Contact-type image sensor based on adaptive multispectrum

Abstract

The invention discloses a contact image sensor based on adaptive multispectrum, which relates to the technical field of image sensing and comprises a self-focusing lens array, a liquid lens array, an optical filter array, a driving circuit, an image sensing element and a control unit, according to the invention, by integrating the multi-band liquid lens array and the one-to-one corresponding optical filter arrays and combining an independent driving circuit and an adaptive control algorithm, mechanical-free automatic focusing in a wide spectral range is realized, a mechanical moving part in a traditional scheme is not needed, the problem of a complex structure caused by mechanical switching in multi-spectral imaging is solved, and the imaging efficiency is improved. By configuring an independent voltage regulation and control mechanism for each liquid lens, focal length requirements of different spectral bands can be accurately matched, the problem that part of band images are blurred due to a fixed focal length is avoided, delay generated by mechanical switching is eliminated by means of the quick response characteristic of the liquid lenses, and the multi-spectral scanning efficiency is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image sensing, in particular to a contact image sensor based on adaptive multi-spectrum. BACKGROUND

[0002] A contact image sensor (CIS) is a close-range imaging device that directly images the information on the surface of an object onto a photosensitive element through an optical system. Its core principle is to use a lens array to focus the reflected light of a scanning object onto an image sensing element to achieve high-precision digital acquisition of text, patterns or biological tissues. As a lightweight and low-cost imaging solution, the contact image sensor has irreplaceable significance in the fields of document scanning, industrial defect detection, biomedical diagnosis, etc. It not only promotes the miniaturization of office automation equipment, but also provides core technical support for on-site rapid detection and portable medical devices.

[0003] However, the existing contact image sensor technology still has some defects when in use. The existing contact image sensor is mostly designed based on a single spectrum. If multi-spectrum information is needed, mechanical switching of filters or light sources is required, which increases the complexity of the structure, delays the response, and increases the cost. Although some multi-spectrum contact image sensors integrate multiple filters, they use a fixed focal length design. Due to the difference in light refraction characteristics of different spectral bands, image blurring may occur in some bands. Few adjustable focus solutions only optimize for a single lens and cannot cover a wide spectral range, making it difficult to meet the multi-spectral imaging needs of complex scenes. Therefore, it is of great significance to develop a contact image sensor based on adaptive multi-spectrum. SUMMARY

[0004] The purpose of the present application is to overcome the shortcomings of the prior art and provide a contact image sensor based on adaptive multi-spectrum. It can integrate a multi-band liquid lens array and a one-to-one corresponding filter, combine an independent driving circuit and an adaptive control algorithm, realize mechanical focusing-free in a wide spectral range, match the focal length requirements of different spectra through independent voltage regulation of each liquid lens, solve the problem of uneven multi-spectrum image clarity, eliminate mechanical switching delay with the fast response characteristics of liquid lenses, and improve scanning efficiency.

[0005] To solve the above technical problems, the present application provides the following technical solution: a contact image sensor based on adaptive multi-spectrum, comprising: a self-focusing lens array, a liquid lens array, a filter array, a driving circuit, an image sensing element, and a control unit. The self-focusing lens array is used to focus the image of a scanning object onto an image sensing element. The liquid lens array is disposed in the optical path between the self-focusing lens array and the image sensing element, and includes three liquid lenses, which are composed of three independently controlled liquid lenses: a visible light liquid lens, a near-infrared liquid lens, and an ultraviolet liquid lens, each liquid lens corresponding to a different spectral band; The filter array is a three-layer structure corresponding to the liquid lens array, including: RGB filter, near-infrared filter, and ultraviolet filter; The driving circuit is connected to the liquid lens array and is used to apply an independent control voltage to each liquid lens to adjust its curvature; The image sensing element receives multispectral image signals processed by a self-focusing lens array, a liquid lens array, and a filter array; The control unit is connected to the drive circuit and the image sensing element, and is used to control the drive circuit to adjust the focal length of each liquid lens according to the preset spectral parameters and real-time image analysis results.

[0006] Furthermore, each liquid lens in the liquid lens array employs a differentiated liquid combination and electrode structure. The visible light liquid lens uses a combination of water and silicone oil, the near-infrared liquid lens uses a combination of ionic liquid and fluorinated oil, and the ultraviolet liquid lens uses a combination of ultraviolet transparent liquid and silicone oil. The focal length adjustment of each liquid lens and the control voltage satisfy the following relationship: In the formula, This is the focal length adjustment amount. This is the lens specificity coefficient (determined by the liquid refractive index and the cavity structure). To control the voltage, To correspond to the center wavelength of the spectral band, the electrodes of each liquid lens are made of ITO transparent conductive film and encapsulated in an independent transparent chamber.

[0007] Furthermore, the filter array is directly deposited on the light-emitting surface of each liquid lens in the liquid lens array using micro-nano fabrication technology. The RGB filter is an integrated red, green, and blue strip structure. The near-infrared filter has a center wavelength of 850nm and a bandwidth of 100nm, and the ultraviolet filter has a center wavelength of 365nm and a bandwidth of 50nm. The transmittance distribution of the filters satisfies the following: In the formula, wavelength Transmittance at that location Peak transmittance (≥80%). The center wavelength, This is the bandwidth factor (calculated from the bandwidth).

[0008] Further, the driving circuit takes FPGA as the control core to generate 0-50V adjustable voltage signal, and the visible light liquid lens is driven by 100Hz sine wave, the near-infrared liquid lens is driven by 50Hz square wave, and the ultraviolet liquid lens is driven by 80Hz triangular wave, and the driving circuit communicates with the control unit through the SPI interface.

[0009] Further, the self-focusing lens array is prepared by ion exchange method using low dispersion glass, the diameter of a single self-focusing lens is 0.3mm, the pitch is 0.35mm, the numerical aperture NA=0.55, and a plurality of self-focusing lenses are arranged into a linear array and fixed on a silicon substrate, and the two end faces are precisely polished.

[0010] Further, the image sensing element is a multi-channel linear array or area array sensor, each channel independently corresponds to a spectral band, and the image signals of the corresponding bands are synchronously collected, and the pixel density is not less than 600DPI.

[0011] Further, the control unit comprises a spectrum recognition module and a focus optimization algorithm, the spectrum recognition module is used to analyze the spectral characteristics of the target object obtained by initial scanning to determine the working spectral combination, and the focus optimization algorithm is used to adjust the control voltage of each liquid lens according to the image sharpness evaluation function, and the sharpness evaluation function is: In the formula, is the sharpness evaluation index, , respectively, , gradient operator in the direction, is the image pixel value, is a nonlinear coefficient (value 2-4).

[0012] Further, when the focus optimization algorithm of the control unit is executed, the initial voltage values of the visible light liquid lens, the near-infrared liquid lens and the ultraviolet liquid lens are calculated according to the preset spectrum-focal length mapping table, the images of each spectral channel are collected and the sharpness evaluation function is calculated, and finally the hill climbing algorithm is used to fine-tune the voltage of each liquid lens until the image sharpness reaches the maximum value, and the optimal parameter combination is saved for subsequent scanning of the same type of object.

[0013] Further, the adjustable focal length range of each liquid lens in the liquid lens array is set according to the characteristics of the corresponding spectral band, the focal length range of the visible light liquid lens is 12-18mm, the focal length range of the near-infrared liquid lens is 10-16mm, and the focal length range of the ultraviolet liquid lens is 14-20mm, and the response time of each liquid lens is less than 10ms.

[0014] Compared with the prior art, the contact image sensor based on adaptive multi-spectrum has the following beneficial effects: The application realizes automatic focusing without mechanical in a wide spectral range by integrating a multi-band liquid lens array with a one-to-one corresponding filter array, combining an independent driving circuit and an adaptive control algorithm, without mechanical moving parts in the traditional scheme, solving the problem of complex structure caused by mechanical switching in multi-spectral imaging, configuring an independent voltage regulation mechanism for each liquid lens, accurately matching the focal length requirements of different spectral bands, avoiding the image blur problem caused by fixed focal length, ensuring clear imaging in the full spectral range, eliminating the delay caused by mechanical switching by the fast response characteristics of the liquid lens, improving the multi-spectral scanning efficiency, effectively reducing the system volume through integrated optical path design, and reducing the equipment cost and power consumption.

[0015] Additional advantages, objects, and features of the application will be set forth in part by the description that follows, and in part will become apparent to those skilled in the art upon examination of the following specification or can be learned by practice of the application. The objects and advantages of the application can be realized and attained by means of the instrumentalities and combinations particularly pointed out in the appended claims. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings from these drawings without creative labor.

[0017] Figure 1 It is a whole structure schematic diagram of a contact image sensor based on adaptive multi-spectrum. Figure 2 It is a working flowchart of a contact image sensor based on adaptive multi-spectrum.

[0018] In the figure: 1, self-focusing lens array; 2, liquid lens array; 3, filter array; 4, driving circuit; 5, image sensing element; 6, control unit; L1, visible light liquid lens; L2, near-infrared liquid lens; L3, ultraviolet liquid lens; F1, RGB filter; F2, near-infrared filter; F3, ultraviolet filter. DETAILED DESCRIPTION

[0019] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined invention purpose, the following will combine the drawings and the preferred embodiments to specifically describe the specific embodiments, structures, features and effects of the present application.

[0020] Reference Figure 1The adaptive multi-spectrum based contact image sensor specifically comprises a self-focusing lens array 1, a liquid lens array 2, a filter array 3, a driving circuit 4, an image sensing element 5 and a control unit 6, and the specific contents of each part are as follows. The self-focusing lens array 1 is prepared by ion exchange method using low dispersion glass, the diameter of a single self-focusing lens is 0.3 mm, the pitch is 0.35 mm, the numerical aperture NA=0.55, and multiple self-focusing lenses are arranged into a linear array and fixed on a silicon substrate, the two end faces of which are precisely polished, and the main function is to focus the image of the scanning object onto the image sensing element 5.

[0021] The liquid lens array 2 is arranged on the light path between the self-focusing lens array 1 and the image sensing element 5, and is composed of three independently controlled liquid lenses corresponding to different spectral bands, including a visible light liquid lens L1, a near-infrared liquid lens L2 and an ultraviolet liquid lens L3.

[0022] The visible light liquid lens L1 adopts a combination of water and silicone oil, the near-infrared liquid lens L2 adopts a combination of ionic liquid and fluorine oil, and the ultraviolet liquid lens L3 adopts a combination of ultraviolet transparent liquid and silicone oil.

[0023] The electrodes of each liquid lens are made of ITO transparent conductive film and are packaged in an independent transparent chamber, and the adjustable focal length range is set according to the corresponding spectral band characteristics, wherein the focal length range of the visible light liquid lens L1 is 12-18 mm, the near-infrared liquid lens L2 is 10-16 mm, and the ultraviolet liquid lens L3 is 14-20 mm, and the response time of each liquid lens is less than 10 ms.

[0024] The filter array 3 is a three-layer structure corresponding to the liquid lens array 2, which is directly deposited on the light exit surface of each liquid lens in the liquid lens array 2 by micro-nano processing technology, and includes an RGB filter F1, a near-infrared filter F2 and an ultraviolet filter F3.

[0025] The RGB filter F1 is a strip structure integrating red, green and blue colors, the central wavelength of the near-infrared filter F2 is 850 nm with a bandwidth of 100 nm, and the central wavelength of the ultraviolet filter F3 is 365 nm with a bandwidth of 50 nm, and the peak transmittance is ≥80%.

[0026] The driving circuit 4 is connected with the liquid lens array 2, takes FPGA as a control core, can generate a 0-50V adjustable voltage signal, is used for applying independent control voltage to each liquid lens to adjust the curvature, different driving modes are adopted for different liquid lenses, wherein the visible light liquid lens L1 is driven by a 100Hz sine wave, the near-infrared liquid lens L2 is driven by a 50Hz square wave, the ultraviolet liquid lens L3 is driven by an 80Hz triangular wave, and the driving circuit 4 communicates with the control unit 6 through an SPI interface.

[0027] The image sensing element 5 is a multi-channel linear array or area array sensor, each channel independently corresponds to one spectral band, can synchronously collect image signals of the corresponding wave band, and the pixel density is not less than 600DPI, and is used for receiving the multi-spectral image signals processed by the self-focusing lens array 1, the liquid lens array 2 and the filter array 3.

[0028] The control unit 6 is connected with the driving circuit 4 and the image sensing element 5, includes a spectrum recognition module and a focus optimization algorithm, the spectrum recognition module is used for analyzing the spectral characteristics of the target object obtained by initial scanning to determine the working spectral combination, and the focus optimization algorithm is used for adjusting the control voltage of each liquid lens according to an image sharpness evaluation function, when the algorithm is executed, first, the initial voltage value of each liquid lens is calculated according to a preset spectrum-focal length mapping table, then the images of each spectral channel are collected and the sharpness evaluation function is calculated, finally, the voltage of each liquid lens is fine-tuned by using a hill climbing algorithm until the image sharpness reaches a maximum value, and the optimal parameter combination is saved for subsequent scanning of the same type of object, and the control unit 6 is used as a whole to control the driving circuit 4 to adjust the focal length of each liquid lens according to the preset spectral parameters and the real-time image analysis result.

[0029] Embodiment one The embodiment is applied to a multi-spectral detection scene of cultural relics surfaces, aims to obtain high-definition images of cultural relics surfaces in visible light, near-infrared and ultraviolet wave bands through non-contact scanning, and provides comprehensive spectral data support for age identification, material analysis and damage evaluation of cultural relics, in traditional cultural relic detection, multi-spectral imaging often causes low scanning efficiency and image distortion due to mechanical switching of filters or lenses, and different wave band focal lengths are not matched, which easily causes partial area blur, the embodiment can efficiently solve the above problems through the mechanical focusing-free and multi-wave band synchronous acquisition capability of the adaptive multi-spectral contact image sensor.

[0030] Referring to Figure 1 and Figure 2 , the implementation process of the embodiment is as follows: After the control unit 6 is started, a preset spectrum-focal length mapping table is first called, which stores the initial voltage reference values of the visible light (400-760 nm), near-infrared (760-1100 nm), and ultraviolet (200-400 nm) bands based on the spectral characteristics of common materials (such as paper, pigments, and metals) of cultural relics.

[0031] The self-focusing lens array 1 converges the reflected light from the surface of the cultural relic into parallel light through the ion exchange characteristics of low dispersion glass, and the linear arrangement of the silicon-based substrate ensures that the light uniformly enters the subsequent optical path. The high numerical aperture (NA=0.55) of the individual lenses ensures the imaging brightness in a weak light environment.

[0032] The image sensing element 5 first performs a rapid initial scan to obtain the rough spectral distribution of the surface of the cultural relic. After the initial image is analyzed by the spectral recognition module, the waveband combination that needs to be collected is determined (for example, if a painted cultural relic is detected, the visible light liquid lens L1 and the RGB filter F1 are preferentially activated; if metal corrosion is detected, the near-infrared liquid lens L2 and the ultraviolet liquid lens L3 are simultaneously activated).

[0033] The control unit 6 sends instructions to the driving circuit 4 through the SPI interface, and the FPGA core of the driving circuit 4 generates voltage signals corresponding to the waveforms according to the instructions: 100Hz sine wave is output to L1, 50Hz square wave is output to L2, and 80Hz triangular wave is output to L3, to ensure that the electrodes (ITO transparent conductive film) of each liquid lens generate a stable electric field.

[0034] Each lens in the liquid lens array 2 changes its curvature according to the voltage signal: the visible light liquid lens L1 uses a combination of water and silicone oil, which adjusts the focal length by changing the interfacial tension under the action of the electric field. The focal length adjustment amount follows the formula where is the center wavelength of visible light, is determined by the refractive indices of water and silicone oil and the chamber structure. The combination of ionic liquid and fluorine oil in the near-infrared liquid lens L2 quickly responds to the focal length requirements of the near-infrared band under square wave voltage due to the difference in electrical conductivity. The combination of ultraviolet transparent liquid and silicone oil in the ultraviolet liquid lens L3 precisely matches the refractive properties of 365nm ultraviolet light through triangular wave voltage.

[0035] The focus optimization algorithm of the control unit 6 calculates the image sharpness evaluation function in real time by , the gradient operator in the direction of , analyzing the edge sharpness of each waveband image. If the value of a certain waveband is below the threshold, the algorithm starts the hill climbing algorithm to fine-tune the driving voltage, for example, when the ultraviolet image is blurred, gradually increase the voltage of the ultraviolet liquid lens L3 until The value reaches the maximum, at which time the focal length is stabilized at the optimal value in the range of 14-20 mm.

[0036] The filter array 3 is integrated with the light output surface of the liquid lens by micro-nano processing technology, ensuring no deviation in the optical path: the RGB strip structure of the RGB filter F1 decomposes visible light into three channels of red, green and blue, the near-infrared filter F2 only allows near-infrared light of 850 nm ± 50 nm to pass, and the ultraviolet filter F3 filters out stray light outside 365 nm ± 25 nm, and its transmittance distribution meets , , The bandwidth is calculated.

[0037] The filtered monochromatic light is received by the multi-channel sensor of the image sensing element 5, which independently converts the light signal into an electrical signal, and the pixel density of more than 600 DPI ensures that the fine cracks, pigment layering and other features on the surface of cultural relics are clearly recorded.

[0038] The image sensing element 5 synchronously transmits the electrical signals of the three wavebands to the control unit 6, which performs noise reduction and splicing processing on the signals to generate a multispectral composite image containing visible light texture, near-infrared material distribution, and ultraviolet fluorescence reaction.

[0039] The storage module saves the original data, clarity evaluation results and voltage adjustment records in real time, and the display module dynamically refreshes the images of each waveband and the energy recovery efficiency (if the energy management module is integrated), which facilitates the real-time monitoring of scanning quality by the operator.

[0040] When the surface material of the cultural relic changes (such as switching from a painted area to a wooden area), the control unit 6 automatically triggers a new round of focal length optimization by analyzing the rate of change of the image pixel value: for example, the near-infrared absorption characteristics of the wooden area are enhanced, and the algorithm will lower the voltage of the near-infrared liquid lens L2 to extend the focal length, so that the value of the near-infrared image rises.

[0041] After scanning is completed, the control unit 6 stores the optimal voltage combination, spectral parameters and other information of this detection into the spectral-focal length mapping table, which can be directly called for detection of similar cultural relics, reducing the repeated calibration time.

[0042] In summary, the embodiment realizes efficient and clear collection of multi-band images of cultural relics surfaces through the cooperative work of the adaptive multi-spectral contact image sensor. The system can complete focal length matching in a wide spectral range without mechanical components, solving the delay problem caused by mechanical switching in traditional equipment; the independent voltage regulation mechanism of each liquid lens ensures that the images of visible light, near-infrared, and ultraviolet bands all reach high definition; the integrated optical path design (self-focusing lens array + liquid lens + filter integration) reduces the size of the equipment and reduces power consumption, while the pixel density of more than 600 DPI meets the identification needs of subtle features in cultural relic detection. This technical solution is not only suitable for the field of cultural relic protection, but also can be expanded to industrial defect detection, biological tissue imaging, and other scenes, and has wide application value.

[0043] Embodiment Two The embodiment is applied to the scene of rapid detection of food quality, aiming to non-invasively detect the internal quality (such as sugar content, water distribution, and mold condition) of fruits, meat, and other foods through multi-spectral imaging technology. In traditional food detection, single spectral imaging cannot fully reflect the physicochemical properties of food, and mechanical switching type multi-spectral equipment has the problems of low detection efficiency and easy image distortion due to vibration. The embodiment can realize high-speed and accurate detection of food quality in a production line environment through the mechanical focusing-free and multi-band synchronous acquisition capabilities of the adaptive multi-spectral contact image sensor, providing technical support for food safety control.

[0044] Referring to Figure 1 and Figure 2 , the implementation process of the embodiment is as follows: After the control unit 6 is powered on, it automatically loads the spectral-focal length mapping table dedicated for food detection. This table stores the initial voltage parameters of the visible light (for appearance color detection), near-infrared (for sugar and water analysis), and ultraviolet (for mold and pesticide residue detection) bands based on the spectral response characteristics of common foods (such as apples, beef, bread, etc.).

[0045] The self-focusing lens array 1 is prepared by ion exchange method of low dispersion glass, which converges the scattered light from the surface of the food into parallel light beams. The linearly arranged silicon substrate and the precisely polished two end faces ensure that the light does not deviate during transmission, and the high numerical aperture (NA = 0.55) design ensures that the edge area of curved food (such as fruits) can also be clearly imaged.

[0046] The image sensing element 5 performs an initial scan on the food on the conveyor belt, and the spectral recognition module analyzes the spectral peaks in the scan data: if the detection object is an apple, the visible liquid lens L1 and the RGB filter F1 are activated in priority, because the maturity and the content of red pigment are related; if the detection object is beef, the intermuscular fat distribution needs to be analyzed through the near-infrared wave band, and the near-infrared liquid lens L2 and the near-infrared filter F2 are activated at the same time; if the risk of mold is detected, the ultraviolet liquid lens L3 and the ultraviolet filter F3 will be activated.

[0047] The control unit 6 sends a spectral combination instruction to the driving circuit 4 through the SPI interface, and the FPGA core of the driving circuit 4 generates a differentiated voltage signal according to the instruction: a 100Hz sine wave is output to the visible liquid lens L1 (adapted to the fast response requirement of the visible light wave band), a 50Hz square wave is output to the near-infrared liquid lens L2 (satisfies the stable regulation and control of the near-infrared wave band), and a 80Hz triangular wave is output to the ultraviolet liquid lens L3 (adapted to the fine adjustment of the ultraviolet wave band), and the voltage range is continuously adjustable within 0-50V.

[0048] Each lens in the liquid lens array 2 changes the interface curvature according to the voltage signal: the visible liquid lens L1 adopts the combination of water and silicone oil, under the action of the sine wave voltage, the focal length is adjusted by the periodic change of the liquid interfacial tension, and the focal length adjustment amount follows the formula , wherein is the visible light center wavelength, is determined by the refractive index of water and silicone oil and the chamber structure, the combination of ionic liquid and fluorine oil of the near-infrared liquid lens L2 changes the focal length of the near-infrared wave band quickly under the action of the square wave voltage due to ionic conductivity, which adapts to the focusing requirement of the scattered light inside the food; the combination of ultraviolet transparent liquid and silicone oil of the ultraviolet liquid lens L3 precisely matches the short wavelength refraction characteristics of ultraviolet light through the linear change of the triangular wave voltage.

[0049] The electrodes of each liquid lens adopt ITO transparent conductive film, which generates a uniform electric field under the action of voltage, and the independent transparent chamber design avoids the mutual interference between different liquids, and the response time is less than 10ms, which can match the high-speed conveying requirement of the production line (usually 1-2m / s).

[0050] The filter array 3 is directly deposited on the light output surface of the liquid lens through micro-nano processing technology, which ensures the integration of the light path: the RGB strip structure of the RGB filter F1 decomposes the visible light into red, green and blue three channels, which correspond to the color parameters of the food (such as the red saturation of apples and the freshness color of meat); the near-infrared filter F2 only allows specific near-infrared wave band to pass, which is used to analyze the molecular vibration spectrum inside the food; the ultraviolet filter F3 filters out the ultraviolet stray light, and only retains the wave band that can excite the fluorescence of mold spores.

[0051] The transmittance distribution of the filter meets the formula , , (Calculated from bandwidth) to ensure efficient transmission of light signals in the target band while effectively suppressing stray light. For example, when detecting the moisture content of beef, the high transmittance of the near-infrared filter F2 in the near-infrared band can enhance the signal intensity of the moisture characteristic peak.

[0052] The focus optimization algorithm of control unit 6 calculates the sharpness evaluation function of images in each band in real time. pass , Gradient operator of direction 、 Analyze the edge sharpness of images in each band (such as the texture of fruit surfaces and the muscle fiber structure of meat), and analyze how the food's edge sharpness is affected by the vibration of the conveyor belt in a certain band. When the value decreases, the algorithm initiates a hill-climbing algorithm to fine-tune the voltage of the corresponding liquid lens: for example, if the mold spots in the ultraviolet image are blurred, the voltage of the ultraviolet liquid lens L3 is gradually increased, and the focal length adjustment formula is used to adjust the voltage. The value has risen back above the threshold.

[0053] The multi-channel sensor of the image sensing element 5 synchronously receives light signals from three wavelengths. Each channel independently completes photoelectric conversion. The pixel density of over 600 DPI ensures that minute defects on the food surface (such as bruises on fruit or bruises on meat) are clearly captured. The synchronous acquisition feature avoids multi-band image misalignment caused by food movement.

[0054] Control unit 6 performs fusion processing on multi-band image data: integrating the color parameters of the visible light image, the component analysis results of the near-infrared image, and the safety indicators of the ultraviolet image into a comprehensive quality score. For example, when inspecting apples, ripeness is determined by the color values ​​of the RGB channel, sugar content is calculated by the absorption peaks of the near-infrared channel, and surface mold is detected by the fluorescence intensity of the ultraviolet channel.

[0055] The storage module saves the original images, clarity evaluation data and voltage adjustment records in real time, while the display module dynamically refreshes the test results (such as "qualified", "excessive sugar content", "risk of mold"). The control unit 6 simultaneously sends the results to the production line control system via the industrial bus to achieve automatic rejection of defective products.

[0056] Dynamic adaptation and parameter iteration When the object being detected is changed (e.g., from apple to bread), the spectral recognition module automatically identifies the spectral characteristics of the new object. The control unit 6 adjusts the output signal of the drive circuit 4 through the SPI interface, and the liquid lens array 2 completes the focal length switching within 10ms without the need for shutdown calibration.

[0057] After each detection is completed, the control unit 6 updates the optimal voltage parameters and the spectrum combination to the spectrum-focal length mapping table, optimizes the detection model of different foods through a machine learning algorithm, and gradually improves the detection accuracy and efficiency.

[0058] To sum up, the embodiment realizes high-speed and multi-dimensional detection of food quality through the cooperative work of the adaptive multi-spectrum contact image sensor. The system can complete focal length matching in a wide spectral range without mechanical components, solves the problem of mechanical switching equipment failure due to vibration in the production line environment, and improves the detection efficiency compared with traditional equipment; the integrated optical path design (self-focusing lens + liquid lens + filter integration) reduces the device volume, adapts to the compact space requirements of the production line, and has low power consumption characteristics suitable for long-time continuous operation. This technical solution is not only suitable for food detection, but also can be extended to the fields of agricultural product grading, pharmaceutical packaging detection, etc., and has significant industrial value.

[0059] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content without departing from the scope of the present application, and any equivalent embodiments with equivalent changes and modifications are still within the scope of the present application.

Claims

1. A contact image sensor based on adaptive multispectral imaging, characterized in that, include: The system comprises a self-focusing lens array (1), a liquid lens array (2), a filter array (3), a driving circuit (4), an image sensing element (5), and a control unit (6). The self-focusing lens array (1) is used to focus the image of the scanned object onto the image sensing element (5); The liquid lens array (2) is set in the optical path between the self-focusing lens array (1) and the image sensing element (5), and includes three liquid lenses, which are composed of three independently controlled liquid lenses: a visible light liquid lens (L1), a near-infrared liquid lens (L2), and an ultraviolet liquid lens (L3), each liquid lens corresponding to a different spectral band; The filter array (3) is a three-layer structure corresponding to the liquid lens array (2), including: RGB filter (F1), near-infrared filter (F2), and ultraviolet filter (F3). The driving circuit (4) is connected to the liquid lens array (2) and is used to apply an independent control voltage to each liquid lens to adjust its curvature; The image sensing element (5) receives the multispectral image signal after being processed by the self-focusing lens array (1), the liquid lens array (2) and the filter array (3); The control unit (6) is connected to the drive circuit (4) and the image sensing element (5) and is used to control the drive circuit (4) to adjust the focal length of each liquid lens according to the preset spectral parameters and real-time image analysis results.

2. The contact image sensor based on adaptive multispectral imaging according to claim 1, characterized in that, The liquid lens array (2) employs differentiated liquid combinations and electrode structures for each liquid lens. The visible light liquid lens (L1) uses a combination of water and silicone oil, the near-infrared liquid lens (L2) uses a combination of ionic liquid and fluorinated oil, and the ultraviolet liquid lens (L3) uses a combination of ultraviolet transparent liquid and silicone oil. The focal length adjustment and control voltage of each liquid lens satisfy the following relationship: In the formula, This is the focal length adjustment amount. This is the lens specificity coefficient. To control the voltage, To correspond to the center wavelength of the spectral band, the electrodes of each liquid lens are made of ITO transparent conductive film and encapsulated in an independent transparent chamber.

3. The contact image sensor based on adaptive multispectral imaging according to claim 1, characterized in that, The filter array (3) is directly deposited on the light-emitting surface of each liquid lens in the liquid lens array (2) using micro-nano fabrication technology. The RGB filter (F1) is a strip structure integrating red, green, and blue colors. The near-infrared filter (F2) has a center wavelength of 850nm and a bandwidth of 100nm. The ultraviolet filter (F3) has a center wavelength of 365nm and a bandwidth of 50nm. The transmittance distribution of the filters satisfies: In the formula, wavelength Transmittance at that location Peak transmittance The center wavelength, This represents the bandwidth coefficient.

4. A contact image sensor based on adaptive multispectral imaging according to claim 1, characterized in that, The driving circuit (4) uses FPGA as the control core to generate a 0-50V adjustable voltage signal. It uses a 100Hz sine wave to drive the visible light liquid lens (L1), a 50Hz square wave to drive the near-infrared liquid lens (L2), and an 80Hz triangular wave to drive the ultraviolet liquid lens (L3). It also communicates with the control unit (6) through the SPI interface.

5. A contact image sensor based on adaptive multispectral imaging according to claim 1, characterized in that, The self-focusing lens array (1) is made of low-dispersion glass by ion exchange method. The diameter of a single self-focusing lens is 0.3 mm, the pitch is 0.35 mm, the numerical aperture NA is 0.55, and multiple self-focusing lenses are arranged in a linear array and fixed on a silicon substrate. Both ends are precision polished.

6. A contact image sensor based on adaptive multispectral imaging according to claim 1, characterized in that, The image sensing element (5) is a multi-channel linear array or area array sensor, each channel independently corresponds to a spectral band, synchronously acquires the image signal of the corresponding band, and the pixel density is not less than 600 DPI.

7. A contact image sensor based on adaptive multispectral imaging according to claim 1, characterized in that, The control unit (6) includes a spectral recognition module and a focus optimization algorithm. The spectral recognition module is used to analyze the spectral characteristics of the target object acquired in the initial scan to determine the working spectral combination. The focus optimization algorithm is used to adjust the control voltage of each liquid lens according to the image sharpness evaluation function, which is: In the formula, As a metric for evaluating sharpness, , They are respectively , Gradient operator of direction, Image pixel values, These are nonlinear coefficients.

8. A contact image sensor based on adaptive multispectral imaging according to claim 1, characterized in that, When the focus optimization algorithm of the control unit (6) is executed, it first calculates the initial voltage values ​​of the visible light liquid lens (L1), near-infrared liquid lens (L2), and ultraviolet liquid lens (L3) according to the preset spectrum-focal length mapping table, then acquires the images of each spectral channel and calculates the sharpness evaluation function, and finally uses the hill climbing algorithm to fine-tune the voltage of each liquid lens until the image sharpness reaches the maximum value, and saves the optimal parameter combination for subsequent scanning of the same type of object.

9. A contact image sensor based on adaptive multispectral imaging according to claim 1, characterized in that, The adjustable focal length range of each liquid lens in the liquid lens array (2) is set according to the characteristics of the corresponding spectral band. The focal length range of the visible light liquid lens (L1) is 12-18mm, the near-infrared liquid lens (L2) is 10-16mm, the ultraviolet liquid lens (L3) is 14-20mm, and the response time of each liquid lens is less than 10ms.

Images