A method of using sensors to reduce risks in wood processing machinery

Abstract

The present invention discloses a method for reducing mechanical risks in wood processing by using sensors. The method includes installing a thermal imaging sensor above the wood processing machine, collecting operator and field data in real time, selecting a batch of field data by sampling, establishing a data model, and reminding the operator in real time on site based on the field data and the model calculation results, so as to reduce mechanical risks in wood processing. The method has the characteristics of low cost, easy installation, no impact on existing production lines, etc., and has significant effects. The method uses sensors and data models to reduce the occurrence of mechanical risks in the wood processing process.

Description

Technical Field

[0001] The invention relates to a method for reducing the risk of wood processing machinery by using sensors. Background Art

[0002] The most common hazard in woodworking is the operator's fingers being cut by the machine, resulting in disability. Usually, the operator's high concentration is relied upon to avoid accidents, but the operator is prone to fatigue due to long-term work.

[0003] Although there are some automated stopping machinery and equipment, the cost is relatively high and compared with the current industry status in wood processing, it is difficult to popularize them. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the defects of the prior art that the automated stopping mechanical equipment for reducing accidents in wood processing is high in cost and difficult to popularize, and to provide a method for reducing the risk of wood processing machinery by using sensors.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A method for reducing the risk of wood processing machinery by using sensors, by installing a thermal imaging sensor above the wood processing machine, using the sensor to collect operator and field data in real time, and establish a data model. In actual application, according to the field data and the model calculation results, the operator is reminded in real time on site to reduce the mechanical risk in wood processing, specifically comprising the following steps:

[0007] S1. Install and initialize the sensor;

[0008] Install the thermal imaging sensor above the wood processing machine, adjust the angle of the sensor, refresh the latest data model for the sensor, and initialize the data model;

[0009] Initialize sensors: curing face angle X1, curing machine speed V1 and curing environment decibel H1;

[0010] S2, collect target data;

[0011] Get thermal sensor video, environmental data-decibel and character data-the direction and angle of the face respectively;

[0012] Specifically, the machine operating speed V1 is used to measure the operator speed; the machine operating speed is fixed, and the machine operating speed is set to the system, and the operator's propulsion speed is consistent with the machine operating speed.

[0013] Face coverage angle X1: to measure whether the operator is paying attention to the work; determine X1 through affine transformation?

[0014] Environmental decibel H1: to measure whether the machine is running; the environmental decibel Hx is obtained every 60 seconds, and the value of Hx is between [Hm, Hn];

[0015] Real-time acquisition of thermal sensor images, analysis of thermal sensor images, the use of sensors with cameras to acquire thermal sensor images, real-time acquisition of video images, analysis of video image content based on the model to find the operator's hands and face, real-time calculation of the distance Sx between the hand and the machine, Sx value between [Sm, Sn], real-time calculation of the facial angle interval Xx, facial angle Xx value between [Xm, Xn];

[0016] Danger distance D1: To measure the distance at which the risk will occur; to set the electronic fence to mark whether it is a dangerous distance;

[0017] S3. Determine the target data risk based on the data model;

[0018] The value of Hx in [Hm, Hn] indicates that the machine is working;

[0019] If the face angle Xx is between [Xm, Xn], it indicates that the face is paying attention, and if it is outside of [Xm, Xn], it indicates that the face is not paying attention.

[0020] Calculate the danger time Tx (milliseconds);

[0021] Tx=(Sx-D1) / V1

[0022] Tx is between [Tn, D1 / V1]

[0023] Pass parameters such as X and T to the data model, and the data model gives the result;

[0024] S4. Retrain and optimize the data model

[0025] Every day, sensors upload alarm data to the data analysis system, which retrains and optimizes the data model based on the alarm data. The data model is then updated to the sensors.

[0026] X is in range

[0027] If the Tx data is less than 1 second, set an orange alarm;

[0028] TX data <500ms, set a red alarm;

[0029] X is out of range

[0030] Calculate the transfer time XT of X

[0031] If the data of XT+Tx is less than 1 second, an orange alarm is set;

[0032] XT+TX data <500ms, set a red alarm;

[0033] Retrain and optimize based on actual results

[0034] The environmental data is calculated based on the data model to determine whether it is working, the position and speed of the hand in the video are calculated based on the data model, and the risk level is identified based on the data model; red: a high-decibel alarm sounds on site; orange: a medium-decibel alarm sounds on site.

[0035] The beneficial effects achieved by the present invention are:

[0036] This method has the characteristics of low cost, easy installation, no impact on existing production lines, etc. It has significant effects and uses sensors and data models to reduce the occurrence of mechanical risks in the wood processing process. DETAILED DESCRIPTION

[0037] The preferred embodiments of the present invention are described below. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0038] Example

[0039] Install the thermal imaging sensor above the wood processing machine, use the sensor to collect operator and field data in real time, select a batch of field data by sampling, and establish a data model. In actual application, according to the field data and the model calculation results, the operator is reminded in real time on site to reduce the mechanical risks in wood processing. The specific steps include:

[0040] S1. Install and initialize the sensor;

[0041] Install the thermal imaging sensor above the wood processing machine, adjust the angle of the sensor, refresh the latest data model for the sensor, and initialize the data model; the angle of the sensor needs to meet 3 points: the camera can cover: the operator's face, the operating table + the cutting edge of the machine.

[0042] Initialize sensors: fix face orientation angle X1, machine speed V1 and ambient decibel level H1;

[0043] S2, collect target data;

[0044] Thermal sensor video; Environmental data: decibel; Character data: the angle of the direction of the face; On-site measurement of the following basic environmental data: Machine operation speed V1 can be achieved through camera + motherboard + CPU: to measure the operator speed;

[0045] Face coverage angle X1: to measure whether the operator is paying attention to the work;

[0046] Environmental decibel H1: to measure whether the machine is running; Environmental decibel: Hx is obtained every 60 seconds, and the value of Hx is between [Hm, Hn];

[0047] Acquire thermal sensor images in real time, analyze the thermal sensor images, find the operator's hands and face, calculate the distance Sx between the hand and the machine in real time, Sx takes the value between [Sm, Sn], calculate the facial angle interval Xx in real time, and the facial angle Xx takes the value between [Xm, Xn];

[0048] Danger distance D1: to measure the distance at which the risk will occur;

[0049] S3. Determine the target data risk based on the data model;

[0050] The value of Hx in [Hm, Hn] indicates that the machine is working;

[0051] If the face angle Xx is between [Xm, Xn], it indicates that the face is paying attention, and if it is outside of [Xm, Xn], it indicates that the face is not paying attention.

[0052] Calculate the danger time Tx (milliseconds);

[0053] Tx=(Sx-D1) / V1

[0054] Tx is between [Tn, D1 / V1]

[0055] Pass parameters such as X and T to the data model, and the data model gives the result;

[0056] S4. Retrain and optimize the data model

[0057] Every day, the sensor uploads the alarm data to the data analysis system, which retrains and optimizes the data model based on the alarm data, and the data model is updated to the sensor.

[0058] X is in range

[0059] If the Tx data is less than 1 second, set an orange alarm;

[0060] TX data <500ms, set a red alarm;

[0061] X is out of range

[0062] Calculate the transfer time XT of X

[0063] If the data of XT+Tx is less than 1 second, an orange alarm is set;

[0064] XT+TX data <500ms, set a red alarm;

[0065] Retrain and optimize based on actual results

[0066] The environmental data is calculated based on the data model to determine whether it is working, the position and speed of the hand in the video are calculated based on the data model, and the risk level is identified based on the data model; red: a high-decibel alarm sounds on site; orange: a medium-decibel alarm sounds on site.

[0067] When in use, the device is installed on the same vertical plane of the machine, and the height depends on whether the coverage area covers the angle area shown in the figure; the machine's operating speed determines the arm's moving speed, so the arm's moving speed is set to V0; considering that the average human reaction time is 300 milliseconds, we set the device alarm limit time to 500 milliseconds. Therefore, the distance between the arm and the machine should be greater than or equal to V0*0.5 seconds, and this distance is set as the danger distance S0

[0068] If the human face is not facing the front of the machine, the distance between the arm and the machine equipment should be greater than or equal to V0*1 second, and this distance is set as the abnormal danger distance S0'.

[0069] If the arm distance is less than the danger distance, the system automatically saves the video in the 5 seconds before and after for subsequent data rule optimization

[0070] The system sets up two electronic fences on the plane, namely the primary fence and the secondary fence; when the arm enters the electronic fence, the device will emit an alarm prompt sound, and when the arm leaves the electronic fence, the alarm will be lifted.

[0071] When recognizing faces, the open source face recognition engine SeetaFaceEngine is used to identify faces in the video. It mainly uses two parts: the face detection module SeetaFace Detection and the facial feature point positioning module SeetaFaceAlignment. SeetaFace Detection uses a cascade structure that combines traditional artificial features with a multi-layer perceptron (MLP), achieving a recall rate of 84.4% on FDDB (when there are 100 false detections), and can process VGA resolution images in real time on a single i7 CPU; Facial feature point positioning module (SeetaFace Alignment): By cascading multiple deep models (stacked autoencoder networks) to regress the positions of 5 key feature points (centers of both eyes, tip of the nose, and two corners of the mouth), it achieves state-of-the-art accuracy on the AFLW database, and the positioning speed exceeds 200fps on a single i7CPU.

[0072] Face deviation recognition

[0073] When the face cannot be recognized, the face coverage angle X1 is set to -1, indicating that the operator is not paying attention to the current work. When the face can be recognized, the face coverage angle is calculated. Generally speaking, the distance between the two eyes of a human is about 5.8 cm to 6.1 cm. The system calculates the distance between the centers of the two eyes in real time by establishing a coordinate system, and assigns the calculated distance between the two eyes to X1.

[0074] Arm recognition and monitoring

[0075] Through facial recognition to the body, and then to the arm; the arm is continuously monitored.

[0076] Alarm control

[0077] Fence takes effect: When X1<3.5 cm, the first-level fence is enabled; when X2>=3.5 cm, the second-level fence is enabled.

[0078] When the arm monitoring program detects that the arm is approaching the fence, the system sounds an alarm.

[0079] When the system triggers an alarm, the system retains the 5 seconds of video before and after and retains the current data.

[0080] Model Optimization

[0081] Collect system alarm data from each terminal every month and conduct data analysis: Is the alarm normal? Is the alarm distance sufficient? How to make a more accurate prediction if an accident occurs? Based on the above content, optimize the existing data model. After optimization, the model can be re-implanted into the device and continue to collect data for the next step of optimization.

[0082] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for reducing the risk of wood processing machinery using sensors, characterized in that: By installing thermal imaging sensors above wood processing machines, real-time data from operators and on-site data is collected. A batch of on-site data is selected by sampling, and a data model is established. Based on the on-site data and the model calculation results, the operator is reminded in real time on-site to reduce mechanical risks in wood processing. The specific steps include: S1. Install and initialize the sensor; Install the thermal imaging sensor above the wood processing machine, adjust the angle of the sensor, refresh the latest data model for the sensor, and initialize the data model; S2, collect target data; The following basic environmental data were measured on site: Machine operating speed V1: to measure the operator speed; Face coverage angle X1: to measure whether the operator is paying attention to the work; Environmental decibel H1: To measure whether the machine is running, the environmental decibel is obtained every 60 seconds Hx, and the value of Hx is between [Hm, Hn]; Acquire thermal sensor images in real time, analyze the thermal sensor images, find the operator's hands and face, calculate the distance Sx between the hand and the machine in real time, Sx takes the value between [Sm, Sn], calculate the facial angle interval Xx in real time, and the facial angle Xx takes the value between [Xm, Xn]; Danger distance D1: to measure the distance at which the risk will occur; S3. Determine the target data risk based on the data model; The value of Hx in [Hm, Hn] indicates that the machine is working; If the face angle Xx is between [Xm, Xn], it indicates that the face is paying attention, and if it is outside [Xm, Xn], it indicates that the face is not paying attention. Calculate the danger time Tx (milliseconds); Tx=(Sx-D1) / V1, Tx is between [Tn, D1 / V1] Pass parameters such as Xx and Tx to the data model, and the data model gives the result; S4. Retrain and optimize the data model Every day, sensors upload alarm data to the data analysis system. The data analysis system retrains and optimizes the data model based on the alarm data, and the data model is updated to the sensors. Retraining and optimization are carried out based on actual results. Environmental data is calculated based on the data model to determine whether it is working. The position and speed of the hand in the video are calculated based on the data model. The risk level is identified based on the data model. Red: a high-decibel alarm sounds on site. Orange: a medium-decibel alarm sounds on site.

2. The method for reducing the risk of wood processing machinery using sensors as claimed in claim 1, characterized in that: The sensors initialized in S1 include the fixed face orientation angle X1, the fixed machine operation speed V1 and the fixed telephone environment decibel H1.

3. The method of reducing the risk of wood processing machinery using sensors as claimed in claim 1, characterized in that: The settings in S3 are as follows: When Tx is within range, If the Tx data is less than 1 second, set an orange alarm; If the Tx data is less than 500 milliseconds, a red alarm is set; When Tx is out of range Calculate the transfer time XT of Tx; If the data of XT+Tx is less than 1 second, an orange alarm is set; If XT+TX data is less than 500 milliseconds, a red alarm is set.