Quality nondestructive testing device and method suitable for thick-peel fruits and vegetables

Abstract

The invention discloses a nondestructive quality detection device and method suitable for thick-peel fruits and vegetables, and relates to the technical field of fruit and vegetable detection. The device comprises a detection head module, a mechanical adaptation module, a data processing module and a display module. The detection head module is integrated with an annular array light source, a silicon-based multispectral chip acquisition unit and an optical adjustment unit, and can acquire visible light and near infrared spectrum at the same time; a telescopic detection arm and a rotating platform of the mechanical adaptation module adapt to the body types of the thick-peel fruits and vegetables; and the data processing module optimizes the detection data through skin interference deduction and multi-modal data fusion to realize quality judgment. The method comprises the steps of placement adjustment, light source light path adjustment, spectrum acquisition, data processing, quality judgment and display. The problem that the effective range of visible light collection is limited in thick-peel fruit and vegetable detection is solved, the quality of thick-peel fruits and vegetables such as watermelons and pumpkins can be accurately and efficiently detected, and good practicability and popularization value are achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of fruit and vegetable quality detection, and in particular to a non-destructive quality detection device and method suitable for thick-skinned fruits and vegetables. Background Art

[0002] In the quality inspection of fruits and vegetables, thick-skinned fruits and vegetables such as watermelon and pumpkin pose many difficulties for non-destructive testing due to their thick skin and large size. Existing non-destructive testing devices and methods for fruit quality often have the problem of limited effective range of visible light collection when testing such thick-skinned fruits and vegetables. Visible light has weak penetrating ability and most of it is reflected or absorbed by the skin of thick-skinned fruits and vegetables, making it difficult to obtain internal quality information. Moreover, due to the large size of thick-skinned fruits and vegetables, the light source illumination range and spectrum collection angle of traditional detection heads are fixed, which can easily lead to incomplete light coverage and weak signals in edge areas, affecting detection accuracy and efficiency. At the same time, relying solely on near-infrared spectroscopy cannot fully reflect skin defects and requires the cooperation of visible light, but the collection efficiency of both is difficult to optimize at the same time, resulting in poor quality detection results for thick-skinned fruits and vegetables. Summary of the Invention

[0003] The purpose of the present invention is to provide a nondestructive quality testing device and method suitable for thick-skinned fruits and vegetables, so as to solve the problems in the prior art of quality testing of thick-skinned fruits and vegetables, such as the limited effective range of visible light collection, poor applicability of the detection head, and difficulty in simultaneously optimizing the efficiency of visible light and near-infrared collection, so as to achieve accurate, efficient and nondestructive testing of thick-skinned fruits and vegetables such as watermelon and pumpkin.

[0004] The adopted technical solution: A nondestructive quality testing device suitable for thick-skinned fruits and vegetables, characterized by including: a detection head module, which integrates a light source unit, a silicon-based multispectral chip acquisition unit and an optical adjustment unit. The light source unit includes visible light LEDs and near-infrared LEDs distributed in a ring array. The silicon-based multispectral chip acquisition unit contains detection channels covering the visible light band (400-760nm) and the near-infrared band (760-2500nm).

[0005] The mechanical adaptation module includes a retractable detection arm and a rotating platform;

[0006] The retractable detection arm is used to adjust the distance between the detection head module and the surface of thick-skinned fruits and vegetables;

[0007] The rotating platform is used to drive thick-skinned fruits and vegetables to rotate to achieve all-round detection.

[0008] The data processing module includes a skin interference subtraction unit, a multimodal data fusion unit and a quality judgment unit;

[0009] The epidermal interference subtraction unit is used to establish and call an epidermal interference spectrum library to eliminate the interference of the epidermis on the visible light signal;

[0010] The multimodal data fusion unit is used to fuse visible light and near-infrared spectrum data;

[0011] The quality determination unit is used to determine the quality of thick-skinned fruits and vegetables based on the fused data.

[0012] The display module is used to display the quality test results of thick-skinned fruits and vegetables.

[0013] The optical adjustment unit includes a wide-angle lens, an adjustable focus lens and an exposure time adjustment component;

[0014] The wide-angle lens is used to expand the range of light source illumination and spectrum collection;

[0015] The adjustable focus lens is used to adjust the focal length according to the detection distance;

[0016] The exposure time adjustment component is used to adjust the exposure time of the silicon-based multi-spectral chip acquisition unit.

[0017] The visible light LED is a red light LED in the 600-700nm band, and the light intensity is adjustable;

[0018] The near-infrared LED uses a wavelength range of 900-1000nm, and the light intensity is adjustable.

[0019] The visible light channel of the silicon-based multispectral chip acquisition unit is integrated with a high-gain photodiode to enhance the signal detection capability of weak penetrating light.

[0020] A non-destructive quality testing method for thick-skinned fruits and vegetables based on the device as described in any one of the above items, characterized by comprising the following steps:

[0021] S1 places thick-skinned fruits and vegetables on a rotating platform and uses the retractable detection arm to adjust the distance between the detection head module and the surface of the thick-skinned fruits and vegetables to 5-10 cm;

[0022] S2 turns on the light source unit, and the ring array of visible light LEDs and near-infrared LEDs simultaneously illuminate the surface of thick-skinned fruits and vegetables. The wide-angle lens of the optical adjustment unit expands the illumination range, and the adjustable focus lens is adjusted to a clear focus state;

[0023] The S3 silicon-based multispectral chip acquisition unit, under the control of the exposure time adjustment component and in conjunction with the rotation of the rotating platform, collects visible light and near-infrared spectral data at different locations on the surface of thick-skinned fruits and vegetables;

[0024] The epidermal interference subtraction unit of the S4 data processing module calls the epidermal interference spectrum library to perform epidermal interference subtraction processing on the collected visible light spectrum data;

[0025] The S5 multimodal data fusion unit fuses the processed visible light spectrum data with the near-infrared spectrum data to obtain comprehensive spectrum data;

[0026] The S6 quality judgment unit compares the comprehensive spectral data with the preset thick-skinned fruit and vegetable quality spectral database to determine the quality of thick-skinned fruits and vegetables and displays the results through the display module.

[0027] As mentioned above, the skin interference spectrum library is established by collecting the reflectance spectra of pure skin samples of thick-skinned fruits and vegetables, and can be updated according to different types of thick-skinned fruits and vegetables.

[0028] As mentioned above, the multimodal data fusion unit uses a convolutional neural network algorithm to fuse visible light and near-infrared spectral data.

[0029] Beneficial effects

[0030] Improve the effective range of visible light collection: The design of a ring array light source, wide-angle lens and rotating platform expands the irradiation and collection range of visible light. Combined with the retractable detection arm, it ensures that all areas on the surface of thick-skinned fruits and vegetables can be effectively detected, reducing signal blind spots.

[0031] Optimize collection efficiency: In view of the thick-skin characteristics, we select visible light and near-infrared light sources of appropriate bands, adjust the light intensity and exposure time, integrate high-gain photodiodes, and optimize the collection efficiency of visible light and near-infrared light at the same time, so that the two can work together to fully reflect the skin and internal quality of thick-skinned fruits and vegetables.

[0032] Improve detection accuracy: The skin interference subtraction unit removes the interference of the skin on the visible light signal. The multimodal data fusion unit fuses two types of spectral data and combines them with the preset quality spectrum database to improve the accuracy of quality judgment of thick-skinned fruits and vegetables.

[0033] Enhanced applicability: The retractable detection arm and rotating platform of the mechanical adaptation module enable the detection head to adapt to thick-skinned fruits and vegetables of different sizes, achieving one-stop applicability without the need to replace the detection head, improving the flexibility and convenience of detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Attached photos:

[0035] Figure 1 is a schematic diagram of the overall structure of the device;

[0036] Figure 2: Details of the detection head module;

[0037] Figure 3: Detail of the mechanical adaptation module;

[0038] Figure 4: Schematic diagram of data processing flow;

[0039] Markings in the figure: detection head module 100, light source unit 11, silicon-based multispectral chip acquisition unit 12, optical adjustment unit 13, mechanical adaptation module 200, retractable detection arm 21, detection head module 211, rotating platform 22, data processing module 300, epidermal interference subtraction unit 31, multimodal data fusion unit 32, quality judgment unit 33, display module 400. DETAILED DESCRIPTION

[0040] The technical solution of the present invention is further described in detail below with reference to specific embodiments.

[0041] like Figure 1 Schematic diagram of the overall structure of the device: A non-destructive quality detection device suitable for thick-skinned fruits and vegetables, consisting of a device housing device including: the detection head module 100, which integrates a light source unit 11, a silicon-based multi-spectral chip acquisition unit 12 and an optical adjustment unit 13, and a mechanical adaptation module 200, which is equipped with a detection area of ​​a retractable detection arm 21 and a rotating platform 22; and the device includes: the data processing module 300, which integrates a control area of ​​an epidermal interference subtraction unit 31, a multimodal data fusion unit 32 and a quality judgment unit 33, and is connected to the display screen of the display module 400. Through the full process of system collaboration-mechanical adjustment-spectral acquisition-data processing-result output, the core innovations for thick-skinned fruits and vegetables are highlighted, such as visible light range expansion, multi-spectral collaboration, and all-round detection.

[0042] As shown in the detailed diagram of the detection head module in FIG2 , the detection head module 100 (which is the core detection part) is equipped with a wide-angle lens, an adjustable focus lens, and an exposure time adjustment component, and its functions are debugged and are normal;

[0043] The annular array light source of the detection head adopts 600-700nm red light LED and 900-1000nm near-infrared LED, and the light intensity is adjustable;

[0044] The visible light channel of the silicon-based multispectral chip integrates a high-gain photodiode, which is combined with a wide-angle lens to expand the acquisition range, and the exposure time is set to 80ms.

[0045] The light source unit 11 is placed in the light source unit position of the detection head module 100 and is connected to the light intensity adjustment circuit. It uses 6-8 visible light LEDs and near-infrared LEDs distributed in a circular array. The visible light LEDs use red light LEDs in the 600-700nm band, and the light intensity is adjustable to reduce the absorption interference of epidermal chlorophyll. The near-infrared LEDs use the 900-1000nm band, and the light intensity is adjustable to facilitate the acquisition of internal quality information.

[0046] The silicon-based multispectral chip acquisition unit 12 is a silicon-based multispectral chip with an integrated high-gain photodiode. It is installed in the acquisition unit position to ensure that its detection channel covers the required band and is connected to the data processing module. The silicon-based multispectral chip acquisition unit 12 includes detection channels covering the visible light band (400-760nm) and the near-infrared band (760-2500nm). The visible light channel integrates a high-gain photodiode to enhance the signal detection capability of weak penetrating light.

[0047] The optical adjustment unit 13 includes a wide-angle lens, an adjustable focal length lens and an exposure time adjustment component. The wide-angle lens expands the light source illumination and spectrum collection range, the adjustable focal length lens adjusts the focal length according to the detection distance, and the exposure time adjustment component adjusts the exposure time of the collection unit to compensate for the light intensity attenuation caused by thick skin.

[0048] The display module 400 is used to display the test results for the convenience of operators.

[0049] The process is: spectral data collected by the detection head → analysis by the data processing module → output on the display screen.

[0050] As shown in the detailed diagram of the mechanical adaptation module in Figure 3: the mechanical adaptation module 200 is composed of a retractable detection arm 21 and a rotating platform 22; the retractable detection arm 21 can adjust the distance between the detection head module 211 and the surface of thick-skinned fruits and vegetables to keep it at 5-10 cm, reducing the attenuation of light in the air; the rotating platform 22 rotates at a speed of 5 r / min, driving the watermelon to achieve 360° scanning to avoid blind spots in detection.

[0051] As shown in the data processing flow diagram of FIG4 : the data processing module 300 includes a skin interference subtraction unit 31 , a multimodal data fusion unit 32 and a quality determination unit 33 ;

[0052] The skin interference subtraction unit 31 extracts effective signals by subtracting the reflection / absorption effect of the skin on visible light from the established skin interference spectrum library (obtained by collecting the reflectance spectra of pure skin samples of thick-skinned fruits and vegetables);

[0053] The multimodal data fusion unit 32 uses a convolutional neural network algorithm to fuse visible light (reflecting epidermal defects) and near infrared (reflecting internal quality) spectral data and correlate the characteristics of the two;

[0054] The quality determination unit 33 compares the fused comprehensive spectral data with a preset database of quality spectra of thick-skinned fruits and vegetables to determine the quality.

[0055] Data processing flow:

[0056] The first step, “spectral acquisition (visible light + near-infrared)”, corresponds to the synchronous acquisition of the detection head in the embodiment, reflecting the innovation of “simultaneously optimizing the acquisition efficiency of two spectra”.

[0057] The second step, “skin interference subtraction (visible light data)”, corresponds to the key processing of the embodiment: calling the pre-established watermelon skin spectrum library to eliminate the reflection / absorption interference of the thick skin on visible light and extract the effective signal of the flesh.

[0058] The third step, “multimodal data fusion,” corresponds to the algorithm design of the embodiment: using a convolutional neural network to correlate skin defects (visible light) with internal quality (near infrared), such as the correlation analysis between watermelon skin damage and sugar content distribution.

[0059] The fourth step "comparison with the quality database" corresponds to the judgment basis of the embodiment: the database contains the spectral characteristics of watermelons of different maturity levels, and the optimal result is matched by calculating the mean square error.

[0060] The fifth step, "outputting quality results," corresponds to the display phase of the embodiment: displaying indicators such as sugar content (error ≤ 0.5° Brix) and maturity on the screen, completing the detection closed loop.

[0061] Thick-skinned fruits and vegetables detection method:

[0062] S1 Placement and Adjustment: Place thick-skinned fruits and vegetables on the rotating platform, and use the retractable detection arm to adjust the distance between the detection head module and the surface of the fruits and vegetables to 5-10 cm.

[0063] S2 light source and light path adjustment: Turn on the light source unit, the circular array of LEDs illuminates the surface of fruits and vegetables, the wide-angle lens of the optical adjustment unit expands the illumination range, and the adjustable focus lens is adjusted to a clear focus.

[0064] S3 spectral acquisition: The silicon-based multispectral chip acquisition unit, under the control of the exposure time adjustment component, rotates in conjunction with the rotating platform to collect visible light and near-infrared spectral data at different positions.

[0065] S4 data processing: The epidermal interference subtraction unit processes the visible light data to remove epidermal interference; the multimodal data fusion unit fuses the two spectral data to obtain comprehensive data.

[0066] Quality determination and display: The quality determination unit compares the comprehensive data with the database, determines the quality and displays it through the display module.

[0067] Example 1 Detection process:

[0068] 1. Taking a watermelon as an example, place the watermelon on the rotating platform 22 and use the retractable detection arm 21 to adjust the distance between the detection head module 100 and the watermelon surface to 8 cm;

[0069] 2. Turn on the light source unit 11 and adjust the light intensity of the visible light LED and near-infrared LED. The wide-angle lens of the optical adjustment unit 13 expands the illumination range, and the adjustable focus lens is adjusted to clearly focus on the watermelon surface. The rotating platform 22 rotates to achieve all-round detection.

[0070] 3. The silicon-based multispectral chip acquisition unit 12, under the control of the exposure time adjustment component (exposure time set to 80ms), cooperates with the rotating platform 22 to rotate at a speed of 5 rpm to collect visible light and near-infrared spectral data at different locations on the watermelon surface;

[0071] 4. The data processing module 300 receives the spectral data collected by the detection head, and the skin interference subtraction unit 31 calls the watermelon skin interference spectrum library to process the collected visible light spectrum data and remove the skin interference;

[0072] 5. The multimodal data fusion unit 32 uses a convolutional neural network algorithm to fuse the processed visible light spectrum data with the near-infrared spectrum data to obtain comprehensive spectrum data;

[0073] 6. The quality determination unit 33 compares the comprehensive spectrum data with the watermelon quality spectrum database to determine the quality indicators of the watermelon, such as sugar content, maturity, and skin defects, and displays them through the display module 400.

[0074] In summary, the device and method highlight the core innovations of the device and process, such as the expansion of the visible light range for thick-skinned fruits and vegetables, multi-spectral collaboration, and all-round detection.

[0075] , which can meet the needs of quality testing of thick-skinned fruits and vegetables.

[0076] The above embodiments are only some of the embodiments of the present invention. In actual application, they can be adjusted and improved according to the types and characteristics of specific thick-skinned fruits and vegetables. The scope of protection of the present invention is not limited to the above embodiments, but also includes various modifications and improvements made based on the technical solution of the present invention.

Claims

1. A non-destructive quality testing device for thick-skinned fruits and vegetables, characterized in that: include: A detection head module (100), wherein the detection head module is integrated with a light source unit (11), a silicon-based multi-spectral chip acquisition unit (12), and an optical adjustment unit (13), wherein the light source unit (11) includes visible light LEDs and near-infrared LEDs distributed in a ring array, and the silicon-based multi-spectral chip acquisition unit (12) includes detection channels covering the visible light band (400-760nm) and the near-infrared band (760-2500nm); The mechanical adaptation module (200) comprises a retractable detection arm (21) and a rotating platform (22), wherein the retractable detection arm (21) is used to adjust the distance between the detection head module (211) and the surface of thick-skinned fruits and vegetables, and the rotating platform (22) is used to drive the thick-skinned fruits and vegetables to rotate to achieve all-round detection; The data processing module (300) includes a skin interference subtraction unit (31), a multimodal data fusion unit (32) and a quality judgment unit (33), wherein the skin interference subtraction unit (31) is used to establish and call a skin interference spectrum library to eliminate the interference of the skin on the visible light signal, the multimodal data fusion unit (32) is used to fuse visible light and near-infrared spectrum data, and the quality judgment unit (33) is used to judge the quality of thick-skinned fruits and vegetables based on the fused data; The display module (400) is used to display the quality test results of thick-skinned fruits and vegetables.

2. The device according to claim 1, characterized in that The optical adjustment unit (13) comprises a wide-angle lens, an adjustable focal length lens and an exposure time adjustment component, wherein the wide-angle lens is used to expand the light source illumination and spectrum collection range, the adjustable focal length lens is used to adjust the focal length according to the detection distance, and the exposure time adjustment component is used to adjust the exposure time of the silicon-based multi-spectral chip collection unit.

3. The device according to claim 1, characterized in that The visible light LED is a red light LED in the 600-700nm band, and the light intensity is adjustable; the near-infrared LED is a red light LED in the 900-1000nm band, and the light intensity is adjustable.

4. The device according to claim 1, characterized in that The visible light channel of the silicon-based multi-spectral chip acquisition unit (12) is integrated with a high-gain photodiode to enhance the signal detection capability of weak penetrating light.

5. A non-destructive quality testing method for thick-skinned fruits and vegetables based on the device according to any one of claims 1 to 4, characterized in that: The following steps are involved: S1 places thick-skinned fruits and vegetables on a rotating platform and uses the retractable detection arm to adjust the distance between the detection head module and the surface of the thick-skinned fruits and vegetables to 5-10 cm; S2 turns on the light source unit, and the ring array of visible light LEDs and near-infrared LEDs simultaneously illuminate the surface of thick-skinned fruits and vegetables. The wide-angle lens of the optical adjustment unit expands the illumination range, and the adjustable focus lens is adjusted to a clear focus state; The S3 silicon-based multispectral chip acquisition unit, under the control of the exposure time adjustment component and in conjunction with the rotation of the rotating platform, collects visible light and near-infrared spectral data at different locations on the surface of thick-skinned fruits and vegetables; The epidermal interference subtraction unit of the S4 data processing module calls the epidermal interference spectrum library to perform epidermal interference subtraction processing on the collected visible light spectrum data; The S5 multimodal data fusion unit fuses the processed visible light spectrum data with the near-infrared spectrum data to obtain comprehensive spectrum data; The S6 quality judgment unit compares the comprehensive spectral data with the preset thick-skinned fruit and vegetable quality spectral database to determine the quality of thick-skinned fruits and vegetables and displays the results through the display module.

6. The method according to claim 5, characterized in that The skin interference spectrum library is established by collecting reflectance spectra of pure skin samples of thick-skinned fruits and vegetables, and can be updated according to different types of thick-skinned fruits and vegetables.

7. The method according to claim 5, characterized in that The multimodal data fusion unit uses a convolutional neural network algorithm to fuse the visible light and near-infrared spectrum data.

Images