CN109663747B

CN109663747B - Intelligent detection method for Chinese chestnut wormholes

Info

Publication number: CN109663747B
Application number: CN201910038084.0A
Authority: CN
Inventors: 何文斌; 明五一; 都金光; 魏爱云; 李安生; 马军; 李晓科; 刘琨; 曹阳; 侯俊剑; 贾豪杰; 楚昌昊; 张笑笑; 王萌
Original assignee: Zhengzhou University of Light Industry
Current assignee: Zhengzhou University of Light Industry
Priority date: 2019-01-16
Filing date: 2019-01-16
Publication date: 2020-09-25
Anticipated expiration: 2039-01-16
Also published as: CN109663747A

Abstract

An intelligent detection method for Chinese chestnut wormholes comprises the following steps: firstly, the conveying belt drives the Chinese chestnuts to move from left to right; secondly, positioning and locking the Chinese chestnuts into the grooves of the conveying belt; thirdly, visual detection is carried out; fourthly, carrying out laser detection; fifthly, carrying out acoustic detection; sixthly, unlocking the Chinese chestnut; and seventhly, the Chinese chestnuts which are determined to have wormholes are sorted by the sorting device, and the intact and qualified Chinese chestnuts are conveyed rightwards by the horizontal conveying belt (4). According to the invention, under the multi-stage combined intelligent information fusion of machine vision, acoustic detection and laser detection, the high-precision Chinese chestnut wormhole detection can be realized, on one hand, the intelligent level of detection can be effectively improved, on the other hand, the sorting accuracy is improved and the device cost is reduced through the multi-stage combined intelligent detection. The method can be used for on-line real-time detection of the wormholes of the Chinese chestnut agricultural products, has important significance for improving the development of deep processing of the Chinese chestnut agricultural products in China, and has good market application prospect.

Description

Intelligent detection method for Chinese chestnut wormholes

Technical Field

The invention belongs to the technical field of agricultural product appearance quality detection used in the agricultural field, and particularly relates to an intelligent detection method for chestnut wormholes.

Background

The chestnut essence has the reputation of 'the king of dried fruits', the chestnuts are rich in nutrition, the fruits contain 70.1 percent of sugar and starch, and 7 percent of protein. In addition, the chestnut cake also contains fat, calcium, phosphorus, iron, various vitamins and trace elements, particularly the contents of vitamin C, B1 and carotene are higher than those of common dry fruits, the chestnut is rich in nutrition, and besides starch, the chestnut cake also contains nutrient substances such as monosaccharide and disaccharide, carotene, thiamine, riboflavin, nicotinic acid, ascorbic acid, protein, fat, inorganic salts and the like. However, in the planting process of the Chinese chestnut, various attacks of plant diseases and insect pests are often encountered, so that the wormhole detection of the Chinese chestnut is very key for improving the grade of the Chinese chestnut.

At present, the detection of the Chinese chestnut moth eyes mostly stays in the stage of identification and judgment by artificial senses, and a lot of inconvenience exists. The mode of choosing by hand is relied on to accomplish the sorting of chinese chestnut wormhole, and firstly a large amount of labours of wasting, along with the harvest season the recruitment is too old and useless, secondly there is the hourglass to examine easily, and the testing result uniformity is poor, efficient, is difficult to satisfy high standard hierarchical requirement, is unfavorable for realizing the automation. With the accelerated progress of the artificial intelligence technology, the accuracy of the insect eye detection can be improved by utilizing the multi-stage combined detection of the multi-source information intelligent fusion.

Although the Chinese chestnut grading method and equipment based on computer vision gradually become hot at present, the method is generally limited to laboratory research or a camera is adopted to take pictures of the Chinese chestnuts, and the adopted technical means is single. Under the conditions that the wormhole is small or the shape and the position are complex, the judgment cannot be accurately carried out, and the risk of missing detection exists; or the hyperspectral imaging technology is adopted to detect the Chinese chestnut to be detected, but the cost and the complexity of the detection device are increased.

In order to overcome the defects of the conventional Chinese chestnut wormhole sorting device, a multistage combined Chinese chestnut wormhole intelligent detection method and device based on machine vision, laser and acoustics are provided.

Some relevant patent documents are found through the search of domestic patent documents, and the following are mainly found:

1. the publication number is CN 108801971A, the name is 'a mold infection Chinese chestnut detection method based on a hyperspectral imaging technology', the invention provides a mold infection Chinese chestnut detection method based on a hyperspectral imaging technology, but the high-precision detection of Chinese chestnut wormholes cannot be realized by adopting a single hyperspectral imaging technical scheme.

2. The invention discloses a Chinese chestnut automatic sorting device and a Chinese chestnut automatic sorting method with the publication number of CN 107350173A, and the name of the invention is 'Chinese chestnut automatic sorting device and method', which adopts a plurality of cameras, computers, PLC, electromagnetic valves and the like, utilizes the algorithm of machine vision to identify and sort Chinese chestnuts, and carries out filtering and contour extraction according to surface color pictures of the Chinese chestnuts to obtain the effective proportion of the area of damaged Chinese chestnuts and realize the sorting of the Chinese chestnuts to be classified, but the classification device can not accurately judge under the conditions of small wormholes or complex shapes and positions.

3. The invention discloses a multi-information fusion electromagnetic crawler type Chinese chestnut detection and screening device with the publication number of CN 108620338A, and is named as a multi-information fusion electromagnetic crawler type Chinese chestnut detection and screening device.

Although the above patent provides a sorting method and a grading device for Chinese chestnuts, some devices can carry out grading according to appearance characteristics, but the detection of the damage and mildew of the Chinese chestnuts can be realized only by adopting a machine vision method or combining with hyperspectral and infrared spectrum technologies. However, the detection of the chestnut wormholes is not comprehensive, and particularly, the probability of missed detection exists under the condition that the wormholes are small or the shapes and the positions are complex.

Disclosure of Invention

The invention aims to overcome the defects in the prior art and provides the intelligent detection method for the chestnut wormhole, which can further improve the accuracy of wormhole detection and reduce the omission factor.

In order to solve the technical problems, the invention adopts the following technical scheme: an intelligent detection method for chestnut wormholes, which adopts intelligent detection equipment for chestnut wormholes to operate; the intelligent detection equipment for the wormholes of the Chinese chestnuts comprises a rack 1, a lifting conveyer belt 2 and a horizontal conveyer belt 4 are sequentially arranged on the rack 1 from left to right, the lifting conveyer belt 2 is arranged in a left-low-right-high-inclination manner, a photographing darkroom detection module 3, a laser detection module 5, an acoustic detection module 6 and a sorting device 7 are sequentially arranged on the rack 1 from left to right, the photographing darkroom detection module 3 is positioned above the horizontal conveyer belt 4, an electric control module 9, a Chinese chestnut locking device and a Chinese chestnut unlocking device are arranged on the side portion of the rack 1, the working portion of the Chinese chestnut locking device is positioned above the lifting conveyer belt 2, and the working portion of the Chinese chestnut unlocking device is positioned between; the lifting conveying belt 2 and the horizontal conveying belt 4 are both made of metal heat conductor materials, grooves 54 for storing Chinese chestnuts to be conveyed are uniformly formed in the upper surfaces of the lifting conveying belt 2 and the horizontal conveying belt 4 at intervals of 5-8 cm in the left-right direction, and the lifting conveying belt 2 and the horizontal conveying belt 4 are combined into a complete conveying belt;

the detection method comprises the following steps:

firstly, the lifting conveyer belt 2 and the horizontal conveyer belt 4 are driven by a stepping motor to move left and right, Chinese chestnuts are placed into the groove 54 on the left side of the top of the lifting conveyer belt 2 one by one, and the lifting conveyer belt 2 and the horizontal conveyer belt 4 drive the Chinese chestnuts to move right;

secondly, the Chinese chestnut locking device positions and locks the Chinese chestnut moved below the Chinese chestnut locking device into the groove;

thirdly, the lifting conveyer belt 2 and the horizontal conveyer belt 4 drive the Chinese chestnuts to move rightwards to the shooting darkroom detection module 3 for visual detection;

fourthly, the lifting conveying belt 2 and the horizontal conveying belt 4 drive the Chinese chestnuts to move rightwards to the lower part of the laser detection module 5 for laser detection;

lifting the conveying belt 2 and the horizontal conveying belt 4 to drive the Chinese chestnuts to move rightwards to the position below the acoustic detection module 6 for acoustic detection;

the Chinese chestnut unlocking device unlocks the Chinese chestnuts moved to the lower part of the Chinese chestnut unlocking device from the grooves;

and seventhly, the Chinese chestnuts which are determined to have wormholes are sorted by the sorting device, and the intact and qualified Chinese chestnuts are conveyed rightwards by the horizontal conveying belt 4.

The Chinese chestnut locking device is a cold air spray pipe 37 and a water dropping pipe which are arranged above the lifting conveyer belt 2 side by side, the water dropping pipe is positioned at the left side of the cold air spray pipe 37, and a water dropping hole of the water dropping pipe and a nozzle of the cold air spray pipe 37 are both arranged downwards and face towards a passing groove 54; the cold air spray pipe 37 is provided with a second electromagnetic valve 52;

the Chinese chestnut unlocking device is a hot air spray pipe 38 arranged above the horizontal conveying belt 4, a nozzle of the hot air spray pipe 38 is arranged downwards and faces towards a passing groove 54, a third electromagnetic valve 53 is arranged on the hot air spray pipe 38, and the second electromagnetic valve 52 and the third electromagnetic valve 53 are connected with the electronic control module 9;

the concrete process of the step II is as follows: the water in the water dropping pipe is dropped into the groove 54 right below, the Chinese chestnut is placed in the groove with liquid water, after the Chinese chestnut moves to the lower part of the nozzle of the cold air spray pipe 37, the electric control module 9 controls the second electromagnetic valve 52 to be opened, the high-pressure cold air is sprayed to the groove 54 with the liquid water, and the liquid water is frozen and fixed in the groove 54;

the concrete process comprises the following steps: when the detected chestnuts are conveyed to the right below the nozzles of the hot air spray pipes 38, the solid ice in the grooves 54 of the horizontal conveying belt 4 of the metal heat conductor is rapidly heated by hot air to be converted into liquid water, the chestnuts are unlocked, and the sorting device 7 is convenient to sort.

The photographing darkroom detection module 3 comprises a cavity box body 10, an annular LED lamp 11 and an industrial camera 12, the bottom of the cavity box body 10 is open and vertically faces downwards to the horizontal conveying belt 4, the left side and the right side of the cavity box body 10 are fixed on the rack 1 through screws, and the left side and the right side of the bottom edge of the cavity box body 10 are provided with horizontal conveying rectangular gaps convenient for Chinese chestnuts; an annular LED lamp 11 and an industrial camera 12 are installed at the top center position in the cavity box 10;

the concrete process of step three is: promote conveyer belt 2 and horizontal transport belt 4 and be driven by carrying step motor, every section width of conveyer belt is 30~50mm, by step motor preliminary location, will wait to detect the chinese chestnut and send into inside camera darkroom detection module 3, the position is further calculated to the machine vision algorithm of electric control module 9, drive step motor precision positioning, make it moves to wait to detect the chinese chestnut and be located industrial camera 12 below region in camera darkroom detection module 3, calculate the chinese chestnut position through electric control module 9's machine vision method, camera darkroom detection module 3's industrial camera 12 is under electric control module 9's control, treat the pick-up plate and shoot the chinese chestnut, send electric control module 9 to through industrial ethernet communication submodule, handle through embedded ARM singlechip, the flow is: by calibrating the industrial camera 12, the height H from the lens of the industrial camera 12 to the real object space is a known quantity, a monocular vision method is adopted, the included angle a can be calculated through a similar triangular proportion, L1 is the length of a pixel and can be obtained through an image processing algorithm, so that the L2 numerical value in the real object space is calculated, and the coordinate position of the Chinese chestnut to be detected in a photographing darkroom is obtained; then, acquiring the size of the Chinese chestnut to be detected and the area information of suspected insect eyes through feature extraction and analysis of the image; the calculated position and size information of the Chinese chestnut to be detected provides positioning information for the subsequent laser detection module 5 and acoustic detection module 6.

The laser detection module 5 comprises a first X-direction stepping motor 13, a first X-direction moving screw rod 14, a first X-direction support 15, a first Y-direction stepping motor 16, a first Y-direction moving screw rod 17, a first Y-direction guide rail 18, a laser detection clamp 19 and a laser detection sub-module 20; the X direction is a left-right horizontal direction, the Y direction is a front-back horizontal direction, a first support guide rail 55 which is arranged along the X direction and is a rectangular frame is respectively arranged on the front side and the back side of the horizontal conveying belt 4 on the machine frame 1, a first X-direction movement screw rod 14 is arranged in each first support guide rail 55, one end of each first X-direction movement screw rod 14 is rotatably connected to the first support guide rail 55, the other end of each first X-direction movement screw rod 14 is in transmission connection with the first X-direction stepping motor 13, the first X-direction movement screw rods 14 are in threaded connection with the first X-direction supports 15 which slide along the first support guide rails 55, the front end and the back end of each first Y-direction guide rail 18 are fixedly connected to the upper ends of the two first X-direction supports 15, each first Y-direction guide rail 18 is of a hollow rod-shaped structure, each first Y-direction movement screw rod 17 is arranged in the first Y-direction guide rail 18, one end of a first Y-direction moving screw rod 17 is rotatably connected to one end inside a first Y-direction guide rail 18, the other end of the first Y-direction moving screw rod 17 is in transmission connection with the first Y-direction stepping motor 16, a laser detection clamp 19 is in threaded connection with the first Y-direction moving screw rod 17 and is arranged along the first Y-direction guide rail 18 in a sliding mode, and a laser detection sub-module 20 is installed on the laser detection clamp 19 and irradiates towards the groove 54;

the specific process of the step IV is as follows: the horizontal conveying belt 4 drives the Chinese chestnuts to move rightwards to the position below the laser detection module 5, the first X-direction stepping motor 13 drives the first X-direction moving screw rod 14 to rotate, so that the first X-direction support 15 of the first X-direction moving screw rod 14 moves leftwards or rightwards along the first support guide rail 55, and the first Y-direction guide rail 18, the laser detection clamp 19 and the laser detection sub-module 20 also move leftwards or rightwards; the first Y-direction stepping motor 16 drives the first Y-direction moving screw rod 17 to rotate, so that the laser detection clamp 19 connected with the first Y-direction moving screw rod 17 moves forwards or backwards along the first Y-direction guide rail 18, and the laser detection sub-module 20 also moves forwards or backwards; moving the laser detection clamp 19 to the position above the Chinese chestnut to be detected through X-Y plane positioning, and aligning the laser detection sub-module 20 to the position vertically above the suspected Chinese chestnut wormhole area; after the laser detection sub-module 20 completes the detection, the detected laser reflection signal intensity and position height information are processed and then sent to the electric control module 9 for analysis.

The acoustic detection module 6 comprises a second X-direction stepping motor 21, a second X-direction movement screw rod 22, a second X-direction bracket 23, a second Y-direction stepping motor 24, a second Y-direction movement screw rod 25, a second Y-direction guide rail 26, a sliding frame, a C-axis movement motor 27, a C-axis speed reducer 28, a high-pressure air needle 29, a high-pressure air pipe 30, an acoustic sensor 31 and a first electromagnetic valve 51; the X direction is a left-right horizontal direction, the Y direction is a front-back horizontal direction, a second support guide rail 56 which is arranged along the X direction and is in a rectangular frame is respectively arranged on the front side and the back side of the horizontal conveying belt 4 on the rack 1, a second X-direction movement screw rod 22 is arranged in each second support guide rail 56, one end of each second X-direction movement screw rod 22 is rotatably connected to the second support guide rail 56, the other end of each second X-direction movement screw rod 22 is in transmission connection with the second X-direction stepping motor 21, the second X-direction movement screw rods 22 are in threaded connection with the second X-direction supports 23 which slide along the second support guide rails 56, the front end and the back end of each second Y-direction guide rail 26 are fixedly connected to the upper ends of the two second X-direction supports 23, the second Y-direction guide rails 26 are in a hollow rod-shaped structure, the second Y-direction movement screw rods 25 are arranged in the second Y-direction guide rails, one end of a second Y-direction motion screw 25 is rotatably connected with one end inside a second Y-direction guide rail 26, the other end of the second Y-direction motion screw 25 is in transmission connection with a second Y-direction stepping motor 24, a C-axis motion motor 27 and a C-axis speed reducer 28 are arranged inside a sliding frame, the sliding frame is in threaded connection with the second Y-direction motion screw 25 and is arranged along the second Y-direction guide rail 26 in a sliding manner, a power output end of the C-axis motion motor 27 is in transmission connection with a power input end of the C-axis speed reducer 28, a high-pressure air needle 29 is arranged on a power output shaft of the C-axis speed reducer 28, the power output shaft of the C-axis speed reducer 28 is horizontally arranged along the left-right direction, the high-pressure air needle 29 is vertical to the power output shaft of the C-axis speed reducer 28, an air inlet of the high-pressure air needle 29 is connected with an air outlet of, the acoustic sensor 31 is arranged at the air jet port of the high-pressure air needle 29;

the concrete process of step five is: after the horizontal conveying belt 4 drives the Chinese chestnuts to move rightwards to the position below the acoustic detection module 6, the second X-direction stepping motor 21 drives the second X-direction movement screw rod 22 to rotate, so that the second X-direction support 23 of the second X-direction movement screw rod 22 moves leftwards or rightwards along the second support guide rail 56, and the second Y-direction guide rail 26, the sliding frame, the C-axis movement motor 27, the C-axis speed reducer 28 and the high-pressure air needle 29 also move leftwards or rightwards; the second Y-direction stepping motor 24 drives the second Y-direction moving screw 25 to rotate, so that the carriage with the second Y-direction moving screw 25 moves forward or backward along the second Y-direction rail 26, and the C-axis moving motor 27, the C-axis reducer 28 and the high-pressure air needle 29 also move forward or backward; moving the C-axis speed reducer 28 to the position above the Chinese chestnut to be detected through X-Y plane positioning, and aligning the high-pressure air needle 29 to be vertically above suspected Chinese chestnut eyes through the rotary motion of the C-axis speed reducer 28; after the high-pressure air needle 29 sprays the air flow, the acoustic sensor 31 collects the air flow echo signal at the suspected Chinese chestnut eye position, and the air flow echo signal is sent to the electronic control module 9 for analysis.

The sorting device 7 comprises a sorting support plate 32, a sorting cylinder 33, a sorting rod 34, a sorting rod head 35 and a slideway 36; the slideway 36 is installed at the right front side of the frame 1 through bolt connection; the sorting support plate 32 is vertically installed at the right rear side of the rack 1 through bolt connection, the sorting cylinder 33 is fixedly installed at the left side surface of the sorting support plate 32, the telescopic rod of the sorting cylinder 33 is coaxially connected with the sorting rod 34, and the sorting rod 34 is horizontally arranged along the left and right directions;

the specific process of the step (c) is as follows: when the chestnuts with the wormholes are conveyed to the right to the positions corresponding to the front and the back of the sorting device 7, the sorting cylinder 33 drives the sorting rod 34 to move, and the sorting rod head 35 is pushed to push the chestnuts with the wormholes out of the inlet of the slideway 36; the sorting cylinder 33 is connected with the electric control module 9, and under the control of the electric control module 9, the sorting rod 34 is pushed to move, so that the Chinese chestnuts with the defect of wormholes fall into defective frames through the slide way 36, and the qualified Chinese chestnuts are transported to the right side direction by the horizontal conveyor belt 4.

The electric control module 9 comprises an embedded ARM singlechip, an input/output submodule, an industrial Ethernet communication submodule, a display submodule and a light warning submodule; the embedded ARM single chip microcomputer is connected with the input/output submodule through an on-chip bus;

the embedded ARM single chip microcomputer is respectively connected with the industrial camera 12 and the laser detection sub-module 20 through an industrial Ethernet communication sub-module; the industrial Ethernet communication submodule is internally provided with a duplex industrial Ethernet port, the input/output submodule is respectively connected with the first X-direction stepping motor 13, the first Y-direction stepping motor 16, the second X-direction stepping motor 21, the C-axis motion motor 27, the second Y-direction stepping motor 24 and the sorting cylinder 33 through control signal lines, and the positioning and sorting are realized according to the ordered actions of the stepping motors corresponding to the control flow control; the input/output sub-module is also connected with a first electromagnetic valve 51, a second electromagnetic valve 52 and a third electromagnetic valve 53, and the electromagnetic valves are opened or closed according to the action flow to realize the functions of detection, freezing and unfreezing respectively; the input/output sub-module is also connected with and controls the opening and closing of the annular LED lamp 11 and the brightness of the annular LED lamp; the input/output submodule is also connected to the acoustic sensor 31.

In the step ④, the laser detection sub-module 20 performs a specific detection process that the laser detection sub-module 20 scans suspected chestnut eye areas at intervals of 2mm in the X direction and the Y direction of the X-Y plane positioning, the scanning length and the scanning width are 14mm, the total number of the scanning times is 8 × 8=64, and the laser detection sub-module 20 obtains a laser reflection intensity data set B₆₄And height dataset H₆₄The height is the distance from a laser emission head in the laser detection submodule to the surface of the Chinese chestnut to be detected, and is as follows:

further, for the laser reflection intensity data set B₆₄Calculating partial derivatives along X direction and Y direction, wherein the partial derivatives are BX₄₉And BY₄₉For height data set H₆₄Calculating partial derivatives along X direction and Y direction, respectively HX₄₉And HY₄₉As follows:

to BX₄₉And BY₄₉The mean of 49 data was obtained as

And

for HX₄₉And HY₄₉The mean of 49 data was obtained as

And

(ii) a To BX₄₉49 data in (1) and

comparing, if the difference is larger than the threshold value of 0.15/mm, the threshold value is relative brightness, the range is 0-1, the zone bit BXF is set to be 1, if the difference is smaller than the threshold value of 0.15/mm, the zone bit BXF is set to be 0, and similarly, for BY, comparing₄₉49 data in (1) and

comparing, if the difference is larger than the threshold value of 0.15/mm, setting the flag bit BYF to be 1, and if the difference is smaller than the threshold value of 0.15/mm, setting the flag bit BYF to be 0; for HX₄₉49 of (a) are respectively connected with

Comparing, if the difference is larger than the threshold value 1.8, setting the flag HXF to 1, if the difference is smaller than the threshold value 1.8, setting the flag HXF to 0, and similarly, for HY₄₉49 data in (1) and

making a comparison, if the difference isGreater than the threshold value of 1.8, the flag bits HYF are set to 1, and if both are less than the threshold value of 1.8, the flag bits HYF are set to 0; counting the number of 1 in the flag bits BXF, BYF, HXF and HYF, if the total number is 3 or 4, judging that the Chinese chestnut to be detected has the defect of the wormhole by the laser detection module 5, otherwise, judging that the Chinese chestnut to be detected does not have the defect of the wormhole.

The specific process of the matching detection of the acoustic detection module 6 and the electric control module 9 in the step ⑤ is that the acoustic detection module 6 moves to the upper part of the Chinese chestnut to be detected through the X-Y plane positioning under the control of the electric control module 9, the acoustic signal data of three positions are collected through the movement of the C-axis speed reducer 28, the position 1 is vertically over the suspected Chinese chestnut eye area, the position 2 is the vertical direction of the surface of the suspected Chinese chestnut eye area, and the acoustic signal data pass through the HX₄₉And HY₄₉Data are calculated, and the position 3 is a symmetrical position of the position 1 relative to a perpendicular line of the surface of the suspected Chinese chestnut moth eye area; the electronic control module 9 opens the first electromagnetic valve 51, high-pressure gas acts on the suspected Chinese chestnut eye area through the high-pressure gas needle 29, the acoustic sensor 31 sends data to the electronic control module 9 through the input/output submodule after acquiring the data for 2 seconds, and the electronic control module 9 closes the first electromagnetic valve 51 after success; similarly, the electric control module 9 collects acoustic signal data from the position 2 and the position 3; after the acoustic signal data acquisition of all three positions is completed, the embedded ARM single chip microcomputer of the electric control module 9 processes according to the following procedures:

A. respectively processing acoustic signal data acquired from the position 1, the position 2 and the position 3 according to the processes B-D;

B. 2-second acoustic signal data acquired at each position are subjected to Fourier transform to transform time domain signal data into frequency domain signal data;

C. compressing the image of the frequency domain signal data into a 256-level gray scale image of 128 multiplied by 256, wherein the higher the frequency is, the larger the corresponding gray scale is; when the frequency is 20Hz, the corresponding gray scale is 0, when the frequency is 20KHz, the corresponding gray scale is 255, and the gray scale calculation corresponding to the middle frequency is solved by linear interpolation; further, when the frequency is lower than 20Hz, the corresponding gray scale is 0, and when the frequency exceeds 20KHz, the corresponding gray scale is 255;

D. sending the 256-level gray level image of 128 × 256 to a residual error network for identification, wherein the identification result is a two-dimensional vector [ P ]₀,P₁]When P is₀Greater than P₁When the Chinese chestnut is detected, the acoustic detection module marks the current position as the Chinese chestnut to be detected has the defect of wormhole, otherwise, the acoustic detection module judges that the Chinese chestnut to be detected does not have the defect of wormhole;

E. summarizing and analyzing the primary results of the position 1, the position 2 and the position 3, if any one of the position 1, the position 2 and the position 3 has a wormhole defect mark, judging that the current Chinese chestnut to be detected has the wormhole defect by the acoustic detection module, otherwise, judging that the current Chinese chestnut to be detected does not have the wormhole defect;

the final identification result after the combined detection of the shooting vision, the laser and the acoustics is displayed by a display sub-module; the chestnuts with the wormholes are also displayed in an alarm mode through the lamplight warning submodule to remind workers; the display sub-module can also carry out statistical analysis on the detected data, and display the statistical data according to hours or batches, so that the data can be conveniently checked manually.

The residual error network comprises 3 residual error blocks, wherein an A1 convolutional layer, an A2Batch Norm layer, an A3 active layer, a B1 convolutional layer, a B2 Batch Norm layer and a B3 active layer form a first residual error block, the A1 convolutional layer, the A2Batch Norm layer, the A3 active layer, a B1 convolutional layer, the B2 Batch Norm layer and the B3 active layer are sequentially connected end to end, and further, data of an inlet of the A1 convolutional layer can be inserted between the B2 Batch Norm layer and the B3 active layer in a short circuit mode; the C1 convolution layer, the C2 Batch Norm layer, the C3 active layer, the D1 convolution layer, the D2Batch Norm layer and the D3 active layer form a second residual block, the C1 convolution layer, the C2 Batch Norm layer, the C3 active layer, the D1 convolution layer, the D2Batch Norm layer and the D3 active layer are sequentially connected end to end, and further, data of an inlet of the C1 convolution layer can be inserted between the D2Batch Norm layer and the D3 active layer in a short circuit mode; the E1 convolution layer, the E2 BatchNorm layer, the E3 active layer, the F1 convolution layer, the F2Batch Norm layer and the F3 active layer form a third residual block, the E1 convolution layer, the E2 Batch Norm layer, the E3 active layer, the F1 convolution layer, the F2Batch Norm layer and the F3 active layer are sequentially connected end to end, and further, data at the inlet of the E1 convolution layer can be inserted between the F2Batch Norm layer and the F3 active layer in a short circuit mode; and the first residual block, the second residual block and the third residual block are connected in sequence in an ending mode, the frequency domain gray scale image enters a G1 full connection layer after being input into the 3 residual blocks of the residual network, data flow into an H1 soft regression layer, and finally an identification result is output.

Furthermore, the residual error network is trained offline, and data of 3160 samples are collected for training in total, wherein 1510 data are data without the bug defect, and 1650 data are data with the bug defect, and the trained model is stored in the embedded ARM single chip microcomputer of the electronic control module 9.

Obtaining the area information of suspected insect eyes according to the machine vision detection result; detecting suspected insect eye position information by machine vision, and obtaining a laser detection result through laser detection; detecting the position information of the suspected insect eye and the curved surface information near the suspected insect eye of the chestnut by using a machine vision detector, and obtaining an acoustic detection result through acoustic detection; the machine vision detection result: the suspected insect eye position and size information, the laser detection result and the acoustic detection result are identified through an SVM model, and the final identification result is output: and normal and bug eyes are output by the display submodule and the lamplight warning submodule.

The SVM model is trained offline, the same training samples are shared with the residual error network, namely data of 3160 samples are collected in total, wherein 1510 data are data without wormhole defects, 1650 data are data with wormhole defects, and trained SVM model parameters are stored in the embedded ARM single chip microcomputer of the electronic control module 9.

By adopting the technical scheme, the invention can also be set as cascade operation, and two devices and one robot with machine vision form a set of detection unit; the machine hand with the machine vision is arranged between the two devices, the machine vision is utilized to carry out auxiliary positioning, the Chinese chestnuts to be detected are moved from one device to the other device, turning of the Chinese chestnuts is realized in the moving process, and then the detection is carried out, so that the detection accuracy can be further improved.

In summary, compared with the prior art, the invention has the following beneficial effects:

according to the invention, under the multi-stage combined intelligent information fusion of machine vision, acoustic detection and laser detection, the high-precision Chinese chestnut wormhole detection can be realized, on one hand, the intelligent level of detection can be effectively improved, on the other hand, the sorting accuracy is improved and the device cost is reduced through the multi-stage combined intelligent detection. The method can be used for on-line real-time detection of the wormholes of the Chinese chestnut agricultural products, has important significance for improving the development of deep processing of the Chinese chestnut agricultural products in China, and has good market application prospect.

Drawings

FIG. 1 is a schematic perspective view of an intelligent detection device for chestnut moth eye in the present invention at a viewing angle;

FIG. 2 is a schematic perspective view of an intelligent detection device for chestnut moth eye in the present invention at another viewing angle;

FIG. 3 is a schematic top view of a laser detection module of the intelligent Chinese chestnut eye detection device according to the present invention;

FIG. 4 is a left side view of FIG. 3;

FIG. 5 is a schematic structural diagram of an acoustic detection module of the intelligent Chinese chestnut eye detection device in the invention;

FIG. 6 is a schematic structural diagram of a sorting device of the intelligent Chinese chestnut wormhole detection equipment in the invention;

FIG. 7 is a schematic bottom view of a camera darkroom detection module of the intelligent Chinese chestnut eye detection device according to the present invention;

FIG. 8 is a schematic diagram of machine vision chestnut location calculation;

FIG. 9 is a schematic diagram of the operation of the laser detection module;

FIG. 10 is a schematic diagram of the operation of the acoustic detection module;

FIG. 11 is a diagram of a residual error network (ResNet) architecture for acoustic signal data processing;

FIG. 12 is a flow chart of a multi-level joint intelligent detection of machine vision, laser and acoustics;

FIG. 13 is a structural diagram of an electric control module of the intelligent detection device for Chinese chestnut moth eye in the invention.

Detailed Description

The invention relates to an intelligent detection method for chestnut wormholes, which adopts intelligent detection equipment for chestnut wormholes to operate; as shown in fig. 1-6, the intelligent detection equipment for the wormhole of the Chinese chestnut comprises a frame 1, wherein a lifting conveyer belt 2 and a horizontal conveyer belt 4 are sequentially arranged on the frame 1 from left to right, the lifting conveyer belt 2 is arranged in a left-low-right high-inclination manner, a photographing darkroom detection module 3, a laser detection module 5, an acoustic detection module 6 and a sorting device 7 which are positioned above the horizontal conveyer belt 4 are sequentially arranged on the frame 1 from left to right, an electric control module 9, a Chinese chestnut locking device and a Chinese chestnut unlocking device are arranged on the side portion of the frame 1, the working portion of the Chinese chestnut locking device is positioned above the lifting conveyer belt 2, and the working portion of the Chinese chestnut unlocking device is positioned between the acoustic detection module 6 and the; the lifting conveying belt 2 and the horizontal conveying belt 4 are both made of metal heat conductor materials, grooves 54 for storing Chinese chestnuts to be conveyed are uniformly formed in the upper surfaces of the lifting conveying belt 2 and the horizontal conveying belt 4 at intervals of 5-8 cm in the left-right direction, and the lifting conveying belt 2 and the horizontal conveying belt 4 are combined into a complete conveying belt;

the detection method comprises the following steps:

the concrete process of step three is: the lifting conveying belt 2 and the horizontal conveying belt 4 are driven by a conveying stepping motor (conventional technology in the field), the width of each section of the conveying belt is 30-50 mm, the Chinese chestnuts to be detected are preliminarily positioned by the stepping motor, sent into the photographing darkroom detection module 3, the positions are further calculated by a machine vision algorithm of the electronic control module 9, and the stepping motor is driven to be precisely positioned, so that the Chinese chestnuts to be detected move to the right to the area below the industrial camera 12 in the photographing darkroom detection module 3, as shown in fig. 7 and 8, the positions of the Chinese chestnuts are calculated by a machine vision method of the electronic control module 9, the industrial camera 12 of the photographing darkroom detection module 3 is controlled by the electronic control module 9 to photograph the detection plate, the Chinese chestnuts are sent to the electronic control module 9 through an industrial Ethernet communication submodule, and are processed by an embedded ARM single chip microcomputer: by calibrating the industrial camera 12, the height H from the lens of the industrial camera 12 to the real object space is a known quantity, a monocular vision method is adopted, the included angle a can be calculated through a similar triangular proportion, L1 is the length of a pixel and can be obtained through an image processing algorithm (conventional technology in the field), so that the L2 numerical value in the real object space is calculated, and the coordinate position of the Chinese chestnut to be detected in a photographing darkroom is obtained; then, acquiring the size of the Chinese chestnut to be detected and the area information of suspected insect eyes through feature extraction and analysis of the image; the calculated position and size information of the Chinese chestnut to be detected provides positioning information for the subsequent laser detection module 5 and acoustic detection module 6.

As shown in fig. 13, the electronic control module 9 includes an embedded ARM single-chip microcomputer, an input/output sub-module, an industrial ethernet communication sub-module, a display sub-module, and a light warning sub-module; the embedded ARM single chip microcomputer is connected with the input/output submodule through an on-chip bus;

In the step ④, the specific detection process of the laser detection sub-module 20 includes, as shown in fig. 9, scanning the suspected chestnut eye area by the laser detection sub-module 20 at intervals of 2mm in the X-Y plane positioning direction and the Y direction, wherein the scanning length and width are 14mm, the total number of the scanning is 8 × 8=64 times, and the laser detection sub-module 20 obtains the laser reflection intensity data set B₆₄And height dataset H₆₄The height is the distance from a laser emission head in the laser detection submodule to the surface of the Chinese chestnut to be detected, and is as follows:

to BX₄₉And BY₄₉The mean of 49 data was obtained as

And

for HX₄₉And HY₄₉The mean of 49 data was obtained as

And

(ii) a To BX₄₉49 data in (1) and

comparing, if the difference is larger than the threshold value 1.8, setting the flag HYF to be 1, and if the difference is smaller than the threshold value 1.8, setting the flag HYF to be 0; counting the number of 1 in the flag bits BXF, BYF, HXF and HYF, if the total number is 3 or 4, judging that the Chinese chestnut to be detected currently has the defect of wormhole by the laser detection module 5, otherwise, judging that the Chinese chestnut to be detected currently has the defect of wormholeNo defect of wormhole.

The specific process of the matching detection of the acoustic detection module 6 and the electric control module 9 in the step ⑤ is that as shown in fig. 10, the acoustic detection module 6 moves to the upper part of the Chinese chestnut to be detected through the X-Y plane positioning under the control of the electric control module 9, the acoustic signal data of three positions are collected through the movement of the C-axis speed reducer 28, the position 1 is vertically over the suspected Chinese chestnut eye area, the position 2 is the vertical direction of the surface of the suspected Chinese chestnut eye area, and the acoustic signal data pass through the HX₄₉And HY₄₉Data are calculated, and the position 3 is a symmetrical position of the position 1 relative to a perpendicular line of the surface of the suspected Chinese chestnut moth eye area; the electronic control module 9 opens the first electromagnetic valve 51, high-pressure gas acts on the suspected Chinese chestnut eye area through the high-pressure gas needle 29, the acoustic sensor 31 sends data to the electronic control module 9 through the input/output submodule after acquiring the data for 2 seconds, and the electronic control module 9 closes the first electromagnetic valve 51 after success; similarly, the electric control module 9 collects acoustic signal data from the position 2 and the position 3; after the acoustic signal data acquisition of all three positions is completed, the embedded ARM single chip microcomputer of the electric control module 9 processes according to the following procedures:

D. sending the 256-level gray level images of 128 × 256 to a residual error network (Res-Net) for identification, wherein the identification result is a two-dimensional vector [ P ]₀,P₁]When P is₀Greater than P₁Then, the acoustic detection module marks the current position as the existence of wormhole defect of the Chinese chestnut to be detectedOtherwise, judging that the Chinese chestnut to be detected does not have the defect of wormhole at present;

As shown in fig. 11, the residual network (Res-Net) includes 3 residual blocks, where the a1 convolutional layer, the A2Batch Norm layer, the A3 active layer, the B1 convolutional layer, the B2 Batch Norm layer, and the B3 active layer constitute a first residual block, the a1 convolutional layer, the A2Batch Norm layer, the A3 active layer, the B1 convolutional layer, the B2 Batch Norm layer, and the B3 active layer are sequentially connected end to end, and further, data of the a1 convolutional layer inlet may be inserted between the B2 Batch Norm layer and the B3 active layer through a short circuit (shortcut); a C1 convolution layer, a C2 Batch Norm layer, a C3 active layer, a D1 convolution layer, a D2Batch Norm layer and a D3 active layer form a second residual block, the C1 convolution layer, the C2 Batch Norm layer, the C3 active layer, the D1 convolution layer, the D2Batch Norm layer and the D3 active layer are sequentially connected end to end, and further, data short circuit (shortcut) of an inlet of the C1 convolution layer is inserted between the D2Batch Norm layer and the D3 active layer; the E1 convolution layer, the E2 Batch Norm layer, the E3 active layer, the F1 convolution layer, the F2Batch Norm layer and the F3 active layer form a third residual block, the E1 convolution layer, the E2 Batch Norm layer, the E3 active layer, the F1 convolution layer, the F2Batch Norm layer and the F3 active layer are sequentially connected end to end, and further, data at the inlet of the E1 convolution layer can be inserted between the F2Batch Norm layer and the F3 active layer in a short circuit (short cut); and the first residual block, the second residual block and the third residual block are connected in sequence in an ending way, the frequency domain gray scale image enters a G1 full connection layer after being input into 3 residual blocks of the residual network (Res-Net), the data flows into an H1 soft regression layer, and finally the identification result is output.

Furthermore, the residual error network (Res-Net) is trained offline, and data of 3160 samples are collected for training in total, wherein 1510 data are data without wormhole defects, and 1650 data are data with wormhole defects, and the trained model is stored in the embedded ARM single chip microcomputer of the electronic control module 9.

As shown in fig. 12, the area information of suspected moth eyes is obtained from the machine vision detection result; detecting suspected insect eye position information by machine vision, and obtaining a laser detection result through laser detection; detecting the position information of the suspected insect eye and the curved surface information near the suspected insect eye of the chestnut by using a machine vision detector, and obtaining an acoustic detection result through acoustic detection; the machine vision detection result (suspected insect eye position and size information), the laser detection result and the acoustic detection result are identified through an SVM model, the final identification result (normal and insect eyes) is output, and then the final identification result is output through a display sub-module and a light warning sub-module.

The SVM model is trained offline, the training samples are the same as those of the residual error network (Res-Net), namely data of 3160 samples collected in total are trained (wherein 1510 are data without the bug defect and 1650 are data with the bug defect), and the trained SVM model parameters are stored in the embedded ARM single chip microcomputer of the electronic control module 9.

The present embodiment is not intended to limit the shape, material, structure, etc. of the present invention in any way, and any simple modification, equivalent change and modification made to the above embodiments according to the technical spirit of the present invention are within the scope of the technical solution of the present invention.

Claims

1. An intelligent detection method for Chinese chestnut wormholes is characterized by comprising the following steps: the detection method adopts intelligent detection equipment for the wormhole of the Chinese chestnut to operate; the intelligent detection equipment for the Chinese chestnut moth eye comprises a rack (1), wherein a lifting conveyer belt (2) and a horizontal conveyer belt (4) are sequentially arranged on the rack (1) from left to right, the lifting conveyer belt (2) is arranged in a left-low-right high-inclination manner, a photographing darkroom detection module (3), a laser detection module (5), an acoustic detection module (6) and a sorting device (7) which are positioned above the horizontal conveyer belt (4) are sequentially arranged on the rack (1) from left to right, an electric control module (9), a Chinese chestnut locking device and a Chinese chestnut unlocking device are arranged on the lateral part of the rack (1), the working part of the Chinese chestnut locking device is positioned above the lifting conveyer belt (2), and the working part of the Chinese chestnut unlocking device is positioned between the acoustic detection module (6) and the sorting device (7) which are positioned; the lifting conveying belt (2) and the horizontal conveying belt (4) are both made of metal heat conductor materials, grooves (54) for storing Chinese chestnuts to be conveyed are uniformly formed in the upper surfaces of the lifting conveying belt (2) and the horizontal conveying belt (4) at intervals of 5-8 cm in the left-right direction, and the lifting conveying belt (2) and the horizontal conveying belt (4) are combined into a complete conveying belt;

the detection method comprises the following steps:

firstly, the lifting conveying belt (2) and the horizontal conveying belt (4) are driven by a stepping motor to move from left to right, Chinese chestnuts are placed into a groove (54) on the left side of the top of the lifting conveying belt (2) one by one, and the lifting conveying belt (2) and the horizontal conveying belt (4) drive the Chinese chestnuts to move right;

thirdly, the lifting conveyer belt (2) and the horizontal conveyer belt (4) drive the Chinese chestnuts to move rightwards to the photographing darkroom detection module (3) for visual detection;

fourthly, the lifting conveying belt (2) and the horizontal conveying belt (4) drive the Chinese chestnuts to move rightwards to the lower part of the laser detection module (5) for laser detection;

fifthly, the lifting conveyer belt (2) and the horizontal conveyer belt (4) drive the Chinese chestnuts to move rightwards to the position below the acoustic detection module (6) for acoustic detection;

seventhly, the sorting device sorts the Chinese chestnuts which are determined to have the wormholes, and the intact and qualified Chinese chestnuts are conveyed rightwards by a horizontal conveying belt (4);

the Chinese chestnut locking device is a cold air spray pipe (37) and a water dropping pipe which are arranged above the lifting conveying belt (2) side by side, the water dropping pipe is positioned on the left side of the cold air spray pipe (37), and a water dropping hole of the water dropping pipe and a nozzle of the cold air spray pipe (37) are both arranged downwards and face towards a passing groove (54); a second electromagnetic valve (52) is arranged on the cold air spray pipe (37);

the Chinese chestnut unlocking device is a hot air spray pipe (38) arranged above the horizontal conveying belt (4), a spray opening of the hot air spray pipe (38) is arranged downwards and faces towards a passing groove (54), a third electromagnetic valve (53) is arranged on the hot air spray pipe (38), and the second electromagnetic valve (52) and the third electromagnetic valve (53) are connected with the electric control module (9);

the concrete process of the step II is as follows: the water in the water dropping pipe is dropped into the groove (54) right below, the Chinese chestnut is placed in the groove with liquid water, after the Chinese chestnut moves to the lower part of the nozzle of the cold air spray pipe (37), the electric control module (9) controls the second electromagnetic valve (52) to be opened, the high-pressure cold air is sprayed to the groove (54) with the liquid water, and the liquid water is frozen and fixed in the groove (54) due to the cold freezing;

the concrete process comprises the following steps: when the detected chestnuts are conveyed to the right below the nozzles of the hot air spray pipes (38), the solid ice in the grooves (54) of the horizontal conveying belt (4) of the metal heat conductor is rapidly heated through hot air to be converted into liquid water, the chestnuts are unlocked, and the sorting device (7) is convenient to sort.

2. The intelligent detection method for chestnut moth eyes according to claim 1, characterized in that: the camera darkroom detection module (3) comprises a cavity box body (10), an annular LED lamp (11) and an industrial camera (12), the bottom of the cavity box body (10) is open and vertically faces downwards to the horizontal conveying belt (4), the left side and the right side of the cavity box body (10) are fixed on the rack (1) through screws, and horizontal conveying rectangular gaps convenient for Chinese chestnuts are reserved on the left side and the right side of the edge of the bottom of the cavity box body (10); the annular LED lamp (11) and the industrial camera (12) are arranged at the top center position in the cavity box body (10);

the concrete process of step three is: promote conveyer belt (2) and horizontal transport area (4) and be driven by carrying step motor, every section width of conveyer belt is 30~50mm, by step motor preliminary location, will wait to detect the chinese chestnut and send into inside camera darkroom detection module (3) of shooing, the machine vision algorithm of rethread electric control module (9) further calculates the position, drive step motor precision positioning, make it moves to wait to detect the chinese chestnut and be located industry camera (12) below region in camera darkroom detection module (3) to move to the camera darkroom detection module (3), calculate the chinese chestnut position through the machine vision method of electric control module (9), industrial camera (12) of camera darkroom detection module (3) are under the control of electric control module (9), treat the pick-up plate and shoot the chinese chestnut, send to electric control module (9) through industry ethernet communication submodule, handle through embedded ARM singlechip, the flow is: the industrial camera (12) is calibrated, the height H from the lens of the industrial camera (12) to the real object space is a known quantity, a monocular vision method is adopted, the included angle a can be calculated through a similar triangle proportion, L1 is the length of a pixel and can be obtained through an image processing algorithm, so that the L2 numerical value in the real object space is calculated, and the coordinate position of the Chinese chestnut to be detected in a photographing darkroom is obtained; then, acquiring the size of the Chinese chestnut to be detected and the area information of suspected insect eyes through feature extraction and analysis of the image; and the calculated position and size information of the Chinese chestnut to be detected provides positioning information for a subsequent laser detection module (5) and an acoustic detection module (6).

3. The intelligent detection method for chestnut moth eyes according to claim 2, characterized in that: the laser detection module (5) comprises a first X-direction stepping motor (13), a first X-direction moving screw rod (14), a first X-direction support (15), a first Y-direction stepping motor (16), a first Y-direction moving screw rod (17), a first Y-direction guide rail (18), a laser detection clamp (19) and a laser detection sub-module (20); the X direction is a left and right horizontal direction, the Y direction is a front and back horizontal direction, a first support guide rail (55) which is arranged along the X direction and is a rectangular frame is respectively arranged on the front side and the back side of the horizontal conveying belt (4) on the rack (1), a first X direction movement screw rod (14) is arranged in each first support guide rail (55), one end of each first X direction movement screw rod (14) is rotatably connected onto the first support guide rail (55), the other end of each first X direction movement screw rod (14) is in transmission connection with the first X direction stepping motor (13), the first X direction movement screw rods (14) are in threaded connection with the first X direction supports (15) which slide along the first support guide rails (55), the front end and the back end of each first Y direction guide rail (18) are fixedly connected onto the upper ends of the two first X direction supports (15), and the first Y direction guide rails (18) are of a hollow rod-shaped structure, the first Y-direction motion screw rod (17) is arranged in the first Y-direction guide rail (18) along the front-back direction, one end of the first Y-direction motion screw rod (17) is rotatably connected to one end in the first Y-direction guide rail (18), the other end of the first Y-direction motion screw rod (17) is in transmission connection with the first Y-direction stepping motor (16), the laser detection clamp (19) is in threaded connection with the first Y-direction motion screw rod (17) and is arranged in a sliding mode along the first Y-direction guide rail (18), and the laser detection sub-module (20) is installed on the laser detection clamp (19) and irradiates towards the groove (54);

the specific process of the step IV is as follows: the horizontal conveying belt (4) drives the Chinese chestnuts to move rightwards to the position below the laser detection module (5), the first X-direction stepping motor (13) drives the first X-direction moving screw rod (14) to rotate, so that the first X-direction support (15) of the first X-direction moving screw rod (14) moves leftwards or rightwards along the first support guide rail (55), and the first Y-direction guide rail (18), the laser detection clamp (19) and the laser detection sub-module (20) also move leftwards or rightwards; a first Y-direction stepping motor (16) drives a first Y-direction moving screw rod (17) to rotate, so that a laser detection clamp (19) of the first Y-direction moving screw rod (17) moves forwards or backwards along a first Y-direction guide rail (18), and a laser detection sub-module (20) also moves forwards or backwards; moving a laser detection clamp (19) to the position above the Chinese chestnut to be detected through X-Y plane positioning, and aligning a laser detection sub-module (20) to the position vertically above a suspected Chinese chestnut wormhole area; after the laser detection sub-module (20) finishes detection, the detected laser reflection signal intensity and position height information are processed and then are sent to the electric control module (9) for analysis.

4. The intelligent detection method for chestnut worms eyes according to claim 3, characterized in that: the acoustic detection module (6) comprises a second X-direction stepping motor (21), a second X-direction movement screw rod (22), a second X-direction support (23), a second Y-direction stepping motor (24), a second Y-direction movement screw rod (25), a second Y-direction guide rail (26), a sliding frame, a C-axis movement motor (27), a C-axis speed reducer (28), a high-pressure air needle (29), a high-pressure air pipe (30), an acoustic sensor (31) and a first electromagnetic valve (51); the X direction is a left-right horizontal direction, the Y direction is a front-back horizontal direction, a second support guide rail (56) which is arranged along the X direction and is a rectangular frame is respectively arranged on the front side and the back side of the horizontal conveying belt (4) on the rack (1), a second X direction movement screw rod (22) is arranged in each second support guide rail (56), one end of each second X direction movement screw rod (22) is rotatably connected to the second support guide rail (56), the other end of each second X direction movement screw rod (22) is in transmission connection with a second X direction stepping motor (21), the second X direction movement screw rods (22) are in threaded connection with second X direction supports (23) which slide along the second support guide rails (56), the front end and the back end of each second Y direction guide rail (26) are fixedly connected to the upper ends of the two second X direction supports (23), and the second Y direction guide rails (26) are of a hollow rod-shaped structure, a second Y-direction motion screw rod (25) is arranged in a second Y-direction guide rail (26) along the front-back direction, one end of the second Y-direction motion screw rod (25) is rotatably connected with one end in the second Y-direction guide rail (26), the other end of the second Y-direction motion screw rod (25) is in transmission connection with a second Y-direction stepping motor (24), a C-axis motion motor (27) and a C-axis speed reducer (28) are arranged in a sliding frame, the sliding frame is in threaded connection with the second Y-direction motion screw rod (25) and is arranged in a sliding way along the second Y-direction guide rail (26), the power output end of the C-axis motion motor (27) is in transmission connection with the power input end of the C-axis speed reducer (28), a high-pressure air needle (29) is arranged on the power output shaft of the C-axis speed reducer (28), the power output shaft of the C-axis speed reducer (28) is horizontally arranged along the left-right direction, and the high-pressure air needle, an air inlet of the high-pressure air needle (29) is connected with an air outlet of the high-pressure air pipe (30), a first electromagnetic valve (51) is arranged on the high-pressure air pipe (30), an air compressor is connected with the air inlet of the high-pressure air pipe (30), and an acoustic sensor (31) is arranged at an air jet of the high-pressure air needle (29);

the concrete process of step five is: after the horizontal conveying belt (4) drives the chestnuts to move rightwards to the position below the acoustic detection module (6), a second X-direction stepping motor (21) drives a second X-direction movement screw rod (22) to rotate, so that a second X-direction support (23) of the second X-direction movement screw rod (22) moves leftwards or rightwards along a second support guide rail (56), and a second Y-direction guide rail (26), a sliding frame, a C-axis movement motor (27), a C-axis speed reducer (28) and a high-pressure air needle (29) also move leftwards or rightwards; a second Y-direction stepping motor (24) drives a second Y-direction moving screw rod (25) to rotate, so that a sliding frame of the second Y-direction moving screw rod (25) moves forwards or backwards along a second Y-direction guide rail (26), and a C-axis moving motor (27), a C-axis speed reducer (28) and a high-pressure air needle (29) also move forwards or backwards; moving a C-axis speed reducer (28) to the position above the Chinese chestnut to be detected through X-Y plane positioning, and aligning a high-pressure air needle (29) to the position vertically above suspected Chinese chestnut eyes through the rotary motion of the C-axis speed reducer (28); after the high-pressure air needle (29) sprays air flow, an acoustic sensor (31) collects an air flow echo signal at a suspected Chinese chestnut eye position, and the air flow echo signal is sent to an electric control module (9) for analysis.

5. The intelligent detection method for chestnut moth eyes according to claim 4, characterized in that: the sorting device (7) comprises a sorting support plate (32), a sorting cylinder (33), a sorting rod (34), a sorting rod head (35) and a slideway (36); the slide way (36) is installed at the right front side of the rack (1) through bolt connection; the sorting support plate (32) is vertically arranged at the right rear side of the rack (1) through bolt connection, the sorting cylinder (33) is fixedly arranged at the left side surface of the sorting support plate (32), a telescopic rod of the sorting cylinder (33) is coaxially connected with the sorting rod (34), and the sorting rod (34) is horizontally arranged along the left and right directions;

the specific process of the step (c) is as follows: when the chestnuts with the wormholes are conveyed to the positions corresponding to the front and the back of the sorting device (7) rightwards, the sorting cylinder (33) drives the sorting rod (34) to move, and the sorting rod head (35) is pushed to push the chestnuts with the wormholes out of the inlet of the slide way (36); the sorting cylinder (33) is connected with the electric control module (9), and the sorting rod (34) is pushed to move under the control of the electric control module (9), so that the Chinese chestnuts with the defect of wormholes fall into the defective goods frame through the slide way (36), and the qualified Chinese chestnuts are transported to the right side direction by the horizontal conveyor belt (4).

6. The intelligent detection method for chestnut moth eyes according to claim 5, characterized in that: the electric control module (9) comprises an embedded ARM singlechip, an input/output submodule, an industrial Ethernet communication submodule, a display submodule and a light warning submodule; the embedded ARM single chip microcomputer is connected with the input/output submodule through an on-chip bus;

the embedded ARM single chip microcomputer is respectively connected with an industrial camera (12) and a laser detection sub-module (20) through an industrial Ethernet communication sub-module; the industrial Ethernet communication submodule is internally provided with a duplex industrial Ethernet port, the input/output submodule is respectively connected with a first X-direction stepping motor (13), a first Y-direction stepping motor (16), a second X-direction stepping motor (21), a C-axis motion motor (27), a second Y-direction stepping motor (24) and a sorting cylinder (33) through control signal lines, and the positioning and sorting are realized according to the ordered actions of the stepping motors corresponding to the control flow control; the input/output sub-module is also connected with a first electromagnetic valve (51), a second electromagnetic valve (52) and a third electromagnetic valve (53), and the electromagnetic valves are opened or closed according to the action flow to realize the functions of detection, freezing and unfreezing respectively; the input/output sub-module is also connected with and controls the opening and closing of the annular LED lamp (11) and the brightness of the annular LED lamp; the input/output submodule is also connected to an acoustic sensor (31).

7. The method of claim 6The intelligent detection method for the chestnut moth eye is characterized in that in the step ④, the specific detection process of the laser detection submodule (20) is that the laser detection submodule (20) scans suspected chestnut moth eye areas at intervals of 2mm in the X direction and the Y direction of X-Y plane positioning, the scanning length and the scanning width are 14mm, the total number is 8 × 8=64 times, and the laser detection submodule (20) acquires a laser reflection intensity data set B₆₄And height dataset H₆₄The height is the distance from a laser emission head in the laser detection submodule to the surface of the Chinese chestnut to be detected, and is as follows:

to BX₄₉And BY₄₉The mean of 49 data was obtained as

And

for HX₄₉And HY₄₉The mean of 49 data was obtained as

And

(ii) a To BX₄₉49 data in (1) and

comparing, if the difference is larger than the threshold value 1.8, setting the flag HYF to be 1, and if the difference is smaller than the threshold value 1.8, setting the flag HYF to be 0; counting the number of 1 in the flag bits BXF, BYF, HXF and HYF, if the total number is 3 or 4, judging that the Chinese chestnut to be detected has the defect of wormhole by the laser detection module (5), otherwise, judging that the Chinese chestnut to be detected does not have wormhole defectIn the case of wormhole defects.

8. The intelligent detection method for chestnut moth eyes according to claim 7, characterized in that the specific process of the cooperation detection of the acoustic detection module (6) and the electronic control module (9) in step ⑤ is that the acoustic detection module (6) moves to the upper side of the chestnut to be detected through the X-Y plane positioning under the control of the electronic control module (9), the acoustic signal data of three positions are collected through the movement of a C-axis reducer (28), the position 1 is vertically above a suspected chestnut moth eye area, the position 2 is in the vertical direction of the surface of the suspected chestnut moth eye area, and the acoustic signal data pass through the HX₄₉And HY₄₉Data are calculated, and the position 3 is a symmetrical position of the position 1 relative to a perpendicular line of the surface of the suspected Chinese chestnut moth eye area; the electric control module (9) opens a first electromagnetic valve (51), high-pressure gas acts on a suspected Chinese chestnut eye area through a high-pressure gas needle (29), after 2-second data are collected by the acoustic sensor (31), the data are sent to the electric control module (9) through the input/output sub-module, and after the data are successfully collected, the electric control module (9) closes the first electromagnetic valve (51); similarly, the electric control module (9) collects acoustic signal data of the position 2 and the position 3; after the acoustic signal data of all three positions are acquired, the embedded ARM single chip microcomputer of the electric control module (9) processes according to the following procedures:

D. feeding 256-level gray level images of 128 × 256 into a residual error network for distinguishingThe recognition result is a two-dimensional vector [ P ]₀,P₁]When P is₀Greater than P₁When the Chinese chestnut is detected, the acoustic detection module marks the current position as the Chinese chestnut to be detected has the defect of wormhole, otherwise, the acoustic detection module judges that the Chinese chestnut to be detected does not have the defect of wormhole;

9. The intelligent detection method for chestnut moth eyes according to claim 8, characterized in that: the residual error network comprises 3 residual error blocks, wherein an A1 convolutional layer, an A2Batch Norm layer, an A3 active layer, a B1 convolutional layer, a B2 Batch Norm layer and a B3 active layer form a first residual error block, the A1 convolutional layer, the A2Batch Norm layer, the A3 active layer, a B1 convolutional layer, the B2 Batch Norm layer and the B3 active layer are sequentially connected end to end, and further, data of an inlet of the A1 convolutional layer can be inserted between the B2 Batch Norm layer and the B3 active layer in a short circuit mode; the C1 convolution layer, the C2 Batch Norm layer, the C3 active layer, the D1 convolution layer, the D2Batch Norm layer and the D3 active layer form a second residual block, the C1 convolution layer, the C2 Batch Norm layer, the C3 active layer, the D1 convolution layer, the D2Batch Norm layer and the D3 active layer are sequentially connected end to end, and further, data at the inlet of the C1 convolution layer can be inserted between the D2Batch Norm layer and the D3 active layer in a short circuit mode; the E1 convolution layer, the E2 Batch Norm layer, the E3 active layer, the F1 convolution layer, the F2Batch Norm layer and the F3 active layer form a third residual block, the E1 convolution layer, the E2 Batch Norm layer, the E3 active layer, the F1 convolution layer, the F2Batch Norm layer and the F3 active layer are sequentially connected end to end, and further, data at the inlet of the E1 convolution layer can be inserted between the F2Batch Norm layer and the F3 active layer in a short circuit mode; the first residual block, the second residual block and the third residual block are connected in sequence in an ending mode, after the frequency domain gray scale image is input into the 3 residual blocks of the residual network, the frequency domain gray scale image enters a G1 full connection layer, data flow into an H1 soft regression layer, and finally an identification result is output;

furthermore, the residual error network is trained offline, data of 3160 samples are collected for training in total, wherein 1510 data are data without the bug defect, 1650 data are data with the bug defect, and the trained model is stored in the embedded ARM single chip microcomputer of the electric control module (9);

obtaining the area information of suspected insect eyes according to the machine vision detection result; detecting suspected insect eye position information by machine vision, and obtaining a laser detection result through laser detection; detecting the position information of the suspected insect eye and the curved surface information near the suspected insect eye of the chestnut by using a machine vision detector, and obtaining an acoustic detection result through acoustic detection; the machine vision detection result: the suspected insect eye position and size information, the laser detection result and the acoustic detection result are identified through an SVM model, and the final identification result is output: normal and wormhole, and then output by the display sub-module and the light warning sub-module;

the SVM model is trained offline, the same training samples are shared with the residual error network, namely data of 3160 samples are collected in total, wherein 1510 data are data without wormhole defects, 1650 data are data with wormhole defects, and trained SVM model parameters are stored in the embedded ARM single chip microcomputer of the electronic control module (9).