CN113269275B - Real-time detection method for cocoons under cocoons - Google Patents
Real-time detection method for cocoons under cocoons Download PDFInfo
- Publication number
- CN113269275B CN113269275B CN202110685414.2A CN202110685414A CN113269275B CN 113269275 B CN113269275 B CN 113269275B CN 202110685414 A CN202110685414 A CN 202110685414A CN 113269275 B CN113269275 B CN 113269275B
- Authority
- CN
- China
- Prior art keywords
- cocoon
- cocoons
- conveying
- linear module
- model
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000011897 real-time detection Methods 0.000 title claims abstract description 16
- 238000001514 detection method Methods 0.000 claims abstract description 35
- 238000013138 pruning Methods 0.000 claims abstract description 30
- 241000255789 Bombyx mori Species 0.000 claims abstract description 15
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 6
- 210000000003 hoof Anatomy 0.000 claims description 32
- 238000012549 training Methods 0.000 claims description 31
- 241000252254 Catostomidae Species 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 15
- 238000002372 labelling Methods 0.000 claims description 6
- 238000013459 approach Methods 0.000 claims description 4
- 238000007621 cluster analysis Methods 0.000 claims description 4
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000007599 discharging Methods 0.000 claims description 3
- 230000001131 transforming effect Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 15
- 238000004519 manufacturing process Methods 0.000 abstract description 13
- 238000012545 processing Methods 0.000 abstract description 9
- 238000013135 deep learning Methods 0.000 abstract description 3
- 230000006872 improvement Effects 0.000 abstract description 3
- 239000010410 layer Substances 0.000 description 23
- 238000010586 diagram Methods 0.000 description 10
- 230000000694 effects Effects 0.000 description 8
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 description 6
- 229910052782 aluminium Inorganic materials 0.000 description 6
- 230000001965 increasing effect Effects 0.000 description 4
- 235000008708 Morus alba Nutrition 0.000 description 3
- 240000000249 Morus alba Species 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 239000002356 single layer Substances 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000005286 illumination Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01K—ANIMAL HUSBANDRY; AVICULTURE; APICULTURE; PISCICULTURE; FISHING; REARING OR BREEDING ANIMALS, NOT OTHERWISE PROVIDED FOR; NEW BREEDS OF ANIMALS
- A01K67/00—Rearing or breeding animals, not otherwise provided for; New or modified breeds of animals
- A01K67/033—Rearing or breeding invertebrates; New breeds of invertebrates
- A01K67/04—Silkworms
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G47/00—Article or material-handling devices associated with conveyors; Methods employing such devices
- B65G47/02—Devices for feeding articles or materials to conveyors
- B65G47/16—Devices for feeding articles or materials to conveyors for feeding materials in bulk
- B65G47/18—Arrangements or applications of hoppers or chutes
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G47/00—Article or material-handling devices associated with conveyors; Methods employing such devices
- B65G47/34—Devices for discharging articles or materials from conveyor
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G47/00—Article or material-handling devices associated with conveyors; Methods employing such devices
- B65G47/74—Feeding, transfer, or discharging devices of particular kinds or types
- B65G47/90—Devices for picking-up and depositing articles or materials
- B65G47/91—Devices for picking-up and depositing articles or materials incorporating pneumatic, e.g. suction, grippers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- 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
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Environmental Sciences (AREA)
- Zoology (AREA)
- Animal Husbandry (AREA)
- Biodiversity & Conservation Biology (AREA)
- Animal Behavior & Ethology (AREA)
- Probability & Statistics with Applications (AREA)
- Sorting Of Articles (AREA)
Abstract
The invention relates to a method for detecting cocoons under cocoons in real time, and belongs to the technical field of processing in the silkworm industry. According to the method, the real-time detection of the cocoons is carried out during cocoon selection in the production and processing process of the cocoons after the parameter presetting based on an anchor point frame, channel pruning and embedding receptive field module improvement are carried out on a YOLOv3 deep learning target detection algorithm. The feeding device of the cocoon sorting machine is used for feeding cocoons, the conveying device is used for conveying cocoons, the image acquisition device is used for photographing cocoons on the conveying device, cocoon images are obtained, the lower cocoon grabbing device is used for grabbing cocoons and removing cocoons, and the upper cocoon storage device is used for storing cocoons. The cocoon sorting machine picks and eliminates the lower cocoons according to the coordinates output by the real-time detection method of the lower cocoons, and collects and stores the upper cocoons, so that the production efficiency and the cocoon silk quality of the production and processing of the cocoons are improved, and the production cost is reduced. The silkworm cocoon sorting machine has the characteristics of simple structure, accurate sorting quality, high working efficiency and the like.
Description
Technical Field
The invention relates to a method for detecting cocoons under cocoons in real time, and belongs to the technical field of processing in the silkworm industry.
Background
The cocoon selecting process in the silk cocoon production and processing process refers to a process of removing the cocoon which cannot be reeled according to the silk making process requirement, and in order to improve the cocoon silk quality, the cocoon selecting process is needed to be carried out on the cocoon before reeling. At present, the common cocoon selecting method mainly comprises manual visual inspection and manual sorting. The method is easily influenced by subjective consciousness and emotion of people, has high labor intensity and low efficiency, and is large in distinction between people and individuals, so that judgment cannot be performed according to unified standards. In order to realize sustainable development of the mulberry industry, realize modernization of the mulberry industry in China, the mechanization and the intellectualization of the cocoon selecting machine are necessary ways for increasing the added value of products of mulberry enterprises, improving the technical competitiveness and realizing sustainable development.
Disclosure of Invention
Aiming at the problem of existing cocoon sorting, the invention provides a method for detecting cocoons in real time, which carries out anchor point frame parameter presetting, channel pruning and embedding receptive field module improvement on a YOLOv3 deep learning target detection algorithm to carry out real-time detection on cocoons in cocoon sorting in the process of cocoon production and processing; the cocoon sorting machine picks and eliminates the lower cocoons according to the coordinates output by the real-time detection method of the lower cocoons, and collects and stores the upper cocoons, so that the production efficiency and the cocoon silk quality of the production and processing of the cocoons are improved, and the production cost is reduced. The silkworm cocoon sorting machine has the characteristics of simple structure, accurate sorting quality, high working efficiency and the like.
The invention adopts the technical proposal for solving the technical problems that:
a method for detecting cocoons under cocoons in real time comprises the following specific steps:
1) Taking thin cocoon, flat cocoon, extra small cocoon and macular cocoon as detection targets, shooting the detection targets at different positions and different angles, and collecting a cocoon discharging detection data set;
2) Transforming the cocoon detection data set into 416×416 images by using an interpolation algorithm, and dividing the images into a training set and a test set; the image quantity ratio of the training set to the test set is 3:1;
3) Labeling images in the training set and the test set by using a label I mg labeling tool; marking a boundary box of a corresponding cocoon type for all the identified objects on each image;
4) Training a model by using an image marked in a training set by taking a YOLOv3 model as a basic model, performing super-parameter searching and evolution to obtain a new generation of evolution, judging whether to evolve or not through a fitness function, wherein the higher the fitness is, the better the performance of the generation is represented, so that the super-parameter suitable for cocoon detection data set is selected through evolution searching;
5) Setting evolution super parameters of the improved YOLOv3 model, and loading the pre-training weight into the improved YOLOv3 model for training to obtain an optimal training weight;
6) Loading the optimal training weight into the improved YOLOv3 model to test the images in the test set;
7) Repeating the steps 5) and 6) to obtain an optimal YOLOv3 model as a cocoon setting real-time detection method.
The step 5) improves the YOLOv3 model to be:
a. selecting anchor frame parameters suitable for the cocoon detection data set by carrying out K-means cluster analysis on the size and the number of anchor frames of the YOLOv3 model;
b. compressing a YOLOv3 model by adopting a model pruning method based on batch regularization layer (BN) scaling factors after channel sparsification;
c. the receptive field module RFB was embedded in the YOLOv3 model to increase the local receptive field of the model.
Further, the specific method for compressing the YOLOv3 model by adopting the model pruning method based on batch regularization layer (BN) scaling factor after channel sparsification comprises the following steps of
And introducing a scaling factor gamma to each channel of the BN layer in the YOLOv3 model to carry out L1 regularization, and if the scaling factor gamma approaches 0 under the constraint of L1 regularization, carrying out pruning by the channel with low contribution degree to feature map prediction in the model.
Furthermore, the scaling factor gamma is a parameter to be trained, and the value is 0< gamma;
the silkworm cocoon sorting machine adopted by the silkworm cocoon lower cocoon real-time detection method comprises a feeding device, a conveying device, an image acquisition device, a lower cocoon grabbing device and an upper cocoon storage device which are connected in sequence,
the feeding device is used for feeding cocoons;
the conveying device is used for conveying cocoons;
the image acquisition device is used for photographing cocoons on the conveying device to obtain cocoon images;
the cocoon picking device is used for picking and removing cocoons;
the cocoon picking device is used for storing cocoons;
the upper cocoon storage device is used for storing upper cocoons;
the feeding device comprises a conveying hopper 1, and the conveying hopper 1 is arranged at the feeding end of the conveying device; the conveying device comprises a conveying belt 6, a conveying device frame 24 and a motor 30, wherein the conveying belt 6 is arranged at the top end of the conveying device frame 24, the motor 30 is arranged at the side end of the top of the conveying device frame 24, and an output shaft of the motor 30 is in transmission connection with the conveying belt 6; the image acquisition device comprises an industrial camera 8, and the industrial camera 8 is arranged right above the middle part of the conveying device; the cocoon picking device comprises a plurality of suckers 10, and the suckers 10 are arranged above the tail end of the conveying device; the cocoon picking device comprises a storage hopper I20, the storage hopper I20 is arranged at the side end of the conveyor belt 6, and when cocoons are stored, the storage hopper I20 is positioned under the sucking disc 10; the upper cocoon storage device is arranged right below the tail end of the conveyor belt 6;
preferably, the upper cocoon storage device is a storage hopper II 36;
preferably, the initial end of the conveyor belt 6 is coated on the driven rotating shaft, the tail end of the conveyor belt 6 is coated on the driving rotating shaft, the driven rotating shaft and the driving rotating shaft are arranged in parallel, and the output shaft of the motor 30 is connected with the end head of the driving rotating shaft;
preferably, the cocoon sorting machine further comprises a controller 25, and the conveying device, the image acquisition device and the cocoon grabbing device are all connected with the controller 25;
more preferably, the motor 30, the industrial camera 8 and the suction cup 10 are all connected to the controller 25;
the feeding device further comprises a conveying hopper mounting support 26, and the conveying hopper 1 is fixedly arranged at the top end of the conveying hopper mounting support 26;
further, the feeding device further comprises a conveying hopper mounting corner fitting 3, foot hooves I27 and a foot hoof mounting plate I28, wherein the output end of the conveying hopper 1 is fixedly arranged at the feeding end of the conveying device through the conveying hopper mounting corner fitting 3, and the foot hooves I27 are arranged at the bottom end of the conveying hopper mounting support 26 through the foot hoof mounting plate I28;
the conveying device further comprises a conveying belt guide rail 29, a front baffle plate 2, a hairbrush 5 and baffle plates 23, wherein the conveying belt guide rail 29 is fixedly arranged at the top end of a conveying device frame 24, the baffle plates 23 are vertically arranged on the conveying belt guide rail 29 and are positioned on two sides of a conveying belt 6, the front baffle plate 2 is fixedly arranged at the initial end of the conveying belt guide rail 29, the hairbrush 5 is fixedly arranged at the top end of the conveying device frame 24 and is positioned at the feeding end of the conveying device; the two sides of the conveyor belt 6 are arranged in the conveyor belt guide rail 29 in a sliding way;
the driven rotating shaft and the driving rotating shaft are respectively arranged at two ends of the conveying device frame 24, and the driven rotating shaft and the driving rotating shaft can rotate at two ends of the conveying device frame 24;
the front baffle plate 2 can prevent cocoons from falling off after being fed by the feeding device, the baffle plate 23 can prevent cocoons from falling off in the conveying process, the hairbrush 5 can brush cocoons on the conveying belt flat, the cocoons are ensured not to be stacked, and one layer of cocoons are ensured to be conveyed forwards each time;
further, the conveying device further comprises foot hooves II and a foot hoof mounting plate II, the foot hooves II are arranged at the bottom end of the conveying device frame 24 through the foot hoof mounting plate II, and the hairbrush 5 is fixedly arranged at the top end of the conveying device frame 24 through the hairbrush mounting plate 4;
the image acquisition device further comprises an image acquisition device mounting support 22 and LED light sources 33, the industrial camera 8 is fixedly arranged at the center of the top end of the image acquisition device mounting support 22, the LED light sources 33 are fixedly arranged at the top end of the image acquisition device mounting support 22, and the LED light sources 33 are positioned on two sides of the industrial camera 8;
preferably, the LED light source 33 is connected to the controller 25; the LED light source is used for providing stable illumination conditions;
further, the image acquisition device further comprises a camera mounting plate 7 and a camera mounting corner piece 9, wherein the camera mounting plate 7 is horizontally arranged at the top end of the image acquisition device mounting support 22, the camera mounting corner piece 9 is vertically arranged at the center of the top end of the camera mounting plate 7, the bottom end of the industrial camera 8 is fixedly arranged on the camera mounting plate 7, the rear side surface of the industrial camera 8 is fixedly arranged on the camera mounting corner piece 9, and the bottom end of the image acquisition device mounting support 22 is provided with a foot shoe III 31 through a foot shoe mounting plate III 32;
the cocoon grabbing device further comprises a linear module I11, a linear module II 16, a linear module III 18, a supporting beam 13, a sucker mounting plate support 14 and a cocoon grabbing device support 21, wherein the linear module I11 and the linear module III 18 are horizontally and parallelly arranged at the top end of the cocoon grabbing device support 21, the linear module I11 and the linear module III 18 are perpendicular to the conveying direction of the conveying belt 6, two ends of the supporting beam 13 are respectively arranged at the top ends of the linear module I11 and the linear module III 18 in a sliding mode, the linear module II 16 is vertically arranged in the middle of the supporting beam 13, the sucker 10 is arranged at the bottom end of the sucker mounting plate 12, the sucker mounting plate 12 is arranged at the bottom end of the linear module II 16 through a connecting beam 17, the storage hopper I20 is arranged in the middle of the cocoon grabbing device support 21 and is positioned at one side of the conveying belt 6, and the storage hopper I20 and the conveying belt 6 are respectively positioned under the linear module I11 and the linear module III 18;
preferably, the linear module I11, the linear module II 16 and the linear module III 18 are all connected with the controller 25; the linear module I11 can be an MDL80 aluminum-based series linear module of Shenzhen Wifar precision technology Co., ltd, and has the specification and model number of: the MDL80-ML1-1600-100BH-E-G-W, and the linear module II 16 can be an MSX45 aluminum-based linear module of Shenzhen Wifar precision technology Co., ltd, and has the specification and model number of: the MDL80-ML1-1600-100BH-E-G-W, and the linear module III 18 is an MDL80 aluminum-based linear module of Shenzhen Wifar precision technology Co., ltd, and has the specification and model number of: MDL80-MR1-1600-100BH-E-G-W;
further, the bottom end of the cocoon grabbing device support 21 is provided with foot hooves IV 34 through foot hoof mounting plates IV 35, the suction disc 10 is arranged at the bottom end of the suction disc mounting plate support 14 through the suction disc mounting plate 12, and the linear module II 16 is arranged in the middle of the supporting beam 13 through the linear module mounting corner fitting 15;
preferably, the conveyor belt 6 is a flexible conveyor belt, a plurality of grooves are uniformly formed in the conveyor belt 6, the sizes of the grooves are matched with the structure of the cocoons, one cocoon is placed in each groove, and the cocoons are always positioned in the grooves in the transportation process, so that the cocoons in the transportation process can be conveniently detected and positioned; the inner wall of the sucker 10 is fixedly provided with a sponge layer, and the sponge layer can prevent the cocoons from being crushed when the cocoons are grabbed;
further, the cocoon loading storage device further comprises a storage hopper mounting support 19, the storage hopper II 36 is fixedly arranged at the top end of the storage hopper mounting support 19, and foot hooves V are arranged at the bottom end of the storage hopper mounting support 19 through foot hoof mounting plates V.
The silkworm cocoon sorting method based on the silkworm cocoon sorting machine comprises the following specific steps:
1) The cocoons are fed by a feeding mechanism and automatically fall on a conveyor belt, and are paved on the conveyor belt in a single layer for conveying;
2) When cocoons reach the lower part of the image acquisition device, the conveyor belt stops transmitting, the industrial camera acquires images of cocoons and transmits the image data to the controller, and the controller detects the center point coordinates of cocoons;
3) The controller outputs the center point coordinates of the cocoons and judges the positions of the suckers corresponding to the center point coordinates;
4) When the cocoons are conveyed below the grabbing device, the conveying belt stops conveying, the sucking disc moves to the position right above the center point of the cocoons to suck the cocoons for grabbing;
5) After the cocoons are grabbed, the sucking discs move to the position right above the storage hopper I, the sucking discs desorb, and the cocoons fall into the storage hopper I, so that the cocoons are collected and stored;
6) The non-grabbed cocoons are upper cocoons, the upper cocoons are continuously conveyed by the conveyor belt, reach the tail end of the conveyor belt, and fall into a collecting and storing device of the upper car cocoons.
The invention has the beneficial effects that:
(1) According to the method, the real-time detection of the cocoons is carried out in the cocoon production and processing process by carrying out parameter presetting based on an anchor point frame, channel pruning and improvement of an embedded receptive field module on a YOLOv3 deep learning target detection algorithm; the cocoon sorting machine picks and eliminates the lower cocoons according to the coordinates output by the real-time detection method of the lower cocoons, and collects and stores the upper cocoons, so that the production efficiency and the cocoon silk quality of the production and processing of the cocoons are improved, and the production cost is reduced;
(2) The silkworm cocoon sorting machine has the characteristics of simple structure, accurate sorting quality, high working efficiency and the like.
Drawings
FIG. 1 is a schematic diagram of a cocoon sorting machine;
FIG. 2 is a schematic diagram of a loading device;
FIG. 3 is a schematic diagram of a conveyor apparatus;
FIG. 4 is a schematic diagram of an image acquisition device;
FIG. 5 is a schematic diagram of the structure of the cocoon picking device;
FIG. 6 is a schematic diagram of the structure of the upper cocoon storage device;
FIG. 7 is a schematic diagram of cocoon picking;
FIG. 8 is a flow chart of a method for detecting cocoons in real time
FIG. 9 is a graph showing the results of target statistics in cocoon detection data set
FIG. 10 is a schematic view of a channel pruning
FIG. 11 is a schematic diagram of a modified YOLOv3 model structure
In the figure: 1-hopper, 2-front baffle, 3-hopper mounting angle, 4-brush mounting plate, 5-brush, 6-flexible groove conveyor belt, 7-camera mounting plate, 8-industrial camera, 9-camera mounting angle, 10-suction cup, 11-straight module I, 12-suction cup mounting plate, 13-support beam, 14-suction cup mounting plate support, 15-straight module mounting angle, 16-straight module II, 17-connection beam, 18-straight module III, 19-storage hopper mounting support, 20-storage hopper I, 21-lower cocoon gripping device support, 22-image acquisition device mounting support, 23-baffle, 24-conveyor frame, 25-controller, 26-hopper mounting support, 27-foot shoe I, 28-foot shoe mounting plate I, 29-conveyor belt guide rail, 30-motor, 31-foot shoe III, 32-foot shoe mounting plate III, 33-LED light source, 34-foot shoe IV, 35-foot mounting plate IV, 36-storage hopper II.
Detailed Description
The invention will be further described with reference to the following specific embodiments.
Example 1: a method for detecting cocoons in real time (see figure 8) comprises the following specific steps:
1) Taking thin cocoon, flat cocoon, extra small cocoon and macular cocoon as detection targets, shooting the detection targets at different positions and different angles, and collecting a cocoon discharging detection data set;
2) Transforming the cocoon detection data set into 416×416 images by using an interpolation algorithm, and dividing the images into a training set and a test set; the image quantity ratio of the training set to the test set is 3:1;
3) Labeling images in the training set and the test set by using a label I mg labeling tool; marking a boundary box of a corresponding cocoon type for all the identified objects on each image;
4) Training the model by using an image marked in a training set by taking a YOLOv3 model as a basic model, performing super-parameter searching and evolution to obtain a new generation of evolution, judging whether to evolve or not by using a fitness function, and selecting super-parameters suitable for cocoon detection data sets by the evolution searching;
5) Setting evolution super parameters of the improved YOLOv3 model, and loading the pre-training weight into the improved YOLOv3 model for training to obtain an optimal training weight;
wherein the improved YOLOv3 model is
a. Selecting anchor frame parameters suitable for the cocoon detection data set by carrying out K-means cluster analysis on the size and the number of anchor frames of the YOLOv3 model;
b. compressing a YOLOv3 model by adopting a model pruning method based on batch regularization layer (BN) scaling factors after channel sparsification;
c. embedding receptive field module RFB in YOLOv3 model to increase local receptive field of the model;
a specific method for compressing the YOLOv3 model by adopting a model pruning method based on batch regularization layer (BN) scaling factors after channel sparsification is as follows
Introducing a scaling factor gamma to each channel of a BN layer in a YOLOv3 model to carry out L1 regularization, and if the scaling factor gamma approaches 0 under the constraint of L1 regularization, carrying out pruning on the channel with low contribution degree to feature map prediction in the model; the scaling factor gamma is a parameter to be trained, and the value of the scaling factor gamma is 0< gamma;
6) Loading the optimal training weight into the improved YOLOv3 model to test the images in the test set;
7) Repeating the steps 5) and 6) to obtain an optimal YOLOv3 model as a cocoon setting real-time detection method.
Example 2: as shown in figure 1, the cocoon sorting machine adopted by the cocoon lower cocoon real-time detection method comprises a feeding device, a conveying device, an image acquisition device, a lower cocoon grabbing device and an upper cocoon storage device which are connected in sequence,
the feeding device is used for feeding cocoons;
the conveying device is used for conveying cocoons;
the image acquisition device is used for photographing cocoons on the conveying device to obtain cocoon images;
the cocoon picking device is used for picking and removing cocoons;
the cocoon picking device is used for storing cocoons;
the upper cocoon storage device is used for storing upper cocoons;
the feeding device comprises a conveying hopper 1, and the conveying hopper 1 is arranged at the feeding end of the conveying device; the conveying device comprises a conveying belt 6, a conveying device frame 24 and a motor 30, wherein the conveying belt 6 is arranged at the top end of the conveying device frame 24, the motor 30 is arranged at the side end of the top of the conveying device frame 24, and an output shaft of the motor 30 is in transmission connection with the conveying belt 6; the image acquisition device comprises an industrial camera 8, and the industrial camera 8 is arranged right above the middle part of the conveying device; the cocoon picking device comprises a plurality of suckers 10, and the suckers 10 are arranged above the tail end of the conveying device; the cocoon picking device comprises a storage hopper I20, the storage hopper I20 is arranged at the side end of the conveyor belt 6, and when cocoons are stored, the storage hopper I20 is positioned under the sucking disc 10; the upper cocoon storage device is arranged right below the tail end of the conveyor belt 6;
the cocoon loading storage device is a storage hopper II 36;
the initial end of the conveyor belt 6 is coated on the driven rotating shaft, the tail end of the conveyor belt 6 is coated on the driving rotating shaft, the driven rotating shaft and the driving rotating shaft are arranged in parallel, and an output shaft of the motor 30 is connected with the end head of the driving rotating shaft;
the cocoon sorting machine further comprises a controller 25, and the conveying device, the image acquisition device and the cocoon grabbing device are all connected with the controller 25; the motor 30, the industrial camera 8 and the sucker 10 are all connected with the controller 25;
the silkworm cocoon sorting method based on the silkworm cocoon sorting machine comprises the following specific operation steps:
1) The cocoons are fed by a feeding mechanism and automatically fall on a conveyor belt, and are paved on the conveyor belt in a single layer for conveying;
2) When cocoons reach the lower part of the image acquisition device, the conveyor belt stops transmitting, the industrial camera acquires images of cocoons and transmits the image data to the controller, and the controller detects the center point coordinates of cocoons;
3) The controller outputs the center point coordinates of the cocoons and judges the positions of the suckers corresponding to the center point coordinates;
4) When the cocoons are conveyed below the grabbing device, the conveying belt stops conveying, the sucking disc moves to the position right above the center point of the cocoons to suck the cocoons for grabbing;
5) After the cocoons are grabbed, the sucking discs move to the position right above the storage hopper I, the sucking discs desorb, and the cocoons fall into the storage hopper I, so that the cocoons are collected and stored;
6) The non-grabbed cocoons are upper cocoons, the upper cocoons are continuously conveyed by the conveyor belt, reach the tail end of the conveyor belt, and fall into a collecting and storing device of the upper car cocoons.
Example 3: the cocoon sorter of this embodiment is basically the same as that of embodiment 1, except that: as shown in fig. 2, the feeding device further comprises a conveying hopper mounting support 26, and the conveying hopper 1 is fixedly arranged at the top end of the conveying hopper mounting support 26;
the feeding device further comprises a conveying hopper mounting corner fitting 3, foot hooves I27 and a foot hoof mounting plate I28, the output end of the conveying hopper 1 is fixedly arranged at the feeding end of the conveying device through the conveying hopper mounting corner fitting 3, and the foot hooves I27 are arranged at the bottom end of the conveying hopper mounting support 26 through the foot hoof mounting plate I28.
Example 4: the cocoon sorting machine of this embodiment is basically the same as that of embodiment 1 or 2, except that: as shown in fig. 3, the conveying device further comprises a conveying belt guide rail 29, a front baffle plate 2, a brush 5 and baffle plates 23, wherein the conveying belt guide rail 29 is fixedly arranged at the top end of the conveying device frame 24, the baffle plates 23 are vertically arranged on the conveying belt guide rail 29 and the baffle plates 23 are positioned on two sides of the conveying belt 6, the front baffle plate 2 is fixedly arranged at the initial end of the conveying belt guide rail 29, the brush 5 is fixedly arranged at the top end of the conveying device frame 24 and the brush 5 is positioned at the feeding end of the conveying device; the two sides of the conveyor belt 6 are arranged in the conveyor belt guide rail 29 in a sliding way;
the driven rotating shaft and the driving rotating shaft are respectively arranged at two ends of the conveying device frame 24, and the driven rotating shaft and the driving rotating shaft can rotate at two ends of the conveying device frame 24;
the front baffle plate 2 can prevent cocoons from falling off after being fed by the feeding device, the baffle plate 23 can prevent cocoons from falling off in the conveying process, the hairbrush 5 can brush cocoons on the conveying belt flat, the cocoons are ensured not to be stacked, and one layer of cocoons are ensured to be conveyed forwards each time;
the conveying device further comprises foot hooves II and foot hoof mounting plates II, foot hooves II are arranged at the bottom end of the conveying device frame 24 through the foot hoof mounting plates II, and the hairbrush 5 is fixedly arranged at the top end of the conveying device frame 24 through the hairbrush mounting plates 4.
Example 5: the cocoon sorting machine of this embodiment is substantially the same as that of embodiment 1, 2 or 3, except that: as shown in fig. 4, the image capturing device further includes an image capturing device mounting support 22 and LED light sources 33, the industrial camera 8 is fixedly disposed at the center of the top end of the image capturing device mounting support 22, the LED light sources 33 are fixedly disposed at the top end of the image capturing device mounting support 22 and the LED light sources 33 are located at both sides of the industrial camera 8;
the LED light source 33 is connected with the controller 25; the LED light source is used for providing stable illumination conditions;
the image acquisition device further comprises a camera mounting plate 7 and a camera mounting corner fitting 9, wherein the camera mounting plate 7 is horizontally arranged at the top end of the image acquisition device mounting support 22, the camera mounting corner fitting 9 is vertically arranged at the center of the top end of the camera mounting plate 7, the bottom end fixing of the industrial camera 8 is arranged on the camera mounting plate 7, the rear side surface of the industrial camera 8 is fixedly arranged on the camera mounting corner fitting 9, and the bottom end of the image acquisition device mounting support 22 is provided with a foot shoe III 31 through a foot shoe mounting plate III 32.
Example 6: the cocoon sorting machine of this embodiment is substantially the same as that of embodiment 1, 2, 3 or 4, except that: as shown in fig. 5, the cocoon grabbing device further comprises a linear module i 11, a linear module ii 16, a linear module iii 18, a support beam 13, a sucker mounting plate support 14 and a cocoon grabbing device support 21, wherein the linear module i 11 and the linear module iii 18 are horizontally and parallelly arranged at the top end of the cocoon grabbing device support 21, the linear module i 11 and the linear module iii 18 are perpendicular to the conveying direction of the conveyor belt 6, two ends of the support beam 13 are respectively and slidably arranged at the top ends of the linear module i 11 and the linear module iii 18, the linear module ii 16 is vertically arranged in the middle of the support beam 13, the sucker 10 is arranged at the bottom end of the linear module ii 16 through the sucker mounting plate support 14, the storage hopper i 20 is arranged in the middle of the cocoon grabbing device support 21 and is positioned at one side of the conveyor belt 6, and the storage hopper i 20 and the conveyor belt 6 are both positioned under the linear module i 11 and the linear module iii 18;
preferably, the linear module I11, the linear module II 16 and the linear module III 18 are all connected with the controller 25; the specification of the MDL80 aluminum-based series linear module of Shenzhen Wifar precision technology Co., ltd.) is as follows: the specification of the MSX45 aluminum-based linear module group of Shenzhen Wifar precision technology Co., ltd is that the MDL80-ML1-1600-100BH-E-G-W and the linear module group II 16 is selected as follows: the MDL80-ML1-1600-100BH-E-G-W and the linear module III 18 is selected from the MDL80 aluminum-based series linear modules of Shenzhen Wifar precision technology Co., ltd, and has the specification that: MDL80-MR1-1600-100BH-E-G-W;
foot hoofs IV 34 are arranged at the bottom end of the lower cocoon grabbing device support 21 through foot hoof mounting plates IV 35, the suction disc 10 is arranged at the bottom end of the suction disc mounting plate support 14 through the suction disc mounting plates 12, and the linear module II 16 is arranged in the middle of the supporting beam 13 through the linear module mounting angle piece 15;
the conveyer belt 6 is a flexible conveyer belt, a plurality of grooves are uniformly formed in the conveyer belt 6, the sizes of the grooves are matched with the structure of the cocoons, one cocoon is placed in each groove, and the cocoons are always positioned in the grooves in the transportation process, so that the cocoons in the transportation process can be conveniently detected and positioned; the inner wall of the sucker 10 is fixedly provided with a sponge layer, and the sponge layer can prevent the cocoons from being crushed when the cocoons are grabbed;
the lower cocoon grabbing schematic diagram is shown in fig. 7, cocoons are positioned in grooves of a conveyor belt in the conveying process, each groove in an image acquired by an image acquisition device is provided with a fixed pixel coordinate range, and when the lower cocoon grabbing device works, the position Nij of each sucking disc is matched with the groove position Mij of the corresponding pixel coordinate range; the position Mij of the groove where the cocoon is positioned can be judged according to the detected center point coordinates of the cocoon which are output by the controller, and when all cocoons in the image are conveyed to the grabbing device, the controller controls the sucker corresponding to the position of the Nij to work, and the corresponding cocoon is grabbed; when a plurality of cocoons are put down in one image, a plurality of corresponding suckers are controlled to work;
the silkworm cocoon sorting method based on the silkworm cocoon sorting machine comprises the following specific steps:
1) The cocoons are fed by a feeding mechanism and automatically fall on a conveyor belt, and are paved on the conveyor belt in a single layer for conveying;
2) When cocoons reach the lower part of the image acquisition device, the conveyor belt stops transmitting, the industrial camera acquires images of cocoons and transmits the image data to the controller, and the controller detects the center point coordinates of cocoons;
3) The controller outputs the center point coordinates of the cocoons and judges the positions of the suckers corresponding to the center point coordinates;
4) When cocoons are conveyed below the grabbing device, the conveying belt stops conveying, the linear module I and the linear module III drive the suckers to move to the position right above the conveying belt, the preset suckers are made to move to the position right above the center point of the cocoon, the linear module II drives the suckers to move towards the direction approaching to the conveying belt, and after the preset position is reached, the suckers suck the cocoons to be grabbed;
5) After the cocoons are grabbed, the linear module II drives the suckers to move in a direction away from the conveyor belt, and after the cocoons reach a preset position, the linear module I and the linear module III drive the suckers to move to the position right above the storage hopper I, namely, the suckers move to the position right above the storage hopper I, the suckers desorb, and the cocoons fall into the storage hopper I, so that the cocoons are collected and stored;
6) The non-grabbed cocoons are upper cocoons, the upper cocoons are continuously conveyed by the conveyor belt, reach the tail end of the conveyor belt, and fall into a collecting and storing device of the upper car cocoons.
Example 6: the cocoon sorting machine of this embodiment is substantially the same as that of embodiment 1, 2, 3, 4 or 5, except that: as shown in fig. 6, the cocoon loading storage device further includes a storage hopper mounting support 19, a storage hopper ii 36 is fixedly provided at the top end of the storage hopper mounting support 19, and the bottom end of the storage hopper mounting support 19 is provided with a foot shoe V through a foot shoe mounting plate V.
Example 8: the cocoon down cocoon real-time detection method of the embodiment is consistent with the cocoon down cocoon real-time detection method of the embodiment 1, wherein 550 images, 412 training sets and 138 verification sets are acquired in the cocoon down detection data set. The number of targets in different categories in the data set is shown in fig. 9, and the targets respectively comprise 399 thin cocoons, 415 flat cocoons, 387 extra-small cocoons and 430 macular cocoons, and each category comprises about 400 cocoons, so that the balance of the targets can be well ensured;
the YOLOv3 model is improved, the anchor point frame parameters suitable for the cocoon descending detection data set are selected through K-means cluster analysis experiments on the size and the number of the anchor point frames, so that the detection precision of the model is improved, and the convergence rate is increased; the effects of the anchor blocks with different numbers are shown in table 1, and as can be seen from table 1, on the premise of ensuring that the model size script is unchanged, when the number of anchor blocks is 3, the average detection speed is fastest relative to the anchor blocks with other numbers, and reaches 31.14 frames/s, while the mAP is highest, and reaches 96.10%. The method has the advantages that when the number of anchor blocks is 3, the detection effect is good for the cocoon drop detection data set. Therefore, the number of anchor blocks selected through experiments is 3, the sizes of the anchor blocks are (30, 28), (31, 35), (31, 41), and each YOLO layer is allocated with 1 anchor block to predict a target.
TABLE 1 different number anchor frame effects
Improving a YOLOv3 model, introducing a scaling factor gamma to each channel of a BN layer in the YOLOv3 for L1 regularization, and if the scaling factor gamma approaches 0 under the constraint of L1 regularization, the contribution degree of the channel to the feature map prediction in the model is lower, so that pruning can be performed; compressing a model by adopting a model pruning method based on batch regularization layer (BN) scaling factors after channel sparsification; channel pruning is shown in opinion fig. 9, the left side is a model original structure, the right side is a model structure after channel pruning, each parallelogram represents a convolution channel, and the weight of a light-colored part after sparsification training is smaller, so that the channel corresponding to the light-colored part needs to be deleted, and a model with fewer parameters is obtained, and the implementation flow of the channel pruning is as follows:
1) Training a YOLOv3 model by using a cocoon detection data set to obtain a basic model;
2) Sparse training is carried out on the basic model, so that a scaling factor gamma corresponding to a channel with low contribution degree to cocoon falling prediction after the characteristic map passes through the BN layer is as close as possible to 0;
3) Determining pruning proportion, and deleting a channel corresponding to a scaling factor lower than a pruning proportion setting threshold value;
4) Fine-tuning the precision of the pruned model, and reducing the precision loss;
5) Generating a new cocoon detecting model file;
performing model pruning based on BN layer scaling factor gamma on the sparsified model, wherein the model effects of different pruning rates are shown in Table 2; as can be seen from table 2, the model size can be continuously reduced by pruning with different pruning rates, and the mAP of the model is reduced as the pruning rate increases, but the overall reduced value is very small, negligible, and the average detection speed is always increased; finally, 80% of the channels of the convolution layer in the model can be subtracted, the size of the model is reduced to 46.90M, the average detection speed is increased from 31.14 frames/s before pruning to 51.05 frames/s after pruning, and mAP is only reduced by 0.04%;
TABLE 2 model effects of different pruning Rate
Since the improved YOLOv3 model is a cocoon based on the detection object, it belongs to a small object at the time of detection, and as the model deepens, the position information of the feature map is partially lost. Embedding a receptive field module (RFB) module into YOLOv3 to increase local receptive field of the model, so that the output characteristic diagram has more abundant context information;
as shown in FIG. 11, in order to ensure that the RFB modules can be obtained by all 3 YOLO layer branches of the model and the effect of enhancing the receptive field is brought, the RFB modules are embedded in the 3 YOLO layer branches to achieve the best initial effect, and the newly added RFB modules are arranged in a dotted line frame; table 3 shows the model effect after RFB embedding.
TABLE 3 model Effect after RFB embedding
As can be seen from table 3, on the premise of ensuring that the size of the model is 46.90M, after RFB modules are added at different positions of the model, the average detection speed of the model is about 50 frames/s, and is reduced less than that of the model after pruning at the pruning rate of 80.00%, wherein after 36 layers of the model, namely, the RFB modules are added at the beginning of 3 YOLO layer branches, the maximum mAP is improved by approximately 1.00% and reaches 96.80% compared with other embedding modes. The model with RFB module added after 36 layers was chosen as the best detection model.
The specific embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the spirit of the present invention.
Claims (8)
1. The silkworm cocoon sorting machine is characterized in that the silkworm cocoon sorting machine performs grabbing and rejecting on lower cocoons according to coordinates output by a real-time detection method of the lower cocoons, and collects and stores upper cocoons;
comprising the following steps:
the feeding device, the conveying device, the image acquisition device, the cocoon picking device and the cocoon feeding storage device are connected in sequence,
the feeding device is used for feeding cocoons;
the conveying device is used for conveying cocoons;
the image acquisition device is used for photographing cocoons on the conveying device to obtain cocoon images;
the cocoon picking device is used for picking and removing cocoons;
the cocoon picking device is used for storing cocoons;
the upper cocoon storage device is used for storing upper cocoons;
the feeding device comprises a conveying hopper (1), and the conveying hopper (1) is arranged at the feeding end of the conveying device; the conveying device comprises a conveying belt (6), a conveying device frame (24) and a motor (30), wherein the conveying belt (6) is arranged at the top end of the conveying device frame (24), the motor (30) is arranged at the side end of the top of the conveying device frame (24), and an output shaft of the motor (30) is in transmission connection with the conveying belt (6); the image acquisition device comprises an industrial camera (8), and the industrial camera (8) is arranged right above the middle part of the conveying device; the cocoon picking device comprises a plurality of suckers (10), and the suckers (10) are arranged above the tail end of the conveying device; the cocoon grabbing device comprises a storage hopper I (20), the storage hopper I (20) is arranged at the side end of the conveyor belt (6), and when cocoons are stored, the storage hopper I
(20) Is positioned right below the sucker (10); the upper cocoon storage device is arranged right below the tail end of the conveyor belt (6);
the cocoon grabbing device further comprises a linear module I (11), a linear module II (16) and a linear module III
(18) The cocoon grabbing device comprises a supporting beam (13), a sucking disc mounting plate supporting device (14) and a cocoon grabbing device supporting device (21), wherein a linear module I (11) and a linear module III (18) are horizontally and parallelly arranged at the top end of the cocoon grabbing device supporting device (21), the linear module I (11) and the linear module III (18) are perpendicular to the conveying direction of a conveying belt (6), two ends of the supporting beam (13) are respectively and slidably arranged at the top ends of the linear module I (11) and the linear module III (18), a linear module II (16) is vertically arranged in the middle of the supporting beam (13), a sucking disc (10) is arranged at the bottom end of the sucking disc mounting plate (12), and the sucking disc mounting plate (12) is arranged at the linear module II through a connecting beam (17)
(16) The storage hopper I (20) is arranged in the middle of the lower cocoon grabbing device support (21), the storage hopper I (20) is positioned on one side of the conveyor belt (6), and the storage hopper I (20) and the conveyor belt (6) are positioned right below the linear module I (11) and the linear module III (18);
the method for detecting cocoons under cocoons in real time comprises the following specific steps:
1) Taking thin cocoon, flat cocoon, extra small cocoon and macular cocoon as detection targets, shooting the detection targets at different positions and different angles, and collecting a cocoon discharging detection data set;
2) Transforming the cocoon detection data set into 416×416 images by using an interpolation algorithm, and dividing the images into a training set and a test set; the image quantity ratio of the training set to the test set is 3:1;
3) Labeling images in the training set and the test set by using a label I mg labeling tool; marking a boundary box of a corresponding cocoon type for all the identified objects on each image;
4) Training a model by using an image marked in a training set by taking a YOLOv3 model as a basic model, performing super-parameter searching and evolution to obtain a new generation of evolution, judging whether to evolve or not through a fitness function, wherein the higher the fitness is, the better the performance of the generation is represented, so that the super-parameter suitable for cocoon detection data set is selected through evolution searching;
5) Setting evolution super parameters of the improved YOLOv3 model, and loading the pre-training weight into the improved YOLOv3 model for training to obtain an optimal training weight;
6) Loading the optimal training weight into the improved YOLOv3 model to test the images in the test set;
7) Repeating the steps 5) and 6) to obtain an optimal YOLOv3 model as a cocoon setting real-time detection method.
2. The cocoon sorting machine of claim 1, wherein: step 5) the improved YOLOv3 model is:
a. selecting anchor frame parameters suitable for the cocoon detection data set by carrying out K-means cluster analysis on the size and the number of anchor frames of the YOLOv3 model;
b. compressing a YOLOv3 model by adopting a model pruning method based on batch regularization layer (BN) scaling factors after channel sparsification;
c. the receptive field module RFB was embedded in the YOLOv3 model to increase the local receptive field of the model.
3. The cocoon sorting machine of claim 2, wherein: a specific method for compressing the YOLOv3 model by adopting a model pruning method based on batch regularization layer (BN) scaling factors after channel sparsification is as follows
And introducing a scaling factor gamma to each channel of the BN layer in the YOLOv3 model to carry out L1 regularization, and if the scaling factor gamma approaches 0 under the constraint of L1 regularization, carrying out pruning by the channel with low contribution degree to feature map prediction in the model.
4. The cocoon sorting machine of claim 1, wherein: the feeding device further comprises a conveying hopper mounting support (26), a conveying hopper mounting corner fitting (3), foot hooves I (27) and a foot hoof mounting plate I (28), wherein the output end of the conveying hopper (1) is fixedly arranged at the feeding end of the conveying device through the conveying hopper mounting corner fitting (3), and the foot hooves I (27) are arranged at the bottom end of the conveying hopper mounting support (26) through the foot hoof mounting plate I (28).
5. The cocoon sorting machine of claim 1, wherein: the conveying device further comprises a conveying belt guide rail (29), a front baffle plate (2), a hairbrush (5) and baffle plates (23), wherein the conveying belt guide rail (29) is fixedly arranged at the top end of the conveying device frame (24), the baffle plates (23) are vertically arranged on the conveying belt guide rail (29) and the baffle plates (23) are positioned on two sides of the conveying belt (6), the front baffle plate (2) is fixedly arranged at the initial end of the conveying belt guide rail (29), the hairbrush (5) is fixedly arranged at the top end of the conveying device frame (24) and the hairbrush (5) is positioned at the feeding end of the conveying device;
the conveying device further comprises foot hooves II and foot hoof mounting plates II, foot hooves II are arranged at the bottom end of the conveying device frame (24) through the foot hoof mounting plates II, and the hairbrush (5) is fixedly arranged at the top end of the conveying device frame (24) through the hairbrush mounting plates (4).
6. The cocoon sorting machine of claim 1, wherein: the image acquisition device further comprises an image acquisition device mounting support (22) and LED light sources (33), the industrial camera (8) is fixedly arranged at the center of the top end of the image acquisition device mounting support (22), the LED light sources (33) are fixedly arranged at the top end of the image acquisition device mounting support (22), and the LED light sources (33) are positioned at two sides of the industrial camera (8);
the image acquisition device further comprises a camera mounting plate (7) and a camera mounting corner fitting (9), wherein the camera mounting plate (7) is horizontally arranged at the top end of the image acquisition device mounting support (22), the camera mounting corner fitting (9) is vertically arranged at the center of the top end of the camera mounting plate (7), the bottom end of the industrial camera (8) is fixedly arranged on the camera mounting plate (7) and the rear side surface of the industrial camera (8) is fixedly arranged on the camera mounting corner fitting (9), and the bottom end of the image acquisition device mounting support (22) is provided with a foot shoe III (31) through a foot shoe mounting plate III (32).
7. The cocoon sorting machine of claim 1, wherein: the bottom end of the cocoon grabbing device support (21) is provided with a foot shoe IV (34) through a foot shoe mounting plate IV (35), and a linear module II (16) is arranged in the middle of the supporting beam (13) through a linear module mounting corner fitting (15).
8. The cocoon sorting machine of claim 1, wherein: the cocoon loading storage device further comprises a storage hopper mounting support (19), a storage hopper II (36) is fixedly arranged at the top end of the storage hopper mounting support (19), and foot hooves V are arranged at the bottom end of the storage hopper mounting support (19) through foot hoof mounting plates V.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110685414.2A CN113269275B (en) | 2021-06-21 | 2021-06-21 | Real-time detection method for cocoons under cocoons |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110685414.2A CN113269275B (en) | 2021-06-21 | 2021-06-21 | Real-time detection method for cocoons under cocoons |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113269275A CN113269275A (en) | 2021-08-17 |
CN113269275B true CN113269275B (en) | 2023-06-06 |
Family
ID=77235628
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110685414.2A Active CN113269275B (en) | 2021-06-21 | 2021-06-21 | Real-time detection method for cocoons under cocoons |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113269275B (en) |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102095481A (en) * | 2010-12-02 | 2011-06-15 | 江苏大学 | Device and method for automatically discriminating cocoon sexuality |
CN109035244A (en) * | 2018-08-14 | 2018-12-18 | 张家港江苏科技大学产业技术研究院 | Automatic cocoon-picking system, automatic cocoon-picking machine and pluck cocoon method |
CN110193476A (en) * | 2019-07-05 | 2019-09-03 | 中国计量大学 | A kind of silk cocoon simple grain sorting unit |
CN110252688A (en) * | 2019-04-18 | 2019-09-20 | 中国计量大学 | A kind of Cocoon Quality identification apparatus and identification method |
CN112446421A (en) * | 2020-11-06 | 2021-03-05 | 湖州市质量技术监督检测研究院 | Silkworm cocoon counting and identifying method based on machine vision |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108522437B (en) * | 2017-05-08 | 2020-10-02 | 杭州蚕鱼科技有限公司 | Full-automatic male and female silkworm cocoon sorting unit |
CN210045602U (en) * | 2019-06-01 | 2020-02-11 | 阳山县兴达蚕业有限公司 | Male and female cocoon sorting machine |
CN110125021B (en) * | 2019-06-25 | 2021-09-24 | 北京工业大学 | Automatic silkworm cocoon sorting device |
CN112395952A (en) * | 2020-10-23 | 2021-02-23 | 西安科锐盛创新科技有限公司 | A unmanned aerial vehicle for rail defect detection |
CN112487899B (en) * | 2020-11-19 | 2023-04-07 | 武汉高德飞行器科技有限公司 | Target identification method and system based on unmanned aerial vehicle, storage medium and electronic equipment |
-
2021
- 2021-06-21 CN CN202110685414.2A patent/CN113269275B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102095481A (en) * | 2010-12-02 | 2011-06-15 | 江苏大学 | Device and method for automatically discriminating cocoon sexuality |
CN109035244A (en) * | 2018-08-14 | 2018-12-18 | 张家港江苏科技大学产业技术研究院 | Automatic cocoon-picking system, automatic cocoon-picking machine and pluck cocoon method |
CN110252688A (en) * | 2019-04-18 | 2019-09-20 | 中国计量大学 | A kind of Cocoon Quality identification apparatus and identification method |
CN110193476A (en) * | 2019-07-05 | 2019-09-03 | 中国计量大学 | A kind of silk cocoon simple grain sorting unit |
CN112446421A (en) * | 2020-11-06 | 2021-03-05 | 湖州市质量技术监督检测研究院 | Silkworm cocoon counting and identifying method based on machine vision |
Non-Patent Citations (3)
Title |
---|
Rui Yang 等.SK-FMYOLOV3: A Novel Detection Method for Urine Test Strips.《Wireless Communications and Mobile Computing》.2020,1-14. * |
刘莫尘 等.基于FCM及HSV模型的方格蔟黄斑茧检测与剔除技术.《农业机械学报》.2018,第49卷(第7期),31-38. * |
贝建设.蚕桑生产机械化应用发展概况与思路.《现代农业科技》.2017,(第20期),155-157. * |
Also Published As
Publication number | Publication date |
---|---|
CN113269275A (en) | 2021-08-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110516570B (en) | Vision-based garbage classification and identification method and device | |
CN104822230B (en) | A kind of multi-functional pcb board production automation system | |
CN102217559B (en) | On-line automatic detection and sorting equipment and method for incubated egg quality | |
CN107185858B (en) | Fruit detection and classification system and method | |
CN113182203A (en) | Intelligent silkworm cocoon sorting method | |
CN110194369B (en) | Automatic egg detecting intelligent device | |
CN108820945A (en) | A kind of film automatic charging device | |
CN201618694U (en) | Fruit grading and detecting device | |
CN202133655U (en) | Online automatic detecting sorting equipment for quality of incubated eggs | |
CN109317420A (en) | Convey the control device of sorting automatically for courier packages | |
CN112044790A (en) | Defective products detects marking machine | |
CN105783990A (en) | Panoramic image measurement system for corn ear indoor accurate seed assessment and testing | |
CN113269275B (en) | Real-time detection method for cocoons under cocoons | |
CN103879788A (en) | Material blocking and separating mechanism | |
CN215656492U (en) | Silkworm cocoon sorting machine | |
CN212306526U (en) | Live fish sorting machine based on AI visual identification | |
CN113714149A (en) | Parallel arm intelligent silkworm cocoon sorting robot and silkworm cocoon sorting method | |
CN219460229U (en) | Full-automatic sorting system for quality grading and defective product removal of large yellow croaker | |
CN219817122U (en) | Series arm silkworm cocoon classifying and sorting machine | |
CN217250743U (en) | Feather detects letter sorting equipment | |
CN117007519A (en) | Appearance detection equipment of semiconductor chip frame | |
CN106238340B (en) | A kind of onion intelligence sorting unit | |
CN216836018U (en) | Planting cup recovery system | |
CN115043009A (en) | Intelligent visual sorting and packaging equipment for eggs, control system and implementation method | |
CN210936013U (en) | Carousel formula veneer grading plant |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |