CN117173533A - Integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel - Google Patents
Integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel Download PDFInfo
- Publication number
- CN117173533A CN117173533A CN202310923981.6A CN202310923981A CN117173533A CN 117173533 A CN117173533 A CN 117173533A CN 202310923981 A CN202310923981 A CN 202310923981A CN 117173533 A CN117173533 A CN 117173533A
- Authority
- CN
- China
- Prior art keywords
- convolution
- module
- theta
- greenhouse
- lamp post
- 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
- 238000012806 monitoring device Methods 0.000 title claims abstract description 26
- 238000012544 monitoring process Methods 0.000 claims abstract description 57
- CURLTUGMZLYLDI-UHFFFAOYSA-N Carbon dioxide Chemical compound O=C=O CURLTUGMZLYLDI-UHFFFAOYSA-N 0.000 claims abstract description 28
- 238000001514 detection method Methods 0.000 claims abstract description 25
- 239000002689 soil Substances 0.000 claims abstract description 15
- 229910002092 carbon dioxide Inorganic materials 0.000 claims abstract description 14
- 239000001569 carbon dioxide Substances 0.000 claims abstract description 14
- 238000004378 air conditioning Methods 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 4
- 210000002414 leg Anatomy 0.000 claims description 34
- 230000009467 reduction Effects 0.000 claims description 27
- 230000001965 increasing effect Effects 0.000 claims description 21
- 238000010586 diagram Methods 0.000 claims description 16
- 210000002683 foot Anatomy 0.000 claims description 16
- 238000000034 method Methods 0.000 claims description 16
- 238000005070 sampling Methods 0.000 claims description 15
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 230000008569 process Effects 0.000 claims description 9
- 238000000926 separation method Methods 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 9
- 230000009471 action Effects 0.000 claims description 8
- 238000005452 bending Methods 0.000 claims description 8
- FQENQNTWSFEDLI-UHFFFAOYSA-J sodium diphosphate Chemical compound [Na+].[Na+].[Na+].[Na+].[O-]P([O-])(=O)OP([O-])([O-])=O FQENQNTWSFEDLI-UHFFFAOYSA-J 0.000 claims description 7
- 235000019818 tetrasodium diphosphate Nutrition 0.000 claims description 7
- 230000007704 transition Effects 0.000 claims description 7
- 210000003423 ankle Anatomy 0.000 claims description 6
- 229910052799 carbon Inorganic materials 0.000 claims description 6
- 230000000694 effects Effects 0.000 claims description 6
- 210000003127 knee Anatomy 0.000 claims description 6
- 238000011176 pooling Methods 0.000 claims description 6
- 230000008707 rearrangement Effects 0.000 claims description 6
- 210000000707 wrist Anatomy 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000008859 change Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005728 strengthening Methods 0.000 claims 1
- 230000010354 integration Effects 0.000 abstract description 2
- 230000008635 plant growth Effects 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 description 5
- 238000005286 illumination Methods 0.000 description 5
- 238000009434 installation Methods 0.000 description 4
- 210000000746 body region Anatomy 0.000 description 3
- 230000033228 biological regulation Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 210000004247 hand Anatomy 0.000 description 2
- 230000017525 heat dissipation Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000012271 agricultural production Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 210000003414 extremity Anatomy 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000029553 photosynthesis Effects 0.000 description 1
- 238000010672 photosynthesis Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Image Processing (AREA)
Abstract
The application relates to the technical field of intelligent lamp post equipment, in particular to an integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel, which comprises an integrated intelligent lamp post and a monitoring device arranged on the integrated intelligent lamp post, wherein the monitoring device comprises an intelligent monitoring module, an LED lamp management control module, a photoresistor module, a temperature monitoring module, an air conditioning module, a humidity monitoring module, a humidifier module, a carbon dioxide monitoring module, a soil pH value monitoring module, a solar panel module and a main control module; the intelligent monitoring module is used for acquiring normal light image information and infrared image information in the greenhouse through the vision sensor, processing the acquired normal light image information by adopting a MOL-POSE algorithm, and providing the distance of a detection target in an infrared image auxiliary mode. The application is used for assisting farmers in judging whether people in the greenhouse are reasonably operating the greenhouse, and maintaining the internal environment of the greenhouse in a state most suitable for the growth of plants in the greenhouse, thereby realizing the monitoring integration of the farm greenhouse.
Description
Technical Field
The application relates to the technical field of intelligent lamp post equipment, in particular to an integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel.
Background
Agriculture is an important pillar industry in China, and as the first agricultural large country in the world, agricultural production takes a significant role in the economic construction and social development of China. There are many places where improvement and improvement of environmental control of our greenhouses are required. The research of the intelligent monitoring system of the greenhouse relates to various technologies and subjects such as machine vision technology, sensor technology, control technology, communication technology, biotechnology and environmental science.
Therefore, the application needs to provide an integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel so as to solve the technical problems.
Disclosure of Invention
The application aims to solve the defects in the prior art, and provides an integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel, which is used for assisting farmers in judging whether people in a greenhouse are reasonably operating the greenhouse, maintaining the internal environment of the greenhouse in a state most suitable for internal plant growth and realizing the monitoring integration of the farm greenhouse.
In order to achieve the above purpose, the present application adopts the following technical scheme: the integrated intelligent lamp post monitoring device comprises an integrated intelligent lamp post and a monitoring device arranged on the integrated intelligent lamp post, wherein the monitoring device comprises an intelligent monitoring module, an LED lamp management control module, a photoresistor module, a temperature monitoring module, an air conditioning module, a humidity monitoring module, a humidifier module, a carbon dioxide monitoring module, a soil pH value monitoring module, a solar panel module and a main control module; the intelligent monitoring module, the LED lamp management control module, the air conditioning module, the humidifier module, the carbon dioxide monitoring module, the soil pH value monitoring module and the solar panel module are electrically connected with the main control module;
the intelligent monitoring module collects normal light image information and infrared image information in the greenhouse through the vision sensor, and adopts MOL-POSE algorithm to process the collected normal light image information, wherein the infrared image information is used for assisting in providing the distance of a detection target. Firstly, selecting an area of a human body by a frame, identifying coordinates of 16 key points of the human body in the area, and obtaining state information of the current human body posture by adopting a state conversion model according to the position relation of the coordinates so as to judge whether detected personnel are reasonably operating the greenhouse;
the temperature monitoring module, the air conditioning module, the humidity monitoring module, the humidifier module, the carbon dioxide monitoring module and the soil pH value monitoring module are used for detecting and adjusting the internal environment state of the greenhouse.
Preferably, the MOL-POSE algorithm can complete the gesture recognition task and the target detection task simultaneously. The MOL-post algorithm includes a backbone network and an output network Head structure. The backbone network selects a lightweight network DI-MobileNet, and a scale pyramid is constructed to extract features of different scales of the input image; the normal light image information and the infrared image information are combined as input to the MOL-post.
Preferably, the one coding layer shared by the target detection and the gesture detection adopts a lightweight network DI-MobileNet, and 1/16, 1/32 and 1/64-level bottom layer features are obtained through downsampling operation; because the 1/64-level bottom layer feature map is smaller and faces the limitation of width and height, a TSPP module is added into the 1/64-level bottom layer feature map for enhancing the multi-scale feature extraction of the single feature map; taking the multi-scale features as three inputs of a Head structure, wherein the Head structure comprises three groups of continuous up-sampling and convolution operations, and the up-sampling adopts a PIX-up module to effectively amplify a small feature map; meanwhile, stacking the results of the up-sampling and convolution operation and the bottom layer characteristics from the backbone network, and performing convolution operation on the stacked results to obtain three paths of outputs; finally, the attitude task directly returns to obtain 16 human body key point normalized coordinates through a Head structure; the target detection task obtains a multi-scale prediction frame endpoint result.
Preferably, the Head structure decodes the different feature maps based on 1/16, 1/32 and 1/64 level bottom features to obtain a prediction result, namely, the upper left endpoint coordinate (C x ,C y ) And lower right endpoint coordinates (D x ,D y ) And according to the reduction ratio of the corresponding feature map to the original map, performing inverse transformation, taking the lower right endpoint coordinate as an example, and the formula is as follows:
in D' x ,D' y Is the inverse transformed abscissa and ordinate, D x ,D y For the inverse transformation pre-coordinates, K is the reduction ratio, taking the reciprocal of the reduction ratio corresponding to the 1/16, 1/32 and 1/64 level bottom feature map for 16, 32 and 64 respectively;
according to the point coordinates based on the original image size obtained by inverse transformation, the average value of the horizontal coordinate and the vertical coordinate is obtained by taking three points with similar positions as a group, and the average value point is taken as the final prediction frame endpoint.
Preferably, the TSPP module uses depth separable convolution to replace a common convolution block, so that a 1/64-level bottom layer characteristic map receptive field is enlarged; taking a 1/64-level bottom layer feature map as input, passing through five channels: in the first channel, depth separable convolution is not added, and the traditional 1X1 point convolution operation is used; processing the image by adopting a 3X3 convolution kernel and a depth separable convolution with a separation degree of 6 in the second channel; the third channel adopts a 3X3 convolution kernel and a depth separable convolution with a degree of separation of 12 to process the image; the fourth channel adopts a 3X3 convolution kernel and a depth separable convolution with a separation degree of 18 to process the image; the fifth channel directly carries out pooling operation on the original image and completes one-time up-sampling operation; finally, the five-channel results are all stacked together, reduced in depth using point convolution and output.
Preferably, the PIX-up module includes a convolution operation with a convolution kernel of 3X3 and a step size of 1, a normalization operation, an activation operation, and a pixel rearrangement operation; the pixel rearrangement operation converts the original characteristic diagram with the size of (h, w, r) into the characteristic with the size of (h, w, r, C), wherein h and w are the height and width of the original characteristic diagram, r is the up-sampling expansion multiple, and C is the number of channels after conversion.
Preferably, the lightweight network DI-MobileNet is improved on the basis of MobileNet, firstly 12 and 13 of MobileNet, an average pooling layer and a full connection layer are removed, then the stride of a 6 th layer is changed to be 1/2 of the original one, and finally a hole convolution layer is added in the 7 th layer;
in the convolution network, the deeper the network layer number is, the larger the receptive field of the network is, the receptive field of the network can be increased by the cavity convolution without stacking the convolution layers, and the sizes of the convolution kernels of the cavity convolution and the common convolution are k s The number of holes in the hole convolution is d r The size of the common convolution kernel of the cavity convolution equivalent is K, and the equivalent conversion formula is as follows:
K=k s +(k s -1)×(d r -1)
wherein K is the size of a common convolution kernel equivalent to cavity convolution, and d r Number of holes, k, which is a convolution of holes s For the convolution kernel size of the cavity convolution, the calculation formula of the receptive field of the cavity convolution is as follows:
t in n Representing the sense of convolution of the layer n voidsSize of the affected field, T n-1 The size of the n-1 layer cavity convolution receptive field is represented, the size of the K cavity convolution equivalent common convolution kernel,representing the current step size.
Preferably, in the state transition model, the states of the arms are divided into 7 types of "lifting the arm", "lowering the arm", "bending the arm", "extending the arm", "hooking the hand", "lifting the hand", "standing the arm", and the leg states are divided into 7 types of "lifting the leg", "putting the leg", "bending the leg", "extending the leg", "lifting the foot", "standing the foot", standing the leg, and the coordinates of 4 key points of each arm and 4 key points of each leg are converted into angles and matched with the 14 types of states; each arm corresponds to 4 key points, namely four points of shoulder (A), elbow (B), wrist (C) and hand (D), from top to bottom, each leg corresponds to 4 key points, namely hip (E), knee (F), ankle (G), foot (H), from top to bottom, and corresponds to 8 groups of coordinates (x 1 ,y 1 ),(x 2 ,y 2 ),(x 3 ,y 3 ),(x 4 ,y 4 ),(x 5 ,y 5 ),(x 6 ,y 6 ),(x 7 ,y 7 ),(x 8 ,y 8 ) And uses the formula:
obtaining the distance between the points;
definition d respectively 1 ,d 2 ,d 3 ,d 4 ,d 5 ,d 6 ,d 7 The distance is A to B, B to C, C to D, D to E, E to F, F to G, G to H;
and then the formula is utilized:
six angles of shoulder (A), elbow (B), wrist (C), hip (E), knee (F), ankle (G) are respectively defined as theta 1 ,θ 2 ,θ 3 ,θ 4 ,θ 5 ,θ 6 ;
And comprehensively measuring the posture of a person at the moment according to the obtained angle change.
Preferably, in the state transition model, the matching method of the obtained angle and the 14 types of states is as follows:
(1) First to theta 1 To determine if theta 1 The movement range is increased to 0-145 degrees and is defined as 'lifting arm', if theta 1 The reduction and the movement range is 145-0 degrees, and the movement range is defined as 'arm drop', theta 1 The constant is defined as 'motionless', and the judgment result is recorded;
(2) And then to theta 2 To determine if theta 2 The reduction, the movement range is 165-0 degree, then the definition is "bent arm", if θ 2 The movement range is increased to be 0-165 degrees and is defined as an extending arm, theta 2 The constant is defined as 'motionless', and the judgment result is recorded;
(3) And then to theta 3 To determine if theta 3 The movement range is increased to 0 to +90 degrees, the operation is defined as pointing, if theta 3 The reduction and the movement range is between 0 and minus 90 degrees, and the definition is "lifting the hand", theta 3 The constant is defined as 'motionless', and the judgment result is recorded;
(4) Then to theta 4 To determine if theta 4 The movement range is increased to 0-120 degrees, which is defined as 'leg lifting', if theta 4 The reduction and the movement range is 120-0 degrees, which is defined as 'leg putting', theta 4 The constant is defined as 'motionless', and the judgment result is recorded;
(5) And then to theta 5 To determine if theta 5 The reduction, the movement range is 135-0 degree, the definition is "leg bending", if θ 5 The movement range is increased to 0 to 135 degrees, and is defined as 'leg extension', theta 5 The constant is defined as 'motionless', and the judgment result is recorded;
(6) And then to theta 6 To determine if theta 6 The reduction, the activity range is 0 to minus 60 degrees, the foot is defined as 'lifting', if theta 6 Increase the range of motionAt 0 to +60 degrees, the term "falling foot" is defined as θ 6 The constant is defined as 'motionless', and the judgment result is recorded;
and finally, comprehensively judging six judging results, wherein the states of the arms and the legs corresponding to each action are different, so that the meaning of the action expression is deduced.
The application has the following beneficial effects:
1. the application combines the normal image and the infrared image as the input of the detection algorithm, and effectively solves the problem of night detection in the greenhouse.
2. The application adopts the method of selecting the human body region by the frame and then positioning the human body key points, thereby greatly improving the accuracy of positioning the human body key points and further increasing the capability of the algorithm for distinguishing human body actions with higher similarity.
3. The application provides a high-efficiency double-task algorithm, which shares the same encoder when the target detection task and the gesture detection task are executed, so that the calculated amount can be reduced, and the detection speed can be improved.
4. The application provides a new algorithm for detecting key points of a human body, which is used in more human-computer interaction scenes through human body actions.
5. The application integrates the functions of temperature, humidity, carbon dioxide concentration and soil pH value monitoring, can monitor and regulate the abnormal conditions in the greenhouse in an omnibearing way, and reduces crop loss.
6. The photosensitive module and the LED lamp management control module adopted by the application can monitor and regulate the illumination environment in the greenhouse in real time, thereby improving the photosynthesis efficiency of crops and improving the crop yield.
7. The intelligent lamp post provided by the application reduces the manpower requirement in the greenhouse to a certain extent, and can accurately and efficiently acquire real-time data in the greenhouse through the sensor, thereby greatly promoting the development of intelligent agriculture in China.
Drawings
FIG. 1 is a schematic diagram of a first intelligent light pole according to the present application;
FIG. 2 is a schematic diagram of a second intelligent light pole according to the present application;
FIG. 3 is a schematic diagram of the structure of the MOL-POSE algorithm of the present application;
FIG. 4 is a schematic diagram of the key parts of 16 human body regions for model detection proposed in the present application;
FIG. 5 is a diagram showing the effect of detection according to the embodiment of the present application;
FIG. 6 is a block diagram of a monitoring device according to the present application;
FIG. 7 is a schematic diagram of a TSPP module used by the gesture detection network of the present application;
fig. 8 is a PIX-up block diagram of the gesture detection network according to the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments.
Referring to fig. 1-8, an integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel comprises an integrated intelligent lamp post and a monitoring device arranged on the integrated intelligent lamp post, wherein the monitoring device comprises an intelligent monitoring module, an LED lamp management control module, a photoresistor module, a temperature monitoring module, an air conditioning module, a humidity monitoring module, a humidifier module, a carbon dioxide monitoring module, a soil pH value monitoring module, a solar panel module and a main control module; the intelligent monitoring module, the LED lamp management control module, the air conditioning module, the humidifier module, the carbon dioxide monitoring module, the soil pH value monitoring module and the solar panel module are electrically connected with the main control module.
The intelligent monitoring module collects normal light image information and infrared image information in the greenhouse through the visual sensor, and adopts MOL-POSE algorithm to process the collected normal light image information, and infrared images assist in providing the distance of a detection target. The method comprises the steps of firstly selecting a region of a human body by a frame, identifying coordinates of 16 key points of the human body in the region, obtaining state information of the current human body posture by adopting a state transition model according to the position relation of the coordinates, and further judging whether detected personnel are reasonably operating the greenhouse.
The system comprises a temperature monitoring module, an air conditioning module, a humidity monitoring module, a humidifier module, a carbon dioxide monitoring module and a soil pH value monitoring module, wherein the temperature monitoring module, the air conditioning module, the humidity monitoring module, the humidifier module, the carbon dioxide monitoring module and the soil pH value monitoring module are used for detecting and adjusting the internal environment state of the greenhouse.
Specifically, the MOL-POSE algorithm can simultaneously complete the gesture recognition task and the target detection task. The MOL-post algorithm includes a backbone network and an output network Head structure. The backbone network selects a lightweight network DI-MobileNet, and a scale pyramid is constructed to extract features of different scales of the input image; the normal light image information and the infrared image information are combined to be used as the input of MOL-POSE, so that the problem of night detection in the greenhouse is solved. The structure of the MOL-POSE algorithm is shown in FIG. 3.
Specifically, an encoder shared by target detection and gesture detection adopts a lightweight network DI-MobileNet, and 1/16, 1/32 and 1/64-level bottom features are obtained through downsampling operation; because the 1/64-level bottom layer feature map is smaller and faces the limitation of width and height, a TSPP module is added into the 1/64-level bottom layer feature map for enhancing the multi-scale feature extraction of the single feature map; taking the multi-scale features as three inputs of a Head model, wherein the Head model comprises three continuous groups of up-sampling and convolution operations, and the up-sampling adopts a PIX-up module to effectively amplify a small feature map; meanwhile, stacking the results of the up-sampling and convolution operation and the features from the Backbone model, and performing convolution operation on the stacked results to obtain three paths of outputs; finally, the attitude task directly returns to obtain 16 human body key point normalized coordinates through a Head structure; the target detection task obtains a multi-scale prediction frame endpoint result.
Specifically, the Head structure decodes the different feature maps based on 1/16, 1/32 and 1/64 level bottom features to obtain a prediction result, namely, the upper left endpoint coordinate (C x ,C y ) And lower right endpoint coordinates (D x ,D y ) And according to the reduction ratio of the corresponding feature map to the original map, performing inverse transformation, taking the lower right endpoint coordinate as an example, and the formula is as follows:
in D' x ,D' y Is the inverse transformed abscissa and ordinate, D x ,D y For the inverse transformation pre-coordinates, K is the reduction ratio, taking the reciprocal of the reduction ratio corresponding to the 1/16, 1/32 and 1/64 level bottom feature map for 16, 32 and 64 respectively;
according to the point coordinates based on the original image size obtained by inverse transformation, the average value of the horizontal coordinate and the vertical coordinate is obtained by taking three points with similar positions as a group, and the average value point is taken as the final prediction frame endpoint.
Specifically, as shown in fig. 7, the TSPP module uses depth separable convolution to replace a common convolution block, so as to enlarge the 1/64-level bottom layer feature map receptive field; taking a 1/64-level bottom layer feature map as input, passing through five channels: in the first channel, depth separable convolution is not added, and the traditional 1X1 point convolution operation is used; processing the image by adopting a 3X3 convolution kernel and a depth separable convolution with a separation degree of 6 in the second channel; the third channel adopts a 3X3 convolution kernel and a depth separable convolution with a degree of separation of 12 to process the image; the fourth channel adopts a 3X3 convolution kernel and a depth separable convolution with a separation degree of 18 to process the image; the fifth channel directly carries out pooling operation on the original image and completes one-time up-sampling operation; finally, the five-channel results are all stacked together, reduced in depth using point convolution and output.
Specifically, as shown in fig. 8, the PIX-up module includes a convolution operation with a convolution kernel of 3X3 and a step size of 1, a normalization operation, an activation operation, and a pixel rearrangement operation; the pixel rearrangement operation converts the original characteristic diagram with the size of (h, w, r) into the characteristic with the size of (h, w, r, C), wherein h and w are the height and width of the original characteristic diagram, r is the up-sampling expansion multiple, and C is the number of channels after conversion.
Specifically, the lightweight network DI-MobileNet is improved on the basis of MobileNet, firstly 12 and 13 of MobileNet, an average pooling layer and a full connection layer are removed, then the stride of a 6 th layer is changed into 1/2 of the original one, and finally a hole convolution layer is added in the 7 th layer;
in a convolutional networkThe deeper the network layer number is, the larger the receptive field of the network is, the receptive field of the network can be increased by the hole convolution under the condition that the convolution layers are not stacked, and the sizes of the hole convolution and the convolution kernel of the common convolution are k s The number of holes in the hole convolution is d r The size of the common convolution kernel of the cavity convolution equivalent is K, and the equivalent conversion formula is as follows:
K=k s +(k s -1)×(d r -1)
wherein K is the size of a common convolution kernel equivalent to cavity convolution, and d r Number of holes, k, which is a convolution of holes s For the convolution kernel size of the cavity convolution, the calculation formula of the receptive field of the cavity convolution is as follows:
t in n Indicating the size of the n-layer cavity convolution receptive field, T n-1 The size of the n-1 layer cavity convolution receptive field is represented, K is the convolution kernel size of the common convolution,representing the current step size.
The 16 key points of the human body are identified as shown in fig. 4, and are connected according to the key points of the human body to form corresponding vectors. Table 1 shows the key parts of the 16 human body regions detected by the model proposed in this example.
TABLE 1 corresponding sequence numbers of key parts of human body
As shown in fig. 5, fig. 5 is an effect diagram of actual detection according to the present application, and key point information of a human body can be accurately located.
Specifically, in the state transition model, the state information of the current human body posture is obtained by adopting the state transition model, namely, a method for converting a set of coordinates of one limb into multiple states. The specific operation is as follows:
the method comprises the steps of dividing the states of arms into 7 types of 'lifting arms', 'lowering arms', 'bending arms', 'extending arms', 'pointing hands', 'lifting hands', 'holding feet', 'bending legs', 'extending legs', 'lifting feet', 'holding feet' at any moment, and then converting coordinates of 4 key points of each arm and 4 key points of each leg into angles and matching the angles with the 14 types of states; each arm corresponds to 4 key points, namely four points of shoulder (A), elbow (B), wrist (C) and hand (D), from top to bottom, each leg corresponds to 4 key points, namely hip (E), knee (F), ankle (G), foot (H), from top to bottom, and corresponds to 8 groups of coordinates (x 1 ,y 1 ),(x 2 ,y 2 ),(x 3 ,y 3 ),(x 4 ,y 4 ),(x 5 ,y 5 ),(x 6 ,y 6 ),(x 7 ,y 7 ),(x 8 ,y 8 ) And uses the formula:
obtaining the distance between the points;
definition d respectively 1 ,d 2 ,d 3 ,d 4 ,d 5 ,d 6 ,d 7 The distance is A to B, B to C, C to D, D to E, E to F, F to G, G to H;
and then the formula is utilized:
six angles of shoulder (A), elbow (B), wrist (C), hip (E), knee (F), ankle (G) are respectively defined as theta 1 ,θ 2 ,θ 3 ,θ 4 ,θ 5 ,θ 6 ;
And comprehensively measuring the posture of a person at the moment according to the obtained angle change.
Specifically, in the state transition model, the matching method of the obtained angle and the 14 types of states comprises the following steps:
(1) First to theta 1 To determine if theta 1 The movement range is increased to 0-145 degrees and is defined as 'lifting arm', if theta 1 The reduction and the movement range is 145-0 degrees, and the movement range is defined as 'arm drop', theta 1 The constant is defined as 'motionless', and the judgment result is recorded;
(2) And then to theta 2 To determine if theta 2 The reduction, the movement range is 165-0 degree, then the definition is "bent arm", if θ 2 The movement range is increased to be 0-165 degrees and is defined as an extending arm, theta 2 The constant is defined as 'motionless', and the judgment result is recorded;
(3) And then to theta 3 To determine if theta 3 The movement range is increased to 0 to +90 degrees, the operation is defined as pointing, if theta 3 The reduction and the movement range is between 0 and minus 90 degrees, and the definition is "lifting the hand", theta 3 The constant is defined as 'motionless', and the judgment result is recorded;
(4) Then to theta 4 To determine if theta 4 The movement range is increased to 0-120 degrees, which is defined as 'leg lifting', if theta 4 The reduction and the movement range is 120-0 degrees, which is defined as 'leg putting', theta 4 The constant is defined as 'motionless', and the judgment result is recorded;
(5) And then to theta 5 To determine if theta 5 The reduction, the movement range is 135-0 degree, the definition is "leg bending", if θ 5 The movement range is increased to 0 to 135 degrees, and is defined as 'leg extension', theta 5 The constant is defined as 'motionless', and the judgment result is recorded;
(6) And then to theta 6 To determine if theta 6 The reduction, the activity range is 0 to minus 60 degrees, the foot is defined as 'lifting', if theta 6 The movement range is increased to 0 to +60 degrees, and the movement range is defined as 'falling foot', theta 6 The constant is defined as 'motionless', and the judgment result is recorded;
and finally, comprehensively judging six judging results, wherein the states of the arms and the legs corresponding to each action are different, so that the meaning of the action expression is deduced.
As shown in fig. 6, fig. 6 is a relationship diagram among the modules of the present application, and the photoresistor module is used for collecting information of ambient illumination intensity, and the LED lamp management control module receives the information and can control the on/off of the LED lamp, the light intensity and the illumination time according to the information. In addition, the illumination lamp outside the greenhouse is also controlled by the module; the temperature monitoring module, the humidity monitoring module, the air conditioning module and the humidifier module are matched to realize the monitoring and regulation of the temperature and the humidity in the greenhouse. When the temperature and humidity sensor monitors that the temperature and humidity information in the greenhouse exceeds a threshold value, an alarm is sent to a user through an application program, and the temperature and humidity state in the greenhouse is automatically adjusted through an air conditioner and a humidifier; and the solar module is powered by solar energy if the electric quantity of the solar panel is sufficient, and automatically converts the electric quantity of the solar panel into power supply if the electric quantity of the solar panel is insufficient. The alarm function can timely inform the user of the damage information of the solar panel, so that the situation that the user cannot timely know the working condition of the solar panel is avoided; the main control module is internally provided with a processor, can automatically regulate and control the system according to various parameters of the environment, and is communicated with a user through Bluetooth to execute the instruction of the user preferentially.
Specifically, the intelligent lamp post is designed into two structures, wherein the first structure is a normal structure, as shown in fig. 1; the second is a simple structure for easy installation, as shown in fig. 2. The first normal structure is characterized in that two symmetrically arranged lamp posts 101 are used as main supporting bodies, the cross section of the main body of the lamp post is circular, and the ground contact area of the bottom end is large, so that the stability of the lamp post is facilitated. The indoor lighting lamp comprises a lamp post, wherein an air conditioner air outlet 102 is formed in the inner side of the lamp post, a first monocular camera 103 with a obliquely downward lens direction is arranged at the top of the lamp post, a temperature sensor 104, a humidity sensor 105 and a carbon dioxide concentration sensor 106 are respectively arranged at the lower end of the lamp post from top to bottom, an outdoor lighting street lamp 107 is arranged at the upper part of the outer side of the lamp post, an external main control module device 108 is arranged at the middle position of one lamp post, heat dissipation ports are formed in the front surface and the side surface of the main control module device, and a good working environment is provided for a built-in processor of the main control module device. The tops of the two lamp posts are connected through a cross beam, a second monocular camera 110 and a humidifier 111 are mounted on the lower side of the cross beam, a solar panel 109 is arranged on the upper surface of the cross beam, and the solar panel and a power supply power for the intelligent lamp post together. The skeleton 112 is equipped with a plurality of soil pH value sensors 114 at the contact place, and skeleton 112 is connected with the lamp pole both sides. An LED lamp tube is connected between the arc keels 113 on two sides of the framework to provide illumination for the greenhouse intelligence. As shown in fig. 1; during installation, after the distance is measured, two lamp posts are symmetrically installed, then the top carrier beam is installed, and then the solar module is installed. And measuring the distance between the brackets (the traditional mounting brackets of the greenhouse are used for supporting the keel structures, the traditional mounting brackets are not displayed in the drawing of the application), mounting a plurality of keel structures of the greenhouse (the traditional keel structures of the greenhouse are used for supporting the surface of the greenhouse, the traditional mounting brackets are not displayed in the drawing of the application), and mounting two ends of the LED lamp tube to the keel structures at two sides of the drawing of the application after all the brackets are mounted, so that the LED lamp tube management control module is connected with a lamp post main body, and the middle of the lamp tube is fixed to prevent the lamp tube from falling or breaking. And then installing a soil pH value sensor, and finally installing an external lighting street lamp outside the lamp post and an external main control module device in the middle.
As shown in fig. 2, two lamp posts 101 are still used as supporting bodies, the area of an air conditioner air outlet 102 is reduced and moved to the side, a monocular camera 103 with a obliquely downward lens direction is arranged at the top of the lamp post 101, a temperature sensor 104, a humidity sensor 105 and a carbon dioxide concentration sensor 106 are respectively arranged at the lower end of the lamp post 101 from top to bottom, a plurality of soil pH value sensors 109 are arranged at the bottom of the lamp post 101, an outdoor lighting street lamp 107 is arranged at the upper part of the outer side of the lamp post 101, an external main control module device 108 is arranged at the middle part of one lamp post, and heat dissipation ports are arranged at the front and the side of the main control module device to provide a good working environment for a built-in processor of the main control module device. The tops of the two lampposts are still connected by a beam 110, and the device on the beam is the same as the first normal structure. As shown in fig. 2; during installation, after the distance is measured, two lamp posts are symmetrically installed, then the top cross beam is installed, and then the solar module is installed. And measuring the distance between the brackets (the traditional bracket for supporting the keel structure is not displayed in the drawing of the application), installing a plurality of keel structures of the greenhouse (the traditional keel structure for supporting the surface of the greenhouse is used for supporting the surface of the greenhouse) after the brackets are installed, and finally installing an external lighting street lamp outside the lamp post and an external main control module device in the middle after the brackets are completely installed. The effect that the monitoring and regulation and control of first kind normal structure to the environment in the big-arch shelter obtained is optimal, and the installation effectiveness can be greatly promoted to the simple and easy structure of second kind.
The foregoing is only a preferred embodiment of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art, who is within the scope of the present application, should make equivalent substitutions or modifications according to the technical scheme of the present application and the inventive concept thereof, and should be covered by the scope of the present application.
Claims (9)
1. The integrated intelligent lamp post monitoring device for the operation management of farm greenhouse personnel is characterized by comprising an integrated intelligent lamp post and a monitoring device arranged on the integrated intelligent lamp post, wherein the monitoring device comprises an intelligent monitoring module, an LED lamp management control module, a photoresistor module, a temperature monitoring module, an air conditioning module, a humidity monitoring module, a humidifier module, a carbon dioxide monitoring module, a soil pH value monitoring module, a solar panel module and a main control module; the intelligent monitoring module, the LED lamp management control module, the air conditioning module, the humidifier module, the carbon dioxide monitoring module, the soil pH value monitoring module and the solar panel module are electrically connected with the main control module;
the intelligent monitoring module collects normal light image information and infrared image information in the greenhouse through the vision sensor, and adopts MOL-POSE algorithm to process the collected normal light image information, wherein the infrared image information is used for assisting in providing the distance of a detection target; firstly, selecting an area of a human body by a frame, identifying coordinates of 16 key points of the human body in the area, and obtaining state information of the current human body posture by adopting a state conversion model according to the position relation of the coordinates so as to judge whether detected personnel are reasonably operating the greenhouse;
the temperature monitoring module, the air conditioning module, the humidity monitoring module, the humidifier module, the carbon dioxide monitoring module and the soil pH value monitoring module are used for detecting and adjusting the internal environment state of the greenhouse.
2. The integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel according to claim 1, wherein the MOL-POSE algorithm is used for simultaneously completing a gesture recognition task and a target detection task and comprises a backbone network and an output network Head structure; the backbone network selects a lightweight network DI-MobileNet, and a scale pyramid is constructed to extract features of different scales of an input image; the normal light image information and the infrared image information are combined as inputs to the MOL-POSE algorithm.
3. The integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel according to claim 2, wherein the lightweight network DI-MobileNet obtains 1/16, 1/32 and 1/64-level bottom layer characteristics through downsampling operation; a TSPP module is added into the 1/64-level bottom layer features and used for strengthening and extracting multi-scale features of the single feature map; taking the multi-scale features as three inputs of a Head structure, wherein the Head structure comprises three groups of continuous up-sampling and convolution operations, and the up-sampling adopts a PIX-up module to effectively amplify a small feature map; meanwhile, stacking the results of the up-sampling and convolution operation and the bottom layer characteristics from the backbone network, and performing convolution operation on the stacked results to obtain three paths of outputs; finally, the attitude task directly returns to obtain 16 human body key point normalized coordinates through a Head structure; the target detection task obtains a multi-scale prediction frame endpoint result.
4. An integrated intelligent light pole monitoring device for operation management of farm greenhouse personnel according to claim 3, wherein the Head structure decodes different feature graphs based on 1/16, 1/32 and 1/64 level bottom features to obtain a prediction result, namely target inspectionMeasuring the upper left endpoint coordinate (C) x ,C y ) And lower right endpoint coordinates (D x ,D y ) And according to the reduction ratio of the corresponding feature map to the original map, performing inverse transformation, taking the lower right endpoint coordinate as an example, and the formula is as follows:
in D' x ,D' y Is the inverse transformed abscissa and ordinate, D x ,D y For the inverse transformation pre-coordinates, K is the reduction ratio, taking the reciprocal of the reduction ratio corresponding to the 1/16, 1/32 and 1/64 level bottom feature map for 16, 32 and 64 respectively;
according to the point coordinates based on the original image size obtained by inverse transformation, the average value of the horizontal coordinate and the vertical coordinate is obtained by taking three points with similar positions as a group, and the average value point is taken as the final prediction frame endpoint.
5. The integrated intelligent light pole monitoring device for operation management of farm greenhouse personnel according to claim 3, wherein the TSPP module replaces a common convolution block with depth separable convolution, and enlarges a 1/64-level bottom layer feature map receptive field; taking a 1/64-level bottom layer feature map as input, passing through five channels: in the first channel, depth separable convolution is not added, and the traditional 1X1 point convolution operation is used; processing the image by adopting a 3X3 convolution kernel and a depth separable convolution with a separation degree of 6 in the second channel; the third channel adopts a 3X3 convolution kernel and a depth separable convolution with a degree of separation of 12 to process the image; the fourth channel adopts a 3X3 convolution kernel and a depth separable convolution with a separation degree of 18 to process the image; the fifth channel directly carries out pooling operation on the original image and completes one-time up-sampling operation; finally, the five-channel results are all stacked together, reduced in depth using point convolution and output.
6. An integrated intelligent light pole monitoring device for farm greenhouse personnel operation management according to claim 3, wherein the PIX-up module comprises a convolution operation with a convolution kernel of 3X3 and a step size of 1, a normalization operation, an activation operation and a pixel rearrangement operation; the pixel rearrangement operation converts the original characteristic diagram with the size of (h, w, r) into the characteristic with the size of (h, w, r, C), wherein h and w are the height and width of the original characteristic diagram, r is the up-sampling expansion multiple, and C is the number of channels after conversion.
7. The integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel according to claim 2, wherein the lightweight network DI-MobileNet is improved on the basis of MobileNet, firstly 12 and 13 of MobileNet, an average pooling layer and a full connection layer are removed, then the stride of a 6 th layer is changed into 1/2 of the original one, and finally a hole convolution layer is added in a 7 th layer;
in the convolution network, the deeper the network layer number is, the larger the receptive field of the network is, the receptive field of the network can be increased by the cavity convolution without stacking the convolution layers, and the sizes of the convolution kernels of the cavity convolution and the common convolution are k s The number of holes in the hole convolution is d r The size of the common convolution kernel of the cavity convolution equivalent is K, and the equivalent conversion formula is as follows:
K=k s +(k s -1)×(d r -1)
wherein K is the size of a common convolution kernel equivalent to cavity convolution, and d r Number of holes, k, which is a convolution of holes s For the convolution kernel size of the cavity convolution, the calculation formula of the receptive field of the cavity convolution is as follows:
t in n Indicating the size of the n-layer cavity convolution receptive field, T n-1 The size of the n-1 layer cavity convolution receptive field is represented, K is the convolution kernel size of the cavity convolution equivalent general convolution,representing the current step size.
8. The integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel according to claim 1, wherein in the state conversion model, the states of arms are divided into 7 types of 'lifting arm', 'lowering arm', 'bending arm', 'extending arm', 'pointing hand', 'lifting hand', 'holding leg', and the leg states are divided into 7 types of 'lifting leg', 'putting leg', 'extending leg', 'lifting foot', 'holding foot', and the like at any time, and then coordinates of 4 key points of each arm and 4 key points of each leg are converted into angles and matched with the 14 types of states; each arm corresponds to 4 key points, namely four points of shoulder (A), elbow (B), wrist (C) and hand (D), from top to bottom, each leg corresponds to 4 key points, namely hip (E), knee (F), ankle (G), foot (H), from top to bottom, and corresponds to 8 groups of coordinates (x 1 ,y 1 ),(x 2 ,y 2 ),(x 3 ,y 3 ),(x 4 ,y 4 ),(x 5 ,y 5 ),(x 6 ,y 6 ),(x 7 ,y 7 ),(x 8 ,y 8 ) And uses the formula:
obtaining the distance between the points;
definition d respectively 1 ,d 2 ,d 3 ,d 4 ,d 5 ,d 6 ,d 7 The distance is A to B, B to C, C to D, D to E, E to F, F to G, G to H;
and then the formula is utilized:
six angles of shoulder (A), elbow (B), wrist (C), hip (E), knee (F), ankle (G) are respectively defined as theta 1 ,θ 2 ,θ 3 ,θ 4 ,θ 5 ,θ 6 ;
And comprehensively measuring the posture of a person at the moment according to the obtained angle change.
9. The integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel according to claim 8, wherein the matching method of the obtained angle and the 14 types of states in the state transition model is as follows:
(1) First to theta 1 To determine if theta 1 The movement range is increased to 0-145 degrees and is defined as 'lifting arm', if theta 1 The reduction and the movement range is 145-0 degrees, and the movement range is defined as 'arm drop', theta 1 The constant is defined as 'motionless', and the judgment result is recorded;
(2) And then to theta 2 To determine if theta 2 The reduction, the movement range is 165-0 degree, then the definition is "bent arm", if θ 2 The movement range is increased to be 0-165 degrees and is defined as an extending arm, theta 2 The constant is defined as 'motionless', and the judgment result is recorded;
(3) And then to theta 3 To determine if theta 3 The movement range is increased to 0 to +90 degrees, the operation is defined as pointing, if theta 3 The reduction and the movement range is between 0 and minus 90 degrees, and the definition is "lifting the hand", theta 3 The constant is defined as 'motionless', and the judgment result is recorded;
(4) Then to theta 4 To determine if theta 4 The movement range is increased to 0-120 degrees, which is defined as 'leg lifting', if theta 4 The reduction and the movement range is 120-0 degrees, which is defined as 'leg putting', theta 4 The constant is defined as 'motionless', and the judgment result is recorded;
(5) And then to theta 5 To determine if theta 5 The reduction, the movement range is 135-0 degree, the definition is "leg bending", if θ 5 The movement range is increased to 0 to 135 degrees, and is defined as 'leg extension', theta 5 The constant is defined as 'motionless', and the judgment result is recorded;
(6) And then to theta 6 To determine if theta 6 The reduction, the activity range is 0 to minus 60 degrees, the foot is defined as 'lifting', if theta 6 The movement range is increased to 0 to +60 degrees, and the movement range is defined as 'falling foot', theta 6 The constant is defined as 'motionless', and the judgment result is recorded;
and finally, comprehensively judging six judging results, wherein the states of the arms and the legs corresponding to each action are different, so that the meaning of the action expression is deduced.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310923981.6A CN117173533A (en) | 2023-07-26 | 2023-07-26 | Integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310923981.6A CN117173533A (en) | 2023-07-26 | 2023-07-26 | Integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117173533A true CN117173533A (en) | 2023-12-05 |
Family
ID=88938268
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310923981.6A Pending CN117173533A (en) | 2023-07-26 | 2023-07-26 | Integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117173533A (en) |
-
2023
- 2023-07-26 CN CN202310923981.6A patent/CN117173533A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN206165346U (en) | Remote visualization intelligence irrigation system based on thing networking | |
CN103891697B (en) | The variable spray method of a kind of indoor autonomous spraying machine device people | |
CN102564593A (en) | Plant growth condition monitoring system based on compute vision and internet of things | |
CN206671924U (en) | A kind of intelligent plant maintenance and management control system based on technology of Internet of things | |
CN209313951U (en) | A kind of greenhouse intelligent movable monitoring system based on ARM9 | |
CN107745756B (en) | The traveling method of six sufficient strawberry picking robots | |
CN103699095A (en) | Greenhouse plant growth posture monitoring system based on binocular stereo vision and greenhouse plant growth posture monitoring method based on binocular stereo vision | |
CN102960197A (en) | Spatial type intelligent seedling-raising robot platform suitable for plant industrialized production | |
CN101356877A (en) | Cucumber picking robot system and picking method in greenhouse | |
CN113406899A (en) | Intelligent agricultural monitoring system based on Internet of things | |
CN108830741A (en) | A kind of farm environment intelligent monitor system | |
CN111160271B (en) | Indoor chicken farm disease chicken detecting system based on deep learning | |
CN206178392U (en) | Real -time reciprocity monitoring devices of greenhouse crop growth information | |
Mount et al. | 2d visual place recognition for domestic service robots at night | |
Ji et al. | Research on key technology of truss tomato harvesting robot in greenhouse | |
CN114745834A (en) | Wisdom street lamp subregion lighting control management system | |
CN117173533A (en) | Integrated intelligent lamp post monitoring device for operation management of farm greenhouse personnel | |
CN114779692A (en) | Linear sliding table type weeding robot and control method thereof | |
CN204790646U (en) | Intelligent greenhouse of agricultural | |
CN113344035A (en) | Banana phenological period monitoring module and planting system | |
CN203724719U (en) | Robot used for picking up table tennis ball | |
CN104751455A (en) | Crop image dense matching method and system | |
CN110930284A (en) | Intelligent scenic spot service method and system based on municipal garden | |
CN205809599U (en) | A kind of greenhouse-environment monitoring system based on Internet of Things | |
Li et al. | Design of multifunctional seedbed planting robot based on MobileNetV2-SSD |
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 |