, so that certain noise interference is generated, and the model is prevented from being over-fitted;
and (3) prediction: after model training is completed, the test set is input into the trained model, the test set is predicted, and then a prediction result is checked.
In the step (3), detection is performed according to the output prediction frame. Firstly, determining whether a worker prediction box and a ' gunhead ' prediction box are approaching according to the IOU ratio of the worker prediction box and the ' gunhead ' prediction box, and if the IOU is always 0, indicating that the prediction boxes between the worker prediction box and the "gunhead ' prediction box are in disjoint positions, so that the invasion condition does not exist; and then, when the IOU of the two frames is larger than 0, the two prediction frames are intersected, the IOU ratio between the two frames is not suitable for judging whether the worker invades the heading machine nose area or not, and for the situation, the distance of the intersection point of the diagonal lines of the two prediction frames is calculated to judge whether the worker invades the heading machine nose area or not.
In the step (4), grading early warning is carried out according to the intrusion detection strategy in the step (3), and early warning is sent out through a buzzer alarm in the process. When the IOU ratio of a worker to a 'gunpoint' prediction frame is always 0, the worker is in a safe area at the moment, and an alarm cannot send out an early warning signal; when the IOU ratio of a worker to a 'shot' prediction frame is larger than 0, whether the worker invades a heading area of the heading machine is judged by calculating the distance between the intersection point of the diagonal lines of the worker and the 'shot' prediction frame, and through investigation of a large number of real underground conditions, when the distance between the worker and the intersection point of the diagonal lines of the 'shot' prediction frame is smaller than 2 meters and larger than 0.5 meter, the worker is already at a relatively unsafe position, a yellow light is turned on by a buzzer alarm to prompt the worker to pay attention to operation safety, and when the distance between the worker and the intersection point of the diagonal lines of the 'shot' prediction frame is smaller than 0.5 meter, the worker is already at a dangerous position, a red light is turned on the alarm to send an alarm to prompt the worker to fast retreat to the safe distance.
Drawings
FIG. 1 is a flow chart of a training process for implementing a heading machine nose area intrusion detection method according to the present invention;
fig. 2 shows intrusion detection results of a heading machine nose area obtained by implementing the present invention.
Detailed Description
The present invention is further described below in conjunction with the appended drawings to enable one skilled in the art to practice the invention with reference to the description.
The invention works in Ubuntu16.04.4LTS environment, is constructed by adopting PyTorch as a frame, and has the following main parameters: the initial learning rate is 0.01, the momentum parameter is 0.937, the weight coefficient is 0.0005, the training threshold is 0.65, the imagesize is 640 × 640, the epoch is 300, and the like.
As shown in fig. 1, we first collect real image data sets of workers and "gun heads" from the underground coal mine, wherein the real image data sets include two parts, namely, a blurred image and a clear image, wherein the blurred image accounts for 20%, the clear image accounts for 80%, and label the acquired data sets with labellmg.
Further, the labeled image data is divided into independent and non-repetitive training sets and test sets according to the proportion of 70% of the training sets and 30% of the test sets.
Further, the training set is input into yolov5 algorithm, and a best model is found and stored after adjusting the size of various parameters for many times, thereby completing the training of the model.
And further, calling a model, testing, inputting a test set to obtain a test result, calculating the IOU occupation ratio between a worker prediction frame and a 'gunhead' prediction frame and the distance between the diagonal intersection points of the prediction frames, establishing a trigger between the grading detection strategy and the buzzer, and triggering the buzzer to send out corresponding early warning when the IOU occupation ratio and the distance between the diagonal intersection points of the prediction frames reach a set threshold value.
Through the combination of two schemes of calculating the IOU ratio between the worker prediction frame and the 'gun head' prediction frame and the distance of the diagonal intersection point of the prediction frame, the intrusion detection of the real underground mine heading machine gun head area is realized from two dimensions, the life safety of mining personnel is ensured by rapid detection under the condition of not influencing the normal operation of workers, and a guarantee is provided for the national coal mine safety.
Claims (5)
1. A heading machine nose area intrusion detection method based on YOLOv5 is characterized by comprising the following steps:
(1) acquiring images of a worker and a 'gun head';
(2) yoolov 5 detects workers and "gun heads";
(3) carrying out intrusion detection according to the prediction frame;
(4) and sending out early warning.
2. The heading machine nose area intrusion detection method based on YOLOv5 as claimed in claim 1, characterized in that: in the step (1), in order to improve the application capability of the invention in an actual scene, the collected data set pictures are all from scenes in a real mine, and then the data set pictures are labeled by using a target detection labeling tool, so that the image information of the worker and the shot head is obtained.
3. The heading machine nose area intrusion detection method based on YOLOv5 as claimed in claim 1, characterized in that: in the step (2) described, the data set collected by us comprises two parts of fuzzy and clear, wherein the fuzzy image accounts for 20% and the clear image accounts for 80%; then, random sampling is carried out from clear and fuzzy images in a random sampling mode, and the clear and fuzzy images are divided into independent and non-repetitive training sets and test sets according to a certain proportion, wherein the training sets account for 70 percent, and the test sets account for 30 percent; because the real underground environment is severe and interference is inevitably generated in the identification process, certain noise is contained in the training process, so that overfitting of the model is prevented, and further the generalization capability of the model is poor.
4. The heading machine nose area intrusion detection method based on YOLOv5 as claimed in claim 1, characterized in that: in the described step (3), intrusion detection is carried out from the IOU ratio of the worker prediction box and the 'gunhead' prediction box and the distance of the intersection point of the diagonals of the two prediction boxes in two dimensions; through the survey of a large number of real underground situations, we count and summarize that when the distance between the intersection of the worker and the diagonal of the prediction box of the 'shot' is less than 2 meters and more than 0.5 meter, the worker is in a relatively unsafe position, and when the distance is less than 0.5 meter, the worker is in a dangerous position.
5. The heading machine nose area intrusion detection method based on YOLOv5 as claimed in claim 1, characterized in that: in the step (4), the safety state of a worker is divided into three states of safety, safety and danger, which respectively correspond to a green light, a yellow light and a red light of a buzzer; and (4) establishing a trigger between the tunneling machine gun head area intrusion detection method in the step (3) and the buzzer, and triggering the buzzer to achieve classification alarm when the prediction result reaches a threshold set by people.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110399552.4A CN113256560A (en) | 2021-04-14 | 2021-04-14 | Heading machine nose area intrusion detection method based on YOLOv5 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110399552.4A CN113256560A (en) | 2021-04-14 | 2021-04-14 | Heading machine nose area intrusion detection method based on YOLOv5 |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113256560A true CN113256560A (en) | 2021-08-13 |
Family
ID=77220890
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110399552.4A Pending CN113256560A (en) | 2021-04-14 | 2021-04-14 | Heading machine nose area intrusion detection method based on YOLOv5 |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113256560A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113808117A (en) * | 2021-09-24 | 2021-12-17 | 北京市商汤科技开发有限公司 | Lamp detection method, device, equipment and storage medium |
-
2021
- 2021-04-14 CN CN202110399552.4A patent/CN113256560A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113808117A (en) * | 2021-09-24 | 2021-12-17 | 北京市商汤科技开发有限公司 | Lamp detection method, device, equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111144263A (en) | Construction worker high-fall accident early warning method and device | |
CN105405150A (en) | Abnormal behavior detection method and abnormal behavior detection device based fused characteristics | |
KR101095528B1 (en) | An outomatic sensing system for traffic accident and method thereof | |
CN111160125A (en) | Railway foreign matter intrusion detection method based on railway monitoring | |
CN106355162A (en) | Method for detecting intrusion on basis of video monitoring | |
CN109446936A (en) | A kind of personal identification method and device for monitoring scene | |
CN106210634A (en) | A kind of wisdom gold eyeball identification personnel fall down to the ground alarm method and device | |
CN106128022A (en) | A kind of wisdom gold eyeball identification violent action alarm method and device | |
CN111126193A (en) | Artificial intelligence recognition system based on deep learning coal mine underground unsafe behavior | |
CN108197575A (en) | A kind of abnormal behaviour recognition methods detected based on target detection and bone point and device | |
CN113484858A (en) | Intrusion detection method and system | |
CN114202711A (en) | Intelligent monitoring method, device and system for abnormal behaviors in train compartment | |
CN110780356A (en) | Subway platform clearance foreign matter detecting system | |
CN106127814A (en) | A kind of wisdom gold eyeball identification gathering of people is fought alarm method and device | |
CN111178424A (en) | Petrochemical production site safety compliance real-time detection system and method | |
CN111179551A (en) | Real-time monitoring method for dangerous chemical transport driver | |
CN113256560A (en) | Heading machine nose area intrusion detection method based on YOLOv5 | |
CN113990101B (en) | Method, system and processing device for detecting vehicles in no-parking area | |
CN112906441A (en) | Image recognition system and method for communication industry survey and maintenance | |
CN109873990A (en) | A kind of illegal mining method for early warning in mine based on computer vision | |
CN110486081A (en) | A kind of development machine operation safety monitoring method, system, equipment and storage medium | |
CN114067244A (en) | Safety operation violation behavior video analysis method and system | |
CN113344882A (en) | Method for counting drill rods of underground drilling machine based on computer vision | |
CN113554682B (en) | Target tracking-based safety helmet detection method | |
CN113653534B (en) | Mine water damage early warning system and method |
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 |