CN103801520A - Method and device for automatically carefully sorting and grading shrimps - Google Patents

Method and device for automatically carefully sorting and grading shrimps Download PDF

Info

Publication number
CN103801520A
CN103801520A CN201410040879.2A CN201410040879A CN103801520A CN 103801520 A CN103801520 A CN 103801520A CN 201410040879 A CN201410040879 A CN 201410040879A CN 103801520 A CN103801520 A CN 103801520A
Authority
CN
China
Prior art keywords
shrimp
image
raw material
channel
moment
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.)
Granted
Application number
CN201410040879.2A
Other languages
Chinese (zh)
Other versions
CN103801520B (en
Inventor
成芳
刘子豪
龚朝勇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN201410040879.2A priority Critical patent/CN103801520B/en
Publication of CN103801520A publication Critical patent/CN103801520A/en
Application granted granted Critical
Publication of CN103801520B publication Critical patent/CN103801520B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Landscapes

  • Image Processing (AREA)

Abstract

The invention discloses a device for automatically carefully sorting and grading shrimps. The device comprises a feeding system, a sorting channel, an image collecting system, a grading system and an image processing system; the feeding system is used for outputting to-be-sorted shrimps to the sorting channel in a single row; the sorting channel consists of a buffering channel and a sampling channel which are sequentially connected, the sampling channel is divided into a plurality of single-row channels, the buffering channel is internally provided with a plurality of direction adjusters, and used for separating raw material shrimps, output by the feeding system, into singles, and adjusting the posture of each raw material shrimp entering single-row channel; the image collecting system is used for collecting images of the raw material shrimps; the grading system comprises spray nozzles corresponding to single-row channels, and the spray nozzles are controlled by the image processing system and used for blowing the shrimps to enter different receiving tanks; the image processing system is used for analyzing the images, grading the image of each shrimp, and emitting the signal for controlling the grading system according to the graded results. The invention further discloses a method for automatically carefully sorting and grading shrimps.

Description

Automatic well-chosen stage division and the device of shrimp
Technical field
The present invention relates to aquatic products field of engineering technology, relate in particular to automatic well-chosen stage division and the device of a seed shrimp.
Background technology
China, as one of the whole world six large shrimps producing country, all will carry to international market a large amount of raw material shrimps every year.But China's raw material shrimp outlet whole price level is low, market share is few, competitiveness is not enough is the reason that causes China's raw material shrimp export volume to reduce.Along with the industrial structure deepens constantly adjustment, how to support numerous and confused rise the in shrimp field and enclose pool cultivation, large-scale cultivation base makes workman be busy with cultivation and the breeding to antenatal shrimp species, but this link that needs most a large amount of human and material resources of input, financial resources of good harvest of raw material shrimp is but often out in the cold for postpartum, cause a lot of raw material shrimps because postpartum degree of attention cause not the mortality of raw material shrimp, caused serious economic loss to enterprise and raw material shrimp aquaculture industry.
For a long time, from the raw material shrimp of just fishing for up, the mode of selected classification raw material shrimp is all to complete by the hand picked mode of workman, not only inefficiency of this mode, and also people is tired in the situation that, easily produces erroneous judgement; Relatively deficient and expensive today labour, use the mode of a large amount of labours selected normal shrimp from raw material shrimp out-of-date already, the mode that our amount in need of immediate treatment is large, equipment that floor space is little, simple in structure, cost is low, automaticity is high replaces artificial work, so not only can improve accuracy rate and the efficiency of the selected classification of raw material shrimp, but also can reduce labour's expense, improve the efficiency of having a good harvest postpartum.
Situation below the general raw material shrimp of fishing for from pond exists conventionally: conventionally can sneak into the raw material shrimp that (1) salvages because of anoxic and downright bad raw material shrimp, if remove not in time, easily normal shrimp is polluted, cause the latter rotten, be unfavorable for post-production and storage, reduced the economic worth of raw material shrimp; (2) many salvagings disembarkations and the shrimp body that floats over splendid attire raw material shrimp bucket surface are easy to be baked by hot sunlight and dead, pollute thereby can produce other normal shrimps normally during the broiling summer the season for a bumper harvest of general raw material shrimp; (3) fish in the raw material shrimp of disembarkation and be also mixed with the small-sized weeds that some are not easy with the naked eye to be found by people, easily sneak in the finished product being processed into.
Through retrieval, document " Automatic grading and packing of prawns " has been developed a kind of raw material shrimp packing and grading plant based on computer vision.This system is obtained the general image of shrimp by camera, then is calculated the parameters such as the shape, the deviation angle, abdomen shell flexibility of shrimp by image information, and mark shrimp head, shrimp tail and shrimp centre position.Patent publication No. be JP3603353B2 Japanese Patent Publication a kind of bottle based on machine vision technique detect grading plant, utilize machine vision technique to detect the outward appearance such as color, shape of plurality of classes bottle, can carry out online real-time grading to bottle.Although this patent has used machine vision method to solve the recovery classification problem of waste and old bottle, this classification method of discrimination is comparatively simple, and accuracy is not high; The Chinese patent that number of patent application is 201010165377.4 provides the impurity elimination sorting unit of a kind of anchovy peculiar to vessel, shrimp processing, comprise multistage screening dish, feed hopper, shaking device and frame, it can selected species only limits to two kinds of anchovy and shrimps, and selected scope awaits further raising; Number of patent application is that 201220713538.3 Chinese patent has used a kind of method based on machine vision to detect the prawn of the different quality ranks of classification, and simple grain induction system is wherein made up of prawn feed hopper, two vibration feed appliances, two sensors, the single-row draw-in groove of individual layer, guide groove and attitude adjusters.In this patent, the shell of prawn is very easily pulverized, and adopts the mode of vibration feeding, and easily raw material shrimp body itself produces microlesion; Transport tape adopts on horizontal transmission mode and its only has a row guide groove can allow material to pass through, and treating capacity is little and speed is slower; Patent publication No. is the classification machine that the BP of GB2160653A has been invented a selected raw material shrimp, probe capable of circulation by one, that can lower, be fixed on rotatable dish directly contacts raw material shrimp body, in the time that probe has been felt abnormal shrimp, damage shrimp, dystopy shrimp, can give executing agency signal, thereby impurity is rejected in starting nozzle air blowing.This method can produce it and pollute because of contact shrimp body; The height of rotating disc need regulate at any time according to the thickness of shrimp body, and automaticity is not high.
Summary of the invention
In order to overcome defect of the prior art, the present invention the method and apparatus that a set of raw material shrimp treating capacity is large, processing speed is fast is provided, integrated application artificial intelligence, electronic technology, Principle of Communication, Machine Design, physics of photography and computer hardware technique raw material shrimp is carried out to the selected classification of external appearance characteristic, result is accurately and reliably.
Automatic well-chosen grading plant of the present invention, in the raw material shrimp season for a bumper harvest, for fishing for the problem that is mixed with under weeds, small fish, anoxic in the raw material shrimp of disembarkation, bakes the impurity such as shrimp, also in order to meet the needs of the selected classification of product in product processing, specialized designs is developed a kind of automatic well-chosen stage division and device being suitable for raw material shrimp aquaculture field and the processing factory seed shrimp based on computer vision technique that use, that floor space is little, intensive degree is high.
The automatic well-chosen grading plant of one seed shrimp, comprises feeding system, sort channel, adopts drawing system, hierarchy system and image processing system;
Feeding system is used for single the shrimp body of the treating sorting sort channel that exports to;
Sort channel is divided into the buffer channel and the sampling channel that are connected successively, sampling channel is separated into multiple single-rowization passages, is provided with multiple directions adjuster for the shrimp of feeding system output is divided into single and regulates each shrimp body to enter the attitude of corresponding single-rowization passage in described buffer channel;
Adopt drawing system for gathering the image of shrimp body in each single-rowization passage, adopt drawing system and comprise and be positioned at two lighting box that sampling channel is upper and lower and dislocation is arranged, in each lighting box, be equipped with camera;
Hierarchy system comprises the air nozzle corresponding with each single-rowization passage, and this air nozzle is controlled by described image processing system, enters the different grooves that connects material for blowing by the shrimp body of single-rowization passage output;
Image processing system, for analyzing described image, to the image grading processing of each shrimp body, according to classification results, and sends the signal of controlling hierarchy system.
Described feeding system comprises support, rack-mount feeding box and the conveyer belt being in tilted layout, the oblique upper of this conveyer belt is provided with backgauge soft board, conveyer belt bottom is stretched in feeding box, and conveyer belt surface is provided with the draw-in groove that holds single shrimp body, one side of feeding box is provided with osculum, and conveyer belt is made up of multiple plastic chain plates with leaking hole.
In feeding box, be placed with the raw material shrimp of pending selected classification, the osculum of its bottom is for discharging the sewage that enters feeding box with shrimp, raw material shrimp is by the conveyer belt output feeding box of persistent movement, conveyer belt is provided with draw-in groove, this draw-in groove is arranged along the width of conveyer belt, row's raw material shrimp only can be allowed in its inside, may on draw-in groove, form accumulation for shrimp body, the raw material shrimp that backgauge soft board is now piled up in can floating draw-in groove, stop that stacking raw material shrimp passes through simultaneously, after the effect of backgauge soft board, the raw material shrimp in draw-in groove is single layer arranging and distributes.
Feeding system also comprises the conveyer frames that conveyer belt is installed, and conveyer frames middle part is coupling on support, and one end of conveyer frames is supported on described support by height regulating rod.
In the present invention, the angle of inclination of conveyer belt can regulate by height regulating rod, to adapt to the needs of different sizes or raw material shrimp species class, one end of height regulating rod is hinged on support side, the other end is bolted on support, by the cooperation position of mobile bolt and height regulating rod, carrys out the height regulating rod length between adjusting bolt and conveyer frames, realize the adjustment of conveyer frames setting angle, thereby drive the angle of inclination that regulates conveyer belt.
Selecting of conveyer belt, larger on the impact of whole selected grading plant, traditional conveyer belt is quality of rubber materials, draw-in groove is straight forming in conveyer belt process; But in the present invention, the conveyer belt preferably using is made up of multiple plastic chain plates with leaking hole, between adjacent two plastic chain plates, connect by rotating shaft, every plastic chain plate is provided with a spacing lug, between adjacent two spacing lugs, be described draw-in groove, raw material shrimp is single to be distributed in draw-in groove, simultaneously raw material shrimp upper with moisture, can see through leaking hole, reduce the moisture that enters sort channel.
Sort channel is divided into the buffer channel and the sampling channel that are connected successively, and two sections of passages are all in tilted layout, and the angle of inclination of sampling channel is greater than the angle of inclination of buffer channel.Buffer channel has cushioning effect, prevents that the raw material shrimp of conveyer belt output from directly entering in sampling channel, to slow down the sliding speed of raw material shrimp, is convenient to adopt the IMAQ of drawing system; Simultaneously, direction adjuster in buffer channel has compartmentation, make each raw material shrimp enter a single-rowization passage, the attitude that can also utilize direction adjuster to regulate raw material shrimp to slide in single-rowization passage, keep gathering the cardinal principle posture of image Raw shrimp, be convenient to comparison and the analysis of image processing system.
Wherein, direction adjuster comprises the locating piece being arranged in buffer channel, and locating piece is fixed with guide plate towards one end of feeding system, and this guide plate has the guiding surface that acts on raw material shrimp.Buffer channel is partitioned into monomer passage by locating piece, each monomer passage is corresponding with single-rowization passage, guide plate is arranged on the end of locating piece, this guide plate comprises two monomer guide plates of integrative-structure, two monomer guide plates are in a certain angle, every monomer guide plate has a guiding surface, guide plate is for separating the raw material shrimp of conveyer belt output, raw material shrimp monomer is entered respectively in different single-rowization passages, between the guiding surface of raw material shrimp by monomer passage both sides, pass, guiding surface has the effect that regulates raw material shrimp attitude.
Between adjacent two single-rowization passages, be fixed with spacing block; the spacing block end that is positioned at single-rowization passage feed end is wedge shape; the wedge shaped tip of spacing block can be for regulating the cruising attitude of raw material shrimp in single-rowization passage; the width of each single-rowization passage and raw material shrimp size adapt, and further impel its simple grain.
Adopt drawing system comprise be fixed on below sampling channel top and passage and inwall scribble barium sulfate lighting box, be arranged in the light source in lighting box and be positioned at the camera at lighting box top, and be also provided with the sensor being connected with image processing system in each single-rowization passage, for responding to the shrimp body that single-rowization passage passes through, and send the signal of controlling described camera and carry out image taking.
Simultaneously, the input of each single-rowization passage and output are fixed with respectively first sensor and the second sensor, two sensors all access described image processing system, first sensor carries out the signal of image taking for sending control camera, the second sensor sends the signal of opening air nozzle; First sensor is for responding to the raw material shrimp that enters single-rowization passage, first sensor is after raw material shrimp touches, send the signal that control camera carries out image taking, the second sensor is for responding to the raw material shrimp that skids off single-rowization passage, after the second sensor is touched, according to the classification results of image processing module, controls the magnetic valve corresponding with substandard products and start, the shake-up unblanking magnetic valve that magnetic valve sends according to the second sensor, controls air nozzle and blows raw material shrimp and enter the substandard products groove that connects material.
If the lighting box of is only installed separately, only can gather shrimp body direct picture analyzes, for back side tool shrimp body defective there is blind area in analysis, therefore, for identification more accurate, reduce error, the base plate that the present invention arranges sampling channel is transparence, and at sampling channel lighting box is set respectively up and down, and upper and lower two lighting box are staggeredly arranged, and prevent from disturbing each other, and according to the sliding speed of shrimp body, between camera in two lighting box, exist certain shooting to postpone, make the position of shrimp body substantially in image medium position.Camera is all controlled by the signal that sensor sends, and makes two cameras can successively gather the image of raw material shrimp front and back, and complete this raw material shrimp image of surface analysis also provides grade discrimination, makes the selected classification of shrimp more accurate.
In the present invention, also be provided with the bracing frame that sort channel is installed, be positioned on the bracing frame of sort channel output the first baffle plate and second baffle are installed, the first baffle effect is in the raw material shrimp freely being glided by each single-rowization passage, second baffle is arc and acts on the raw material shrimp of being blown by air nozzle, and the first baffle plate is equipped with supple buffer layer with second baffle with the face that raw material shrimp contacts.
After analyzing by image processing system, the raw material shrimp that sort channel is skidded off is divided into raw material and impurity or certified products and two class of substandard products, wherein certified products skids off rear free-falling along passage, after stopping by the first baffle plate, falling into certified products connects material in groove, impurity skids off after track under the blowing of air nozzle, fall on second baffle and fall into substandard products along second baffle and connect material in groove, realize the selected classification of raw material shrimp quality; On two baffle plates, be equipped with supple buffer layer, this cushion can be foam-rubber cushion or silicagel pad, to reduce the collsion damage of raw material shrimp.
The present invention also provides the automatic well-chosen stage division of a seed shrimp, comprises the following steps:
1) feeding system is by the single shrimp body sort channel that exports to, and each shrimp body slides at the single-rowization passage of sort channel with the attitude adapting to;
2) utilize the image of collected by camera shrimp body, and utilize image processing system to carry out pretreatment to the image collecting, obtain the area-of-interest in raw material shrimp image, the gray level co-occurrence matrixes information in raw material shrimp of extracting, as textural characteristics, uses decision tree classifier to classify to the textural characteristics extracting;
3) hierarchy system is according to described classification results, and control air nozzle blows the raw material shrimp being skidded off by single-rowization passage and enters the different grooves that connects material.
In step 2) in, in image pretreatment, use 9 × 9 windows to carry out mean filter processing to raw material shrimp image, each point average gray in 9 × 9 neighborhoods around each pixel of this image is replaced to the original each point gray value of this image; Region of interesting extraction process is as follows:
If the visual field actual (tube) length of camera is x, wide is y, and raw material shrimp actual (tube) length that image occupies is x 1, wide is y 1, according to formula:
x 640 = x 1 l , y 480 = y 1 w
Can calculate the size of area-of-interest, wherein, l represents the length of area-of-interest, and w represents the wide of area-of-interest.
In step 2) in, use gray scale symbiosis battle array to extract that entropy, angle second moment, contrast divide and the feature of the moment of inertia to area-of-interest, again according to 15 °, 30 °, 45 °, 60 °, 90 °, 120 °, 150 °, 180 ° eight direction calculating mean entropies and, average angle second moment and, average contrast divide and, average the moment of inertia and, specific formula for calculation is:
Mean entropy and:
Figure BDA0000463051510000062
Average angle second moment and:
Figure BDA0000463051510000063
Average contrast divide and:
Figure BDA0000463051510000064
Average the moment of inertia and:
Figure BDA0000463051510000071
Wherein ENT, ASM, IDM, GXJ represent that respectively entropy, angle second moment, contrast divide and the moment of inertia, by above-mentioned formula calculate mean entropy and, average angle second moment and, average contrast divide and, average the moment of inertia and, be described textural characteristics.
According to textural characteristics obtained above, use decision tree classifier to carry out Decision Classfication to the textural characteristics that best embodies original image, travel through whole decision tree, in the time that leaf node is " just ", corresponding is certified products; In the time that leaf node is " bearing ", corresponding is substandard products.
Compared with prior art, the beneficial effect that the present invention has is:
This device has used the mode of single material loading to replace traditional vibrations feeding style, is difficult for prawn body and causes damage; Utilize the difference of the speed on speed and the buffer board of shrimp in sort channel can make raw material shrimp form simple grain effect; Downslide slideway adopts two lighting box designs can obtain the full superficial makings information of shrimp, and more comprehensively, accuracy of identification is higher for feature.The design of multi-channel mode, multipair sensor and multiple nozzles makes each passage can independently carry out operation separately, and treating capacity increases and speed is accelerated; The corresponding image processing algorithm of capable of regulating adapts to the variation of detected object; Whole system has overcome traditional mechanical type classifying equipoment and has detected the few shortcoming of index, and accuracy of detection is higher simultaneously, and human-computer interaction interface close friend is applicable to workman's operation, and intelligent degree is high.
Accompanying drawing explanation
Fig. 1 is system architecture schematic diagram of the present invention;
Fig. 2 is side view in Fig. 1;
Fig. 3 is the structural representation of feeding system in Fig. 1;
Fig. 4 is the local structural graph of backgauge soft board in Fig. 3;
Fig. 5 is the structural representation of sort channel;
Fig. 6 is the structural representation of two baffle plates;
Fig. 7 is PC-AVR host-guest architecture figure;
Fig. 8 is the workflow diagram of system;
Fig. 9 is image capture module flow chart;
Figure 10 is decision tree classification flow process.
The specific embodiment
Using raw material shrimp as selected classification object, in conjunction with Figure of description, a point automatic well-chosen grading plant for invention is described in further detail below.
As depicted in figs. 1 and 2, in the present embodiment, the automatic well-chosen grading plant of a seed shrimp, comprises feeding system, sort channel, adopts drawing system, hierarchy system and image processing system.
Feeding system is used for single the raw material shrimp of the treating sorting sort channel that exports to, and specifically as shown in Figure 3, it comprises support 1 to the structure of feeding system, is arranged on the feeding box 2 on support 1 and the conveyer belt 4 being in tilted layout, and these conveyer belt 4 bottoms are stretched in feeding box 2.
The interior placement of feeding box 2 needs the raw material shrimp of sorting, and side is provided with osculum 3, for discharging the waste water that enters feeding box with shrimp; Conveyer belt 4 is for constantly exporting the raw material shrimp in feeding box 2, it is arranged on corresponding conveyer frames 5, the both sides of support 1 are fixed with installing plate 7, conveyer frames 5 both sides are provided with the rotating shaft being rotatably assorted with installing plate 7, the both sides of conveyer frames 5 are also hinged with height regulating rod 6, this height regulating rod 6 is bolted on support 1, and the inclination of the adjustable conveyer frames 5 of height regulating rod 6, realizes the adjustment at conveyer belt 4 angles of inclination.
The surface of conveyer belt 4 is provided with the draw-in groove that holds raw material shrimp, and this draw-in groove is along the width layout of conveyer belt 4, and the degree of depth of draw-in groove is no more than the average thickness of raw material shrimp, and the width of shrimp body and draw-in groove adapts, and the raw material shrimp in each draw-in groove is single layer arranging distribution.As shown in Figure 4, the top of conveyer belt 4 is furnished with backgauge soft board 23, this backgauge soft board 23 is bolted on support bar 25, the two ends of support bar 25 are fixed on the base 24 of conveyer frames 5 both sides, and the end face of base 24 is provided with U-lag, and support bar 25 ends are bolted in U-lag, the end of backgauge soft board 23, is along the surface of pressing close to conveyer belt 4, its effect is by floating the raw material shrimp that overlaps on conveyer belt 4, prevents that raw material shrimp from piling up on conveyer belt 4, is convenient to single-rowization of follow-up raw material shrimp.
Sort channel is engaged on the output at conveyer belt 4 tops, for accepting the raw material shrimp of being exported by conveyer belt 4, its concrete structure as shown in Figure 5, comprise the buffer channel 8 and the sampling channel 12 that are connected successively, two sections of passages are all in tilted layout, and the angle of inclination of sampling channel 12 is greater than the angle of inclination of buffer channel 8, the angle of inclination is here the angle between horizontal plane.
The raw material shrimp that buffer channel 8 falls for receiving conveyer belt 4, slows down the speed entering in sampling channel, and on buffer channel 8, both direction adjuster is also installed, and for the attitude adjustment of raw material shrimp, and the raw material shrimp of conveyer belt output is carried out to monomer separately; Direction adjuster comprises the locating piece 9 being arranged in buffer channel 8, and locating piece 9 is fixed with guide plate 10 towards one end of feeding system, and this guide plate 10 has the guiding surface that acts on raw material shrimp; Locating piece 9, for installing and fixed guide plate 10, and is divided into multiple monomer passages by buffer channel 8; Guide plate 10 comprises the monomer guide plate of two integrative-structures, and two monomer guide plate groups are at a certain angle, the tip of angle is towards conveyer belt 4, there is the separately effect of shrimp body of guiding, between single raw material shrimp monomer guide plate by each monomer passage both sides, pass through, the monomer guide plate capable of regulating of monomer passage both sides, by the attitude of raw material shrimp, makes shrimp body enter in sampling channel 12 to be fixed on attitude.
On the base plate of sampling channel 12, be fixed with two spacing blocks 11, sampling channel 12 is separated into three single-rowization passages, each single-rowization passage receives respectively the raw material shrimp that corresponding monomer passage skids off, and spacing block 11 ends of sampling channel 12 feed ends are wedge shape, two inclined-planes of tapered end also have the effect of guiding and attitude adjustment, guiding shrimp body enters in corresponding single-rowization passage, and suitably correct the cruising attitude of raw material shrimp in single-rowization passage, be convenient to the IMAQ of camera 21 and follow-up graphical analysis.
Buffer channel 8 is hinged on conveyer frames 5 tops by rotating shaft 20, sampling channel 12 is arranged on bracing frame 14, and each single-rowization passage both sides are all provided with sensor 26, for responding to the raw material shrimp passing through, sampling channel 12 has transparent base plate simultaneously, and it is all provided with and adopts drawing system up and down.
As shown in Figure 5, adopt drawing system and comprise lighting box 13, light source 22 and camera 21.Lighting box 13 is fixed on bracing frame 14, two lighting box 13 difference arranged in dislocation are at sampling channel 12 upper and lowers, the inwall of lighting box 13 scribbles barium sulfate, can increase uniformity and intensity of illumination that light source 22 irradiates, light source 22 is fixed on the inwall surrounding of lighting box 13, is specially the annular light source being fixed on inwall.The two ends up and down of sampling channel 12 are provided with sensor 26 and sensor 30, wherein camera 21 is controlled by sensor 26, in the time that raw material shrimp touches the sensor 26 of single-rowization passage both sides, sensor 26 sends signal to image processing system, image processing system receives this unblanking camera 21 and gathers image, and image is analyzed, then send according to analysis result the instruction of controlling hierarchy system.The raw material shrimp that sensor 30 skids off for responding to each single-rowization passage, control the magnetic valve of air nozzle 27 receiving after the instruction of hierarchy system, in holding state, in the time that raw material shrimp touches sensor 30, send the signal that starts magnetic valve, open air nozzle 27 and blow raw material shrimp.Adopting in figure process, be the impact of minimizing external environment on raw material shrimp image, the side that sampling channel 12 is relative with each lighting box 13 is provided with background board 29, for isolating external environment.
Hierarchy system comprises that being positioned at sampling channel 12 output bracing frames 14 is provided with cross bar 15, this cross bar 15 is fixed with three air nozzles 27, and be provided with the magnetic valve of controlling each air nozzle 27, the corresponding single-rowization passage of each air nozzle 27, and magnetic valve is controlled by described image processing system.After analyzing by image processing system, the raw material shrimp that sort channel is skidded off is divided into raw material and impurity or certified products and two class of substandard products, wherein certified products skids off rear free-falling along passage, for impurity, image processing system sends control instruction, and corresponding magnetic valve is received after instruction, open air nozzle 27, change the track that impurity skids off single-rowization passage, make impurity and certified products fall into different connecting material in groove, realize the selected classification of raw material shrimp.
Simultaneously, as shown in Figure 6, the first baffle plate 16 and second baffle 17 are also installed on bracing frame 14, the first baffle plate 16 is positioned on the whereabouts track of certified products shrimp, and second baffle 17 is positioned on the whereabouts track of impurity or substandard products, and certified products collides after the first baffle plate 16, falling to certified products connects material in groove 19, impurity or substandard products, under the effect of air nozzle 27 air-flows, fall on second baffle 17, and then landing to substandard products connect material in groove 18.Damage when reducing raw material shrimp collision baffle plate, the face that two baffle plates contact with raw material shrimp, all with supple buffer layer 28, can be spongy layer or silica gel layer etc., meanwhile, each groove that connects material is all contained with water, further keeps the quality of raw material shrimp.
In the present embodiment, the computer in image processing system is connected with camera by cable interface, and computer interface RS-233 is connected with the electromagnetic valve controlling system of controlling three magnetic valves by USB converter.
Wherein, camera adopts the 23 serial DMK-23G618 area array cameras that German imaging company produces; Camera lens is that the model that Chinese Wei Tu vision Science and Technology Ltd. produces is VT-LEM0618-MP3; Conveyer belt is the acetal plastic carrier bar that white has leaking hole and draw-in groove; Electromagnetic valve controlling system is AVR; Buffer channel is stainless steel; Sensor in single-rowization passage is laser type sensor, specifically adopts Japanese Omron company to produce a pair of model to be respectively the optoelectronic switch of E3Z-T61-D and E3Z-T61-L.
The selected hierarchy system of this cover raw material shrimp adopts the primary and secondary structure of PC-AVR, and its structure as shown in Figure 7.In feeding system, conveyer belt 4 inclinations angle can regulate, to adapt to different shrimp body thicknesses.Conveyer belt 4 can be continual by the several rows of raw material shrimp draw-in groove of serving, and raw material shrimp in batches is upwards transmitted with single form; In each draw-in groove, can place at most three raw material shrimps, due to the difference of shrimp body size, may on another draw-in groove, form accumulation; If there are several raw material shrimps to be accumulated in together, backgauge soft board 23 will, floating overlapping raw material shrimp, make only to exist in each draw-in groove the effect of single shrimp.Can make raw material shrimp form simple grain effect by the wedge shaped tip of draw-in groove, direction adjuster and spacing block 11.Buffer channel 8 plays the effect of buffering to rising to the raw material shrimp of peak preparation whereabouts.Be fixed on the direction adjuster on buffer channel 8, the stance adjustment of raw material shrimp is consistent at downslide traffic direction with it to cephlad-caudal.There is certain angle of inclination with ground in sampling channel 12, raw material shrimp can move downward gradually under the effect of gravity.The inclination angle of buffer channel 8 can regulate, the angle theta of sampling channel 12 2can indirectly determine the tiltangleθ of buffer channel 8 1, θ 1be greater than the angle of friction of shrimp and buffer channel 8 storerooms, must guarantee θ 1< θ 2the speed that just can make raw material shrimp glide on buffer channel 8 is less than its speed on sampling channel 12, at this moment just can guarantee that raw material shrimp is unlikely to, in the interior accumulation of sampling channel 12, to impel the singulation of shrimp.
Before whole system brings into operation, first to carry out image registration to adopting drawing system: three onesize round plastic balls are fixed on to the position, middle perpendicular to two lighting box up and down on inclined-plane, open the camera 21 being connected with PC, open and the coexist annular light source of a side of camera 21 simultaneously, make round plastic ball be just arranged in camera 21 under and the position that shows at camera 21 be that four faces are symmetrical.Regulate camera self parameter, comprise time, frame per second, shutter opening time and the gain of aperture, delay exposure.In native system in the time that these parameters are adjusted to following numerical value, the photo that camera is taken is the most clear: aperture f/16, wherein f represents the diameter of the effective aperture of camera, postpones time for exposure 50000us, frame per second 120fps, shutter opening time 1/2000s, gain-adjusted 17.86db.At this moment PC can show the situation of adopting in drawing system in real time, prepares to adopt figure.
The course of work of installing in the present embodiment as shown in Figure 8, in the time of raw material shrimp process sensor 26, can produce a pulse daley signal, while arriving under camera 21 Deng raw material shrimp, camera 21 is accepted sensor and is produced triggering signal and take pictures, and gives image processing module in the image processing system of host computer using image transmitting, below each step be all mutually related, previous step processing result image is supplied with next step and is used, as shown in Figure 9.
1, the hard trigger module flow process of camera: because the upside of lighting box has up been installed three pairs of laser flip flops, the downside of below lighting box has also been installed three pairs of laser flip flops, but their effect is different: the former adopts figure for triggering camera, and the latter opens air nozzle for triggering magnetic valve.A pair of trigger Laser emission mouth is just to putting under normal circumstances, mutually send laser, as long as have object through launching by blocking laser, trigger can moment produces the single-chip microcomputer of an electric pulse to camera or connected electromagnetic valve, after the signal of passing to camera can be selected the moment of moment of taking and exposure to catch image to pass to PC and be for further processing according to the time of its delay; Pass to the signal of single-chip microcomputer and can control magnetic valve according to result.
2, image pretreatment module flow process: because extraneous factor is larger to the attribute interference of camera own, we need to use image noise reduction and the spinning solution in image pretreatment module necessarily to process the image obtaining.If G for original image (m, n) represent, use 9 × 9 windows to carry out mean filter processing to image, replace the original gray value of this image, image use after treatment with the average gray of each point in 9 × 9 neighborhood of pixels
Figure BDA0000463051510000121
represent:
G &OverBar; ( m , n ) = 1 M &Sigma; ( m , n ) &Element; s G ( m , n ) , ( m , n = 0,1,2 . . . N - 1 )
Wherein, S is the set of the point set coordinate in point (m, n) neighborhood, but does not wherein comprise (m, n) point, and M is the sum of set internal coordinate point.
Image
Figure BDA0000463051510000125
in the gray value of each pixel determine by the average gray that is included in several pixels in the neighborhood of appointment of (m, n).In the time that raw material shrimp image departs from calibrating position, depart from the angle of calibration chart picture according to distorted image and utilize clockwise or counterclockwise linear interpolation method infinitely to approach the pixel of distorted image Plays image along the angle of its skew, thereby make its result just be positioned at the center position of camera.
3, module flow process is cut apart in target area: image pretreatment module image after treatment is carried out to region of interesting extraction and cut apart.In this module, the minimum boundary rectangle that in the domain representation entire image of region of interest, prawn occurs.If the visual field actual (tube) length of camera is x, wide is y, and image Raw actual (tube) length that shrimp occupies is x 1, wide is y 1from the attribute of camera own, the image pixel abscissa span of its shooting is [0,640], and ordinate span is [0,480], first take in advance multiple raw material shrimp images, add up the image pixel zone leveling value that its Raw shrimp occupies, i.e. interested pixel region, be made as the length that l × w(l represents interested pixel region, w represents the wide of interested pixel region) pass through formula:
x 640 = x 1 l - - - ( 1 )
y 480 = y 1 w - - - ( 2 )
Can calculate the scope of area-of-interest because the abscissa of this viewing field of camera scope is three passages just, by formula (2) can calculate actual area-of-interest transverse axis, longitudinal axis range be respectively [0, x] and
Figure BDA0000463051510000131
actual area-of-interest length and width be respectively x and
Figure BDA0000463051510000132
in this regional extent, comprise minimum target image.
The target image that this width is comprised to three single width raw material shrimp images below carries out trisection to be cut apart, because the actual length and width in the visual field of camera is x and y, and actual (tube) length that raw material shrimp occupies, wide for x 1and y 1, therefore basis (1) can calculate actual area-of-interest transverse axis, longitudinal axis excursion is respectively:
Figure BDA0000463051510000133
with actual area-of-interest length and width is respectively
Figure BDA0000463051510000135
with
Figure BDA0000463051510000136
therefore the redundancy of initial feed shrimp image is removed, and comprises minimum single raw material shrimp image only in this regional extent.
4, gray level co-occurrence matrixes computing module:
Gray level co-occurrence matrixes is exactly the common method that a kind of spatial correlation characteristic by research gray scale is described texture, it is used for describing from certain angle and specific range direction the probability of two pixels from this pixel to one other pixel, the integrated information of direction, interval, amplitude of variation and the speed of reflection image.This module has been used the feature of following conventional gray level co-occurrence matrixes: entropy, angle second moment, contrast are divided and the moment of inertia.
(1), entropy (ENT)
ENT is the tolerance of the information content that has of image, is the tolerance of a randomness, has reflected the randomness of pixel value in raw material shrimp image.If image rill depth information does not exist, in gray level co-occurrence matrixes, element is almost nil, and entropy is close to zero; If when in gray level co-occurrence matrixes, all values all equates, entropy is got maximum; If when the value in gray level co-occurrence matrixes is very inhomogeneous, entropy is very little.
(2), angle second moment (ASM)
ASM is the quadratic sum of gray level co-occurrence matrixes element value, can be called again energy, has reflected raw material shrimp gradation of image be evenly distributed degree and texture fineness.In the time being distributed on leading diagonal in element set in co-occurrence matrix, illustrate that the intensity profile of raw material shrimp image is comparatively even, macroscopic view, fineness and ASM that image texture distributes are closely bound up: ASM is larger, and image texture degree is thicker; Otherwise image texture degree is thinner.
(3), contrast is divided (IDM)
IDM be metric space gray level co-occurrence matrixes element be expert at or column direction on similarity degree, its value size has reflected the correlation of local gray-value in raw material shrimp image.If gray level co-occurrence matrixes diagonal element have higher value, contrast divides the value will be larger.In the time that matrix element value evenly equates, contrast score value is larger; Otherwise contrast score value is just less.
(4), the moment of inertia (GXJ)
GXJ be image compared with the total amount of zonule grey scale change, reflected the readability of raw material shrimp image and the degree of the texture rill depth.Texture rill is darker, and GXJ is just large, and it is more comprehensive that raw material shrimp superficial makings information embodies; Otherwise raw material shrimp image is just fuzzyyer.
Cut apart the single width raw material shrimp after resume module through image pretreatment module and target area, k represents gray level, and (i, j) is pixel coordinate, and I (i, j) represents the probability that i and j calculate by horizontal direction, has following definition:
Entropy: ENT = - &Sigma; i = 1 k &Sigma; j = 1 k I ( i , j ) log 2 I ( i , j ) - - - ( 3 )
Angle second moment: ASM = &Sigma; i = 1 k &Sigma; j = 1 k I ( i , j ) 2 - - - ( 4 )
Contrast is divided: IDM = &Sigma; i = 1 k &Sigma; j = 1 k I ( i , j ) ( 1 + ( i - j ) 2 ) - - - ( 5 )
The moment of inertia: GXJ = &Sigma; i = 1 k &Sigma; j = 1 k | i - j | I ( i , j ) - - - ( 6 )
Bring in formula (3), formula (4), formula (5) and formula (6) according to 15 °, 30 °, 45 °, 60 °, 90 °, 120 °, 150 °, 180 ° eight directions cutting apart calculated for pixel values direction in the single width raw material shrimp image after resume module through image pretreatment module and target area, we can calculate entropy, angle second moment, the contrast of each width raw material shrimp image and divide and the moment of inertia, corresponding four matrixes of result of each angle calculation, they are first averaging to summation again can obtain:
Mean entropy and:
Figure BDA0000463051510000145
Average angle second moment and:
Figure BDA0000463051510000146
Average contrast divide and:
Figure BDA0000463051510000147
Average the moment of inertia and:
Figure BDA0000463051510000151
The mean entropy calculating by above-mentioned formula and, average angle second moment and, average contrast divide and, average the moment of inertia and, be required textural characteristics.
5, decision tree classifier module
Decision tree (ID3) is a forecast model, he can representative object attribute and object value between a kind of mapping relations.In decision tree, each node represents certain object, certain possible property value that each bifurcated path represents, and each leaf node is the value of the represented object in the corresponding path experiencing to this leaf node from root node.Decision tree only has single output, has plural number output if want, and can set up independently decision tree and export to process difference.
Through first three step can obtain the corresponding mean entropy of each width raw material shrimp image and, average angle second moment and average contrast is divided and, on average the moment of inertia and, our characteristic value using these four values as classified image, in the situation that the various parameters of camera all mix up, adopt online figure, altogether gather N width image, its Raw shrimp image is designated as " just " (certified products), non-raw material shrimp image is designated as " bearing " (substandard products), the first half image is used for training pattern, later half image is used as inspection set, the robustness of verification model.Any one data characteristics screening scope all can be selected as the top that is positioned at tree, is performed and enters next condition case statement in the time that each node satisfies condition feature, provides rank judgement if meet leaf node condition; If do not met, continue to carry out downwards, until leaf node condition meets this characteristic value.The training process of decision tree is as follows: the first width raw material shrimp image of collected by camera is through after image pretreatment and region of interesting extraction, mean entropy and (ENT`), average angle second moment and (ASM`), average contrast divide and (IDM`), average the moment of inertia and (GXJ`) be made as successively a 11, a 12, a 13, a 14; The mean entropy of the second width raw material shrimp image and, average angle second moment and, average contrast divide and, average the moment of inertia and be made as successively a 21, a 22, a 23, a 24.Make the top of ENT` as this decision tree, its span is [a 21, a 11], ASM` is as first leaf left sibling, and its span is [a 22, a 12], ASM` is as the right node of first leaf, and its span is [a 42, a 32] ... by that analogy, when collected the completing of last n/2 width image, IDM` is as Far Left leaf node, and span is
Figure BDA0000463051510000152
rank below it provides, and is all certified products; ENT` is as rightmost leaf node, and its span is
Figure BDA0000463051510000161
rank below it provides, and is also all certified products, now declares that training pattern set up.The concrete result of decision as shown in figure 10, wherein [a 21, a 11], [a 41, a 31] [a 22, a 12], [a 42, a 32]
Figure BDA0000463051510000163
[a 23, a 13], [a 43, a 33]
Figure BDA0000463051510000164
[a 24, a 14], [a 44, a 34]
Figure BDA0000463051510000165
represent respectively mean entropy that characteristic attribute that training pattern is met is corresponding and, average angle second moment and, average contrast divide and, average the moment of inertia and the span of textural characteristics, can find out and in the time that each node satisfies condition feature, be given to left branch, otherwise be assigned to right branch.After training pattern is set up, use the hard trigger module collecting test of camera image, will repeat successively above image pretreatment module-target area for each test sample book and cut apart module-gray level co-occurrence matrixes computing module, now system can provide best classification grade by automatic analysis.
Start below to carry out the image measurement stage, image that each width camera gathers is through after pretreatment and region of interesting extraction, calculate mean entropy and, average angle second moment and, average contrast divide and, judge that whether their feature is at [a 21, a 11], [a 41, a 31]
Figure BDA0000463051510000166
[a 22, a 12], [a 42, a 32]
Figure BDA0000463051510000167
[a 23, a 13], [a 43, a 33] [a 24, a 14], [a 44, a 34]
Figure BDA0000463051510000169
within interval, allow their numerical value start successively downward traversal from the top of decision-tree model of just now setting up, satisfy condition and just give accordingly the rank that leaf node is corresponding, rank exists " just " and " bear " two kinds possible, respectively correspondence certified products groove and the substandard products groove that connects material that connects material; When not satisfying condition, continue traversal, till through satisfying condition, final leaf node corresponding to qualified region can be considered to final result.
6, display module
Result can be real-time in display module, show, the demonstration comprising characteristic value: mean entropy and, average angle second moment and, average contrast divide and, average the moment of inertia and, image total amount, certified products shrimp quantity, substandard products shrimp quantity, training is modeled as power, test success rate.
Image processing module by design can obtain classification results, and result is passed to the executing agency's delivery nozzle control instruction in image processing system, and then three magnetic valves of control open or close.Because the position of nozzle and raw material shrimp terminal position are very approaching, in the time that raw material shrimp runs to shower nozzle below, magnetic valve is opened, and nozzle moment blows.For certified products shrimp, air nozzle keeps closing, and certified products shrimp collides and falls into certified products after the first plate washer 16 and connect material in groove 19, for impurity, spray at air nozzle 27 under the effect of air-flow, change the whereabouts track of impurity, it is touched and fall into substandard products after second baffle 18 and connect material in groove.

Claims (10)

1. the automatic well-chosen grading plant of a seed shrimp, is characterized in that, comprises feeding system, sort channel, adopts drawing system, hierarchy system and image processing system;
Feeding system is used for single the shrimp body of the treating sorting sort channel that exports to;
Sort channel is divided into the buffer channel (8) and the sampling channel (12) that are connected successively, sampling channel (12) is separated into multiple single-rowization passages, described buffer channel is provided with multiple directions adjuster in (8), for the shrimp of feeding system output is divided into single and regulates each shrimp body to enter the attitude of corresponding single-rowization passage;
Adopt drawing system for gathering the image of shrimp body in each single-rowization passage, adopt drawing system and comprise and be positioned at two lighting box (13) that sampling channel (12) is upper and lower and dislocation is arranged, in each lighting box, be equipped with camera (21);
Hierarchy system comprises the air nozzle (27) corresponding with each single-rowization passage, and this air nozzle (27) is controlled by described image processing system, enters the different grooves that connects material for blowing by the shrimp body of single-rowization passage output;
Image processing system, for analyzing described image, to the image grading processing of each shrimp body, according to classification results, and sends the signal of controlling hierarchy system.
2. the automatic well-chosen grading plant of shrimp as claimed in claim 1, is characterized in that, described buffer channel (8) is all in tilted layout with sampling channel (12), and the angle of inclination of sampling channel (12) is greater than the angle of inclination of buffer channel (8).
3. the automatic well-chosen grading plant of shrimp as claimed in claim 1, it is characterized in that, described direction adjuster comprises the locating piece (9) being arranged in buffer channel (8), locating piece (9) is fixed with guide plate (10) towards one end of feeding system, this guide plate (10) has the guiding surface that acts on shrimp body.
4. the automatic well-chosen grading plant of shrimp as claimed in claim 1, it is characterized in that, light source (22) is installed in described lighting box (13), the inwall of lighting box (13) scribbles barium sulfate, and the sampling channel corresponding with each lighting box (13) above or below be provided with black background plate (29).
5. the automatic well-chosen grading plant of shrimp as claimed in claim 1, it is characterized in that, the input of each single-rowization passage and output are fixed with respectively first sensor (26) and the second sensor (30), two sensors all access described image processing system, first sensor (26) carries out the signal of image taking for sending control camera (21), the second sensor (30) sends the signal of opening air nozzle (27).
6. the automatic well-chosen grading plant of shrimp as claimed in claim 1, it is characterized in that, also be provided with the bracing frame (14) that sort channel is installed, be positioned on the bracing frame (14) of sort channel output the first baffle plate (16) and second baffle (17) are installed, the first baffle plate (16) acts on the shrimp body freely being glided by each single-rowization passage, second baffle (17) is for arc and act on the shrimp body being blown by air nozzle (27), and the face that the first baffle plate (16) contacts with shrimp body with second baffle (17) is equipped with supple buffer layer (28).
7. the automatic well-chosen stage division of a seed shrimp, is characterized in that, comprises the following steps:
1) feeding system is by the single shrimp body sort channel that exports to, and each shrimp body slides at the single-rowization passage of sort channel with the attitude adapting to;
2) utilize the image of collected by camera shrimp body, and utilize image processing system to carry out pretreatment to the image collecting, obtain the area-of-interest in raw material shrimp image, the gray level co-occurrence matrixes information in raw material shrimp of extracting, as textural characteristics, uses decision tree classifier to classify to the textural characteristics extracting;
3) hierarchy system is according to described classification results, and control air nozzle blows the raw material shrimp being skidded off by single-rowization passage and enters the different grooves that connects material.
8. the automatic well-chosen stage division of shrimp as claimed in claim 7, it is characterized in that, in step 2) in, in image pretreatment, use 9 × 9 windows to carry out mean filter processing to raw material shrimp image, each point average gray in 9 × 9 neighborhoods around each pixel of this image is replaced to the original each point gray value of this image; Region of interesting extraction process is as follows:
If the visual field actual (tube) length of camera is x, wide is y, and raw material shrimp actual (tube) length that image occupies is x 1, wide is y 1, according to formula:
Figure FDA0000463051500000021
can calculate the size of area-of-interest, wherein, l represents the length of area-of-interest, and w represents the wide of area-of-interest.
9. the automatic well-chosen stage division of shrimp as claimed in claim 7, it is characterized in that, in step 2) in, use gray scale symbiosis battle array to extract that entropy, angle second moment, contrast divide and the feature of the moment of inertia to area-of-interest, again according to 15 °, 30 °, 45 °, 60 °, 90 °, 120 °, 150 °, 180 ° eight direction calculating mean entropies and, average angle second moment and, average contrast divide and, average the moment of inertia and, specific formula for calculation is:
Mean entropy and:
Figure FDA0000463051500000031
Average angle second moment and:
Figure FDA0000463051500000032
Average contrast divide and:
Figure FDA0000463051500000033
Average the moment of inertia and:
Figure FDA0000463051500000034
Wherein ENT, ASM, IDM, GXJ represent that respectively entropy, angle second moment, contrast divide and the moment of inertia, by above-mentioned formula calculate mean entropy and, average angle second moment and, average contrast divide and, average the moment of inertia and, be described textural characteristics.
10. the automatic well-chosen stage division of shrimp as claimed in claim 7, is characterized in that, according to described textural characteristics, use decision tree classifier to carry out Decision Classfication to the textural characteristics that best embodies original image, travel through whole decision tree, in the time that leaf node is " just ", corresponding is certified products; In the time that leaf node is " bearing ", corresponding is substandard products.
CN201410040879.2A 2014-01-27 2014-01-27 The automatic well-chosen stage division of shrimp and device Active CN103801520B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410040879.2A CN103801520B (en) 2014-01-27 2014-01-27 The automatic well-chosen stage division of shrimp and device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410040879.2A CN103801520B (en) 2014-01-27 2014-01-27 The automatic well-chosen stage division of shrimp and device

Publications (2)

Publication Number Publication Date
CN103801520A true CN103801520A (en) 2014-05-21
CN103801520B CN103801520B (en) 2015-11-25

Family

ID=50699185

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410040879.2A Active CN103801520B (en) 2014-01-27 2014-01-27 The automatic well-chosen stage division of shrimp and device

Country Status (1)

Country Link
CN (1) CN103801520B (en)

Cited By (20)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104174598A (en) * 2014-09-01 2014-12-03 安徽唯嵩光电科技有限公司 Sorting machine for sorting hard materials and sorting method thereof
CN105293149A (en) * 2015-11-26 2016-02-03 倪煌斌 Conveying table for collected paper piles
CN105344613A (en) * 2015-11-26 2016-02-24 浙江大学 Brazil pine nut single grain achieving device
CN106455594A (en) * 2014-04-24 2017-02-22 莱特拉姆有限责任公司 Shrimp processing system and methods
CN107444855A (en) * 2017-09-01 2017-12-08 深圳市艾克斯自动化技术开发有限公司 A kind of multiple row conveying aligns machine
CN107661864A (en) * 2017-11-27 2018-02-06 江苏食品药品职业技术学院 Lobster classified screening device based on Machine Vision Detection
CN108072668A (en) * 2016-11-18 2018-05-25 中国科学院沈阳自动化研究所 Bullet open defect automatic recognition system based on Photoelectric Detection
CN108921057A (en) * 2018-06-19 2018-11-30 厦门大学 Prawn method for measuring shape of palaemon, medium, terminal device and device based on convolutional neural networks
CN109092702A (en) * 2018-08-10 2018-12-28 河南科技大学 A kind of agaricus bisporus primary election stage division and device
CN109169465A (en) * 2018-09-04 2019-01-11 中国水产科学研究院黄海水产研究所 The system and method for the underwater living body prawn growth parameter(s) of intelligent determination
CN109829465A (en) * 2019-01-03 2019-05-31 上海交通大学 A method of it searching prawn and most preferably draws position and identification uropodium feature
KR20190068663A (en) * 2017-12-08 2019-06-19 한국로봇융합연구원 Object auto sorting, classifying system using image processing algorithm
CN110385277A (en) * 2019-08-30 2019-10-29 贵州大学 A kind of apple sorting equipment and its method for sorting
CN111050931A (en) * 2017-07-10 2020-04-21 阿朗新科德国有限责任公司 Inspection device and method for visual inspection of elastomeric particles
CN111093377A (en) * 2017-09-05 2020-05-01 莱特拉姆有限责任公司 Singulating and separating system for shrimp processing machines
CN112077012A (en) * 2020-08-31 2020-12-15 林海云 Pepper piling-prevention sorting equipment
CN112474388A (en) * 2020-10-27 2021-03-12 天津合荣钛业有限公司 Reduced titanium separator and subpackaging machine applying same
CN112841099A (en) * 2021-03-19 2021-05-28 东营市阔海水产科技有限公司 Detection apparatus for detect shrimp head pathological change based on image
CN112907503A (en) * 2020-07-24 2021-06-04 嘉兴学院 Penaeus vannamei Boone quality detection method based on adaptive convolutional neural network
CN113857065A (en) * 2021-11-01 2021-12-31 合肥泰禾智能科技集团股份有限公司 Sorting mechanism and sorting device

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5346424A (en) * 1993-08-20 1994-09-13 Horng Shen Machinery Co., Ltd. Automatic size-grading and shrimp peeling machinery
JPH0731424A (en) * 1993-07-15 1995-02-03 Shizuo Hikita Wind selector for boiled and dried larval fish
KR20060091088A (en) * 2005-02-14 2006-08-18 김윤호 Grader for shrimps
CN201108008Y (en) * 2007-11-12 2008-09-03 彭德权 Fish and shrimp automatic numerical control grading screen grader
CN201470593U (en) * 2009-08-14 2010-05-19 江苏九寿堂生物制品有限公司 Lobster grading machine
CN202918945U (en) * 2012-11-22 2013-05-08 福建省丰盛食品有限公司 Sorting device for shrimps
CN203018326U (en) * 2012-12-21 2013-06-26 浙江大学 Prawn quality detecting-grading device based on machine vision technology

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH0731424A (en) * 1993-07-15 1995-02-03 Shizuo Hikita Wind selector for boiled and dried larval fish
US5346424A (en) * 1993-08-20 1994-09-13 Horng Shen Machinery Co., Ltd. Automatic size-grading and shrimp peeling machinery
KR20060091088A (en) * 2005-02-14 2006-08-18 김윤호 Grader for shrimps
CN201108008Y (en) * 2007-11-12 2008-09-03 彭德权 Fish and shrimp automatic numerical control grading screen grader
CN201470593U (en) * 2009-08-14 2010-05-19 江苏九寿堂生物制品有限公司 Lobster grading machine
CN202918945U (en) * 2012-11-22 2013-05-08 福建省丰盛食品有限公司 Sorting device for shrimps
CN203018326U (en) * 2012-12-21 2013-06-26 浙江大学 Prawn quality detecting-grading device based on machine vision technology

Cited By (31)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106455594A (en) * 2014-04-24 2017-02-22 莱特拉姆有限责任公司 Shrimp processing system and methods
CN104174598A (en) * 2014-09-01 2014-12-03 安徽唯嵩光电科技有限公司 Sorting machine for sorting hard materials and sorting method thereof
CN105293149A (en) * 2015-11-26 2016-02-03 倪煌斌 Conveying table for collected paper piles
CN105344613A (en) * 2015-11-26 2016-02-24 浙江大学 Brazil pine nut single grain achieving device
CN105344613B (en) * 2015-11-26 2018-01-26 浙江大学 A kind of Brazilian pine nut list granulating device
CN108072668A (en) * 2016-11-18 2018-05-25 中国科学院沈阳自动化研究所 Bullet open defect automatic recognition system based on Photoelectric Detection
CN111050931A (en) * 2017-07-10 2020-04-21 阿朗新科德国有限责任公司 Inspection device and method for visual inspection of elastomeric particles
CN107444855B (en) * 2017-09-01 2023-01-17 深圳市艾克斯自动化技术开发有限公司 Multi-row conveying and directional arranging machine
CN107444855A (en) * 2017-09-01 2017-12-08 深圳市艾克斯自动化技术开发有限公司 A kind of multiple row conveying aligns machine
US11632961B2 (en) 2017-09-05 2023-04-25 Laitram, L.L.C. Singulation and separation system for a shrimp processor
US11632962B2 (en) 2017-09-05 2023-04-25 Laitram, L.L.C. Singulation and separation system for a shrimp processor
CN115444022A (en) * 2017-09-05 2022-12-09 莱特拉姆有限责任公司 Singulating and separating system for shrimp processing machines
CN111093377B (en) * 2017-09-05 2022-07-15 莱特拉姆有限责任公司 Singulating and separating system for shrimp processing machines
CN111093377A (en) * 2017-09-05 2020-05-01 莱特拉姆有限责任公司 Singulating and separating system for shrimp processing machines
CN107661864B (en) * 2017-11-27 2019-04-02 江苏食品药品职业技术学院 Cray classified screening device based on Machine Vision Detection
CN107661864A (en) * 2017-11-27 2018-02-06 江苏食品药品职业技术学院 Lobster classified screening device based on Machine Vision Detection
KR20190068663A (en) * 2017-12-08 2019-06-19 한국로봇융합연구원 Object auto sorting, classifying system using image processing algorithm
KR102109698B1 (en) 2017-12-08 2020-05-12 한국로봇융합연구원 Object auto sorting, classifying system using image processing algorithm
CN108921057B (en) * 2018-06-19 2021-06-01 厦门大学 Convolutional neural network-based prawn form measuring method, medium, terminal equipment and device
CN108921057A (en) * 2018-06-19 2018-11-30 厦门大学 Prawn method for measuring shape of palaemon, medium, terminal device and device based on convolutional neural networks
CN109092702A (en) * 2018-08-10 2018-12-28 河南科技大学 A kind of agaricus bisporus primary election stage division and device
CN109169465A (en) * 2018-09-04 2019-01-11 中国水产科学研究院黄海水产研究所 The system and method for the underwater living body prawn growth parameter(s) of intelligent determination
CN109829465A (en) * 2019-01-03 2019-05-31 上海交通大学 A method of it searching prawn and most preferably draws position and identification uropodium feature
CN109829465B (en) * 2019-01-03 2023-03-24 上海交通大学 Method for searching optimum sucking position of prawn and identifying tail limb characteristics
CN110385277A (en) * 2019-08-30 2019-10-29 贵州大学 A kind of apple sorting equipment and its method for sorting
CN112907503A (en) * 2020-07-24 2021-06-04 嘉兴学院 Penaeus vannamei Boone quality detection method based on adaptive convolutional neural network
CN112907503B (en) * 2020-07-24 2024-02-13 嘉兴学院 Penaeus vannamei Boone quality detection method based on self-adaptive convolutional neural network
CN112077012A (en) * 2020-08-31 2020-12-15 林海云 Pepper piling-prevention sorting equipment
CN112474388A (en) * 2020-10-27 2021-03-12 天津合荣钛业有限公司 Reduced titanium separator and subpackaging machine applying same
CN112841099A (en) * 2021-03-19 2021-05-28 东营市阔海水产科技有限公司 Detection apparatus for detect shrimp head pathological change based on image
CN113857065A (en) * 2021-11-01 2021-12-31 合肥泰禾智能科技集团股份有限公司 Sorting mechanism and sorting device

Also Published As

Publication number Publication date
CN103801520B (en) 2015-11-25

Similar Documents

Publication Publication Date Title
CN103801520B (en) The automatic well-chosen stage division of shrimp and device
CN106733701B (en) Potato multi-stage combination dysnusia detecting system and detection method
CN109784204B (en) Method for identifying and extracting main fruit stalks of stacked cluster fruits for parallel robot
CN103033512B (en) Method for recognizing hatching egg incubation based on hyperspectrum
CN110472575B (en) Method for detecting ripeness of tomatoes stringing together based on deep learning and computer vision
Jin et al. Design and implementation of anti-leakage planting system for transplanting machine based on fuzzy information
CN104597052A (en) High-speed lossless potato grading detection method and system based on multi-characteristic fusion
Thakur et al. An innovative approach for fruit ripeness classification
CN103808669A (en) Rapid nondestructive apple wormhole testing method based on hyperspectral imaging technology
CN104070016A (en) Granular agricultural material real-time detecting and grading device and granular agricultural material real-time detecting and grading method
CN113145492A (en) Visual grading method and grading production line for pear appearance quality
CN103646251B (en) Apple postharvest field classification detection method and system based on embedded technology
CN114354637A (en) Fruit quality comprehensive grading method and device based on machine vision and X-ray
CN107103608A (en) A kind of conspicuousness detection method based on region candidate samples selection
CN102680488B (en) Device and method for identifying massive agricultural product on line on basis of PCA (Principal Component Analysis)
CN115082719A (en) Wood quality grading method
CN113313692B (en) Automatic banana young plant identification and counting method based on aerial visible light image
Ni et al. Convolution neural network based automatic corn kernel qualification
CN116758539B (en) Embryo image blastomere identification method based on data enhancement
Esehaghbeygi et al. Digital image processing for quality ranking of saffron peach
CN203750879U (en) Image capturing equipment for automatically sorting and grading shrimps
CN116188489A (en) Wheat head point cloud segmentation method and system based on deep learning and geometric correction
Gan et al. Design and testing of a machine-vision-based air-blow sorting platform for famous tea fresh leaves production
CN112164120A (en) Dark spotted egg detection method based on convolutional neural network GoogLeNet model
WO2024108141A2 (en) Technologies for analysis of agricultural products

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant