Full-automatic logistics parcel high-speed sorting device based on visual recognition

Abstract

The present application relates to the technical field of industrial vision intelligence, and particularly relates to a full-automatic logistics package high-speed sorting device based on visual identification, which comprises a visual identification high-speed sorting system, the visual identification high-speed sorting system comprises a visual image acquisition module, a visual image processing module, a visual image light supplementing module and a sorting time control module, the system realizes differentiated visual data collection through preview image size evaluation, realizes adaptive light supplementing and partition optimization through a grid division difference comparison method, globally plans and downgrades the light supplementing strategy through global time balance, establishes a closed-loop optimization from visual identification to time scheduling, and significantly improves the complex package barcode recognition rate and the overall throughput of the sorting line.

Description

Technical Field

[0001] This invention relates to the field of industrial vision intelligence technology, and in particular to a fully automated high-speed sorting device for logistics parcels based on visual recognition. Background Technology

[0002] Automated parcel sorting is a core component of the express delivery and warehousing industry. Existing sorting vision systems are mainly divided into two categories: one is barcode scanning equipment based on fixed exposure and constant resolution. Although it can complete the recognition of regular parcels, it is completely unable to cope with complex working conditions such as black light absorption, reflective coating, and damaged barcodes; the other is industrial barcode readers based on simple image enhancement. Although it has preliminary grayscale adaptive capabilities, its supplementary lighting strategy is fixed, and it has not established a hierarchical supplementary lighting decision mechanism based on light absorption characteristic coefficients. Furthermore, it lacks dynamic budget management of visual processing time, resulting in timeouts for complex parcels and wasted time for simple parcels, making it difficult to improve overall throughput.

[0003] The patent with publication number CN121551293A discloses an adaptive sorting system based on machine vision, but it still has problems such as not introducing preview image size evaluation and dynamic resolution adjustment, and not building a graded supplementary lighting decision and differential optimization mechanism based on light absorption characteristic coefficient, resulting in low collection efficiency of packages of different sizes and high failure rate of recognition of local reflection overexposure.

[0004] The patent with publication number CN121353096A discloses an infrared and visible light image fusion method, but it is still not applied to barcode recognition scenarios in logistics sorting. It has not established a supplementary light mode switching mechanism based on the comparison of light absorption characteristic coefficient and grid difference, and has not made a global plan for the time consumption of visual processing, which leads to the blockage of sorting lines in complex package processing.

[0005] The patent with publication number CN120605873A discloses a sorting device that facilitates the identification of express waybills, but it still relies solely on mechanical tilting to avoid glare and does not adjust the processing time based on the light absorption coefficient and size evaluation results, resulting in barcode decoding failure and insufficient overall throughput. Summary of the Invention

[0006] In view of this, the present invention provides a fully automated high-speed sorting device for logistics parcels based on visual recognition, which overcomes the problems in the prior art that lack a hierarchical supplementary lighting decision and differential optimization mechanism based on light absorption characteristic coefficient, and that do not establish a global time balance for global overall planning and degraded scheduling of visual processing time, resulting in low recognition rate of complex parcels and insufficient overall throughput of sorting lines.

[0007] Specifically, the present invention is achieved through the following technical solution: The fully automated high-speed sorting device for logistics parcels based on vision recognition provided by the present invention includes a main conveyor belt, a V-shaped guide plate, a flexible pressure roller, a pre-triggered light curtain sensor, an auxiliary area array camera, a supplementary light, a main vision camera, a push rod, a slewing bearing, a drive wheel, a driven wheel, a frame, and a vision recognition high-speed sorting system, wherein: The frame is connected to the main conveyor belt, the main conveyor belt is mounted on the frame, and the driving wheel and the driven wheel are respectively mounted at both ends of the frame and connected to the main conveyor belt. The V-shaped guide plate is fixed above the main conveyor belt, and the flexible pressure roller is connected to the V-shaped guide plate; The pre-triggered light curtain sensor is mounted on the auxiliary area scan camera, which is fixed above the main conveyor belt. The supplementary light is installed above the main conveyor belt and connected to the main vision camera; The slewing bearing is fixed to the frame; The push rod is mounted on the slewing bearing; The high-speed visual recognition sorting system is mounted on a frame and is electrically connected to the auxiliary area array camera, the main visual camera, the fill light, and the push rod.

[0008] Optionally, a main conveyor belt is used to transport logistics packages; V-shaped deflectors are used to orient and center logistics packages. Flexible pressure rollers are used to smooth out logistics packages; A pre-triggered light curtain sensor is used to trigger the auxiliary area scan camera; An auxiliary area scan camera is used to acquire preview images; Supplemental lighting is used to provide supplemental lighting for logistics packages based on target lighting decisions; The main vision camera is used to acquire visual image data; Push rods are used to push packages into the corresponding sorting exit chute. Slewing bearing, used to provide slewing support for push rods; The drive pulley is used to drive the main conveyor belt; Driven pulleys are used to support and tension the main conveyor belt; The frame is used to support the main conveyor belt; The high-speed sorting system based on visual recognition is used to obtain the light absorption characteristic coefficient based on the processed visual data, output the target supplementary lighting decision, perform differential optimization on the output process of the target supplementary lighting decision, and downgrade the target supplementary lighting decision based on the global time balance.

[0009] Optionally, the supplementary light includes white LEDs and infrared LEDs, and the white LEDs and infrared LEDs are switched according to the target supplementary light decision. The white LEDs refer to a ring-shaped white LED light group with a color temperature between 5000K and 6500K, divided into 8 independent control sectors, each sector having 12 LED beads, and using PWM dimming. The infrared LEDs refer to a ring-shaped infrared LED light group with a peak wavelength between 850nm and 940nm, divided into 8 independent control sectors, each sector having 4 infrared LED beads, and using PWM dimming.

[0010] Optionally, the visual recognition high-speed sorting system includes: The visual image acquisition module is used to acquire the preview image, evaluate the size of the preview image, obtain the size evaluation result, and acquire the visual image data based on the size evaluation result; The visual image processing module is used to process visual image data to obtain processed visual data. The visual image supplementary lighting module is used to obtain the light absorption characteristic coefficient based on the processed visual data, and output the target supplementary lighting decision based on the light absorption characteristic coefficient. It is also used to perform differential optimization on the output process of the target supplementary lighting decision based on the grid division difference comparison method. Furthermore, it is used to obtain the proportion of overexposed pixels and perform fusion update on the differential optimization process based on the proportion of overexposed pixels. The sorting time control module is used to obtain the global time balance and downgrade the target illumination decision based on the global time balance.

[0011] Optionally, the visual image acquisition module acquires the preview image using an auxiliary area scan camera, performs edge detection on the preview image to obtain the barcode outline, calculates the area S of the circumscribed rectangle of the barcode outline, compares the area S of the circumscribed rectangle with a preset area threshold S0, and outputs the size evaluation result based on the comparison result, wherein: When S≤S0, the visual image acquisition module outputs the small-sized barcode as the size evaluation result; When S > S0, the visual image acquisition module outputs the large-size barcode as the size evaluation result.

[0012] Optionally, the visual image acquisition module acquires visual image data based on the size evaluation result, wherein: When the size level is large barcode, the acquisition resolution of the main vision camera is set to the first resolution, the ROI area is set to full frame, and the visual image data is acquired through the main vision camera. When the size category is small barcode, the acquisition resolution of the main vision camera is set to the second resolution, the ROI area is set to the barcode prediction area, and the visual image data is acquired through the main vision camera.

[0013] Optionally, the visual image processing module performs image processing on the visual image data, specifically by: converting the visual image data to grayscale to obtain a grayscale image; performing median filtering on the grayscale image to obtain a denoised image; performing Gamma correction on the denoised image to obtain a corrected image; performing Canny edge enhancement on the corrected image to obtain an edge-enhanced image; and outputting the edge-enhanced image as processed visual data.

[0014] Optionally, the visual image supplementary lighting module obtains the average gray value Gaavg and the proportion of low gray pixels Rlow based on the processed visual data, and calculates the light absorption characteristic coefficient Rcg based on the average gray value Gaavg, the proportion of low gray pixels Rlow, the weight of average gray value w1, and the weight of low gray pixel proportion w2, and sets Rcg=Gavg×w1+Rlow×w2; The light absorption characteristic coefficient Rcg is compared with the preset light absorption threshold Rc0. Based on the comparison result, the light absorption characteristic coefficient is judged, and the target supplementary lighting decision is output based on the judgment result. When Rcg≥Rc0, the visual image supplementary lighting module determines that the light absorption characteristic coefficient is light absorption and outputs the decision to enable infrared supplementary lighting mode as the target supplementary lighting decision. When Rcg < Rc0, the visual image supplementary lighting module determines that the light absorption characteristic coefficient is not absorbed, and outputs the white light supplementary lighting mode as the target supplementary lighting decision.

[0015] Optionally, the visual image supplementary lighting module performs differential optimization on the output process of the target supplementary lighting decision based on the grid partitioning difference comparison method, wherein the grid partitioning difference comparison method includes: Step A01: Divide the processed visual data into grids to obtain gridded visual data, and calculate the local absorption coefficient Rcj of each grid in the gridded visual data. Step A02: Obtain the maximum absorption value Rcmax and the minimum absorption value Rcmin based on the local absorption coefficient Rcj; Step A03: Calculate the grid absorption non-uniformity ΔRc based on the maximum absorption value Rcmax and the minimum absorption value Rcmin, where ΔRc = Rcmax - Rcmin; Step A04: Compare the grid absorption non-uniformity ΔRc with the preset non-uniformity threshold ΔRc0, determine the grid absorption state based on the comparison result, and perform differential optimization on the target supplementary lighting decision output process based on the determination result, wherein: When ΔRc < ΔRc0, the visual image supplementary lighting module determines that the grid light absorption state is uniform and does not perform differential optimization on the output process of the target supplementary lighting decision; When ΔRc≥ΔRc0, the visual image supplementary lighting module determines that the grid light absorption state is non-uniform, and performs differential optimization on the output process of the target supplementary lighting decision: If the target illumination decision is white light illumination mode, the white light illumination mode is processed by PWM reduction to obtain white light reduction mode, and the white light reduction mode is output as the target illumination decision. If the target illumination decision is infrared illumination mode, and image fusion processing is performed on the infrared illumination mode to obtain the infrared image fusion mode, and the infrared image fusion mode is output as the target illumination decision; The visual image supplementary lighting module acquires the overexposed pixel ratio Z, compares Z with a preset overexposed pixel ratio Z0, determines the state of the overexposed pixel ratio based on the comparison result, and performs a fusion update on the differential optimization process based on the determination result, wherein: When Z≤Z0, the visual image supplementary lighting module determines that the overexposed pixel ratio is normal and does not perform fusion update in the differential optimization process; When Z > Z0, the visual image illumination module determines that the proportion of overexposed pixels is abnormal and performs a fusion update on the differential optimization process: If the target illumination decision is white light illumination mode, the PWM reduction process is reduced a second time; If the target illumination decision is infrared illumination mode, the image fusion process is adjusted a second time.

[0016] Optionally, the sorting time control module obtains the global time balance B and sets... ,in, This refers to the barcode processing time of the previous logistics package. Using the barcode processing time Tpred as the base time, the processing time Tpred is obtained, and then the data is processed based on the barcode processing time Tpred and the base time. The global time balance B is used to downgrade the target illumination decision, where: when At that time, the target illumination decision is downgraded: If the target illumination decision is white light illumination mode, only a single frame of white light is retained for illumination; If the target illumination decision is infrared illumination mode, only a single frame of red light is retained for illumination. Otherwise, the target illumination decision is not downgraded, and the global time balance B is updated. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0019] Figure 1 A schematic diagram of the structure of a fully automated high-speed logistics parcel sorting device based on vision recognition provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of the high-speed visual recognition sorting system provided in an embodiment of the present invention; Figure 3 A schematic diagram illustrating the workflow of the visual image illumination module provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the mesh division difference comparison method provided in an embodiment of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] In related technologies, logistics parcel sorting vision systems mostly use barcode scanning devices with constant resolution, which are difficult to handle complex working conditions such as black light absorption, film reflection, and soiled barcodes. Furthermore, they lack dynamic budget management for visual processing time, resulting in timeouts for complex parcels and wasted time for simple parcels, making it difficult to improve overall throughput. Therefore, it is necessary to build a sorting system that can make hierarchical supplementary lighting decisions based on light absorption characteristic coefficients, achieve partitioned differential optimization through grid division difference comparison method, and establish a global time balance to degrade the supplementary lighting strategy, so as to improve the barcode recognition rate of complex parcels and the overall throughput of the sorting line.

[0022] This embodiment provides a fully automated high-speed sorting device for logistics parcels based on visual recognition. The device uses V-shaped guide plates and flexible pressure rollers to orient and center the logistics parcels, ensuring the effective orientation and stable posture of the barcode surface. This provides a standardized material basis for visual recognition, improving the success rate of visual acquisition and image quality. The device also uses supplementary lighting to adaptively supplement the logistics parcels based on target lighting decisions, addressing complex conditions such as black light absorption, reflective coatings, and damaged barcodes, thereby improving barcode imaging. To improve quality and recognition accuracy, the device also acquires high-resolution visual image data through a main vision camera to obtain clear barcode details, thereby improving decoding success rate. The high-speed visual recognition sorting system in the device achieves differentiated visual data acquisition through preview image size evaluation, realizes supplementary lighting mode switching and differential optimization through light absorption characteristic coefficient and grid division difference comparison method, and establishes a closed-loop optimization from visual recognition to time scheduling through global time balance for global overall planning and degraded scheduling, thereby significantly improving the recognition rate of complex package barcodes and the overall throughput of the sorting line.

[0023] See Figure 1 This invention provides a fully automated high-speed parcel sorting device based on visual recognition, comprising: The system consists of: 1. Main conveyor belt; 2. V-shaped guide plate; 3. Flexible pressure roller; 4. Pre-triggered light curtain sensor; 5. Auxiliary area array camera; 6. Fill light; 7. Main vision camera; 8. Push rod; 9. Slewing bearing; 10. Drive wheel; 11. Driven wheel; 12. Frame; 13. Vision recognition high-speed sorting system. The frame 12 is connected to the main conveyor belt 1, the main conveyor belt 1 is mounted on the frame 12, and the driving wheel 10 and the driven wheel 11 are respectively mounted at both ends of the frame 12 and connected to the main conveyor belt 1. The V-shaped guide plate 2 is fixed above the main conveyor belt 1, and the flexible pressure roller 3 is connected to the V-shaped guide plate 2; The pre-triggered light curtain sensor 4 is mounted on the auxiliary area array camera 5, which is fixed above the main conveyor belt 1. The supplementary light 6 is installed above the main conveyor belt 1 and connected to the main vision camera 7; The slewing bearing 9 is fixed on the frame 12; The push rod 8 is mounted on the slewing bearing 9; The high-speed visual recognition sorting system 13 is mounted on the frame 12 and is electrically connected to the auxiliary area array camera 5, the main visual camera 7, the fill light 6 and the push rod 8 respectively.

[0024] In this embodiment, the main conveyor belt 1 is used to transport logistics packages; V-shaped deflector 2 is used to orient and center logistics packages. Flexible pressure roller 3 is used to smooth out logistics packages; The pre-triggered light curtain sensor 4 is used to trigger the auxiliary area array camera; Auxiliary area array camera 5 is used to acquire preview images; The supplementary light 6 is used to provide supplementary lighting for logistics packages based on target supplementary lighting decisions; The main vision camera 7 is used to acquire visual image data; Push rod 8 is used to push the package into the corresponding sorting exit chute; Slewing bearing 9 is used to provide slewing support for push rod 8; The drive wheel 10 is used to drive the main conveyor belt 1; Driven wheel 11 is used to support and tension the main conveyor belt 1; The frame 12 is used to support the main conveyor belt 1; The high-speed visual recognition sorting system 13 is used to acquire the light absorption characteristic coefficient based on the processed visual data, output the target supplementary lighting decision, perform differential optimization on the output process of the target supplementary lighting decision, and perform downgrade processing on the target supplementary lighting decision based on the global time balance.

[0025] In this embodiment, the orientation alignment refers to guiding the logistics package to the center of the main conveyor belt, making the longitudinal axis of the package parallel to the conveying direction, and centering the package to ensure that its barcode side faces upwards. The electrical connection refers to the method of transmitting signals by establishing an electrical path through a cable.

[0026] In this embodiment, the supplementary lighting includes white LEDs and infrared LEDs. The white LEDs and infrared LEDs are switched according to the target supplementary lighting decision. The white LEDs refer to a ring-shaped white LED light group with a color temperature between 5000K and 6500K, divided into 8 independent control sectors, each sector having 12 LED beads, and using PWM dimming. The infrared LEDs refer to a ring-shaped infrared LED light group with a peak wavelength between 850nm and 940nm, divided into 8 independent control sectors, each sector having 4 infrared LED beads, and using PWM dimming.

[0027] In this embodiment, the auxiliary area scan camera refers to a low-resolution imaging device used for pre-triggered preview image acquisition. This embodiment uses a 2-megapixel global shutter industrial camera with an acquisition resolution set to 512 pixels × 512 pixels, a frame rate set to 30fps, and an exposure time set to 5ms. The main vision camera refers to a line scan camera used for acquiring visual image data. This embodiment uses a 4K resolution line scan camera with a maximum line frequency of 80kHz, supporting external triggering and ROI output.

[0028] Figure 2 This is a schematic diagram of the structure of the high-speed visual recognition sorting system provided by the present invention, as shown below. Figure 2 As shown, the high-speed visual recognition sorting system includes: The visual image acquisition module 101 is used to acquire the preview image, perform size evaluation on the preview image, obtain the size evaluation result, and acquire visual image data based on the size evaluation result; The visual image processing module 102 is used to perform image processing on visual image data to obtain processed visual data. The visual image supplementary lighting module 103 is used to obtain the light absorption characteristic coefficient based on the processed visual data, and output the target supplementary lighting decision based on the light absorption characteristic coefficient. It is also used to perform differential optimization on the output process of the target supplementary lighting decision based on the grid division difference comparison method, and to obtain the proportion of overexposed pixels and perform fusion update on the differential optimization process based on the proportion of overexposed pixels. The sorting time control module 104 is used to obtain the global time balance and downgrade the target illumination decision based on the global time balance.

[0029] In this embodiment, as an optional implementation, the visual image acquisition module 101 acquires the preview image through an auxiliary area array camera, performs edge detection on the preview image to obtain the barcode outline, calculates the area S of the circumscribed rectangle of the barcode outline, compares the area S of the circumscribed rectangle with a preset area threshold S0, and outputs the size evaluation result based on the comparison result, wherein: When S≤S0, the visual image acquisition module 101 outputs the small-sized barcode as the size evaluation result; When S > S0, the visual image acquisition module 101 outputs the large-size barcode as the size evaluation result.

[0030] In this embodiment, the preview image refers to a low-resolution preview image of the logistics package quickly acquired by an auxiliary area array camera. Edge detection refers to the process of identifying the boundary between the barcode and the background in the preview image using the Canny edge detection algorithm. In this embodiment, the Canny low threshold is set to 50 and the high threshold to 150. The barcode outline refers to the smallest circumscribed closed curve surrounding the barcode area, composed of continuous edge pixels after edge detection. The area S of the circumscribed rectangle refers to the pixel area of ​​the smallest circumscribed rectangle of the barcode outline, calculated as S = w × h, where w is the pixel width of the circumscribed rectangle and h... The height of the circumscribed rectangle is the pixel value. The preset area threshold S0 refers to the preset critical area for determining the boundary between small-sized and large-sized barcodes. In this embodiment, S0 is determined by the historical barcode size distribution statistics method. The historical barcode size distribution statistics method includes: collecting the area of ​​the circumscribed rectangle of the barcode outline in the preview image of 10,000 sets of historical sorting tasks, arranging the area distribution set in ascending order, and taking the 50th percentile as S0, resulting in S0 = 250,000 square pixels. The size evaluation result refers to the discrete identifier that characterizes the current barcode size level, including small-sized and large-sized barcodes.

[0031] In this embodiment, the visual image acquisition module 101 performs edge detection and calculates the area of ​​the circumscribed rectangle on the preview image, and compares it with a preset area threshold to determine the size level, so as to classify the barcode size and improve the targeting and efficiency of visual data acquisition.

[0032] In this embodiment, as an optional embodiment, the visual image acquisition module 101 acquires visual image data based on the size evaluation result, wherein: When the size level is large barcode, the acquisition resolution of the main vision camera is set to the first resolution, the ROI area is set to full frame, and the visual image data is acquired through the main vision camera. When the size category is small barcode, the acquisition resolution of the main vision camera is set to the second resolution, the ROI area is set to the barcode prediction area, and the visual image data is acquired through the main vision camera.

[0033] In this embodiment, the first resolution refers to a low acquisition resolution set for large-size barcodes. In this embodiment, the first resolution is set to 2K. The second resolution refers to a standard acquisition resolution set for small-size barcodes. In this embodiment, the second resolution is set to 4K. The full frame refers to the complete and effective photosensitive image area output by the main vision camera without ROI cropping. The ROI area refers to the image area actually output by the main vision camera. The barcode prediction area refers to the rectangular area including the barcode that is predicted based on the center position of the barcode outline in the preview image and the probability distribution of historical barcode occurrences. In this embodiment, the barcode prediction area is set to a rectangular range extending outward from the center of the barcode outline by 20% of the width and height of the circumscribed rectangle.

[0034] In this embodiment, the visual image acquisition module 101 dynamically selects the acquisition resolution and ROI area of ​​the main visual camera based on the size evaluation results, so as to avoid the waste of computing power caused by excessive acquisition in large-size barcode scenarios, and at the same time prevent parsing failure caused by insufficient resolution in small-size barcode scenarios, thereby improving the accuracy and efficiency of visual data acquisition.

[0035] In this embodiment, as an optional embodiment, the visual image processing module 102 performs image processing on the visual image data, specifically: converting the visual image data to grayscale to obtain a grayscale image; performing median filtering on the grayscale image to obtain a denoised image; performing Gamma correction on the denoised image to obtain a corrected image; performing Canny edge enhancement on the corrected image to obtain an edge-enhanced image; and outputting the edge-enhanced image as processed visual data.

[0036] In this embodiment, the grayscale conversion refers to the process of converting color visual image data from the RGB color space to a single-channel grayscale image. This embodiment uses a weighted average method for grayscale conversion. According to the standard weighting formula, Gray = 0.299 × R + 0.587 × G + 0.114 × B, where Gray represents the single-channel grayscale value obtained through weighted calculation, ranging from 0 to 255; R represents the red channel grayscale value of the pixel; G represents the green channel grayscale value of the pixel; and B represents the blue channel grayscale value of the pixel. The median filtering denoising refers to using a 3×3... The process of replacing the center pixel value with the median gray value in the pixel neighborhood to eliminate salt-and-pepper noise; the Gamma correction refers to the process of adjusting the dynamic range of image grayscale through a nonlinear power function; in this embodiment, the Gamma value is set to 0.9; the Canny edge enhancement refers to the process of extracting image edges through Gaussian smoothing, gradient calculation, non-maximum suppression, and dual threshold detection; in this embodiment, the Canny low threshold is set to 30 and the high threshold to 100; the processed visual data refers to the edge-enhanced image obtained after grayscale conversion, denoising, Gamma correction, and edge enhancement.

[0037] In this embodiment, the visual image processing module 102 performs image processing on the visual image data to eliminate random noise, compress the dynamic range, and enhance edge contrast, thereby improving the recognizability of the barcode area against complex material backgrounds.

[0038] Figure 3 This is a schematic diagram illustrating the workflow of the visual image illumination module provided by the present invention, as shown below. Figure 3 As shown, the visual image supplementary lighting module 103 obtains the average gray value Gaavg and the proportion of low gray pixels Rlow based on the processed visual data, and calculates the light absorption characteristic coefficient Rcg based on the average gray value Gaavg, the proportion of low gray pixels Rlow, the weight of average gray value w1, and the weight of low gray pixel proportion w2, and sets Rcg=Gavg×w1+Rlow×w2; The light absorption characteristic coefficient Rcg is compared with the preset light absorption threshold Rc0. Based on the comparison result, the light absorption characteristic coefficient is judged, and the target supplementary lighting decision is output based on the judgment result. When Rcg≥Rc0, the visual image supplementary lighting module 103 determines that the light absorption characteristic coefficient is light absorption and outputs the infrared supplementary lighting mode as the target supplementary lighting decision. When Rcg < Rc0, the visual image supplementary lighting module 103 determines that the light absorption characteristic coefficient is not absorbed and outputs the white light supplementary lighting mode as the target supplementary lighting decision.

[0039] In this embodiment, the average gray value refers to the arithmetic mean of the gray values ​​of all pixels in the processed visual data, and the calculation formula is as follows: in This represents the total number of pixels in the processed visual data. For the first grayscale value of each pixel. This represents the pixel index in the processed visual data, and It is a positive integer. The low gray pixel ratio refers to the proportion of pixels with gray values ​​lower than the preset low gray threshold in the processed visual data to the total number of pixels. The preset low gray threshold is the critical gray value for determining whether a pixel belongs to a low gray dark area. In this embodiment, the preset low gray threshold is determined by the gray quantile method of light-absorbing material. The gray quantile method of light-absorbing material includes: collecting 1000 sets of processed visual data of black light-absorbing wrapping, extracting the gray values ​​of all pixels in each image to obtain a set of gray values ​​of black light-absorbing wrapping, arranging the set of gray values ​​of black light-absorbing wrapping in ascending order, and taking the 80th percentile as the preset low gray threshold, resulting in a preset low gray threshold of 30.

[0040] In this embodiment, the light absorption characteristic coefficient refers to a quantitative index used to characterize the light absorption characteristics of the surface of a logistics package, obtained by combining the average gray value and the proportion of low gray pixels. The average gray value weight w1 refers to the weight coefficient of the average gray value in the calculation of the light absorption characteristic coefficient, and the low gray pixel proportion weight w2 refers to the weight coefficient of the low gray pixel proportion in the calculation of the light absorption characteristic coefficient. In this embodiment, the values ​​of w1 and w2 are determined by the analytic hierarchy process (AHP). The AHP includes: constructing a judgment matrix, comparing the average gray value and the proportion of low gray pixels pairwise, setting the relative importance ratio based on expert experience, obtaining an importance ratio of 2:1 between the average gray value and the proportion of low gray pixels, calculating the feature vector after obtaining the judgment matrix, and performing a consistency check to obtain normalized weights w1=0.667 and w2=0.333. Therefore, w1=0.667 and w2=0.333 are set.

[0041] In this embodiment, the preset light absorption threshold Rc0 refers to the critical value for determining whether the light absorption characteristic coefficient reaches the light absorption standard. In this embodiment, Rc0 is valued by the material light absorption characteristic calibration method. The material light absorption characteristic calibration method includes: collecting processed visual data of 500 groups of black light-absorbing packages and 500 groups of ordinary packages, calculating the light absorption characteristic coefficient Rcg of each group, and taking the minimum Rcg value at the intersection of the two groups as Rc0, thus obtaining Rc0=120. The light absorption characteristic coefficient refers to the light absorption degree of the logistics package surface obtained by comparing the light absorption characteristic coefficient with the preset light absorption threshold, including light absorption and no light absorption. The infrared supplementary lighting mode refers to the supplementary lighting configuration of turning off the white light LED, turning on the infrared LED, and cutting the infrared filter into the camera optical path. The white light supplementary lighting mode refers to the supplementary lighting configuration of turning on the white light LED, turning off the infrared LED, and removing the infrared filter.

[0042] In this embodiment, the visual image supplementary lighting module 103 calculates the light absorption characteristic coefficient by weighting the average gray value and the proportion of low gray pixels, and compares it with the preset light absorption threshold to determine the light absorption situation, so as to switch the appropriate supplementary lighting mode according to the brightness and darkness characteristics of the logistics package surface, thereby improving the barcode imaging quality of black light-absorbing and low-contrast packages, and further improving the barcode recognition accuracy.

[0043] Figure 4 This is a flowchart illustrating the mesh generation difference comparison method provided by the present invention, as shown below. Figure 3 As shown, the visual image supplementary lighting module 103 performs differential optimization on the output process of the target supplementary lighting decision based on the grid partitioning difference comparison method. The grid partitioning difference comparison method includes: Step A01: Divide the processed visual data into grids to obtain gridded visual data, and calculate the local absorption coefficient Rcj of each grid in the gridded visual data. Step A02: Obtain the maximum absorption value Rcmax and the minimum absorption value Rcmin based on the local absorption coefficient Rcj; Step A03: Calculate the grid absorption non-uniformity ΔRc based on the maximum absorption value Rcmax and the minimum absorption value Rcmin, where ΔRc = Rcmax - Rcmin; Step A04: Compare the grid absorption non-uniformity ΔRc with the preset non-uniformity threshold ΔRc0, determine the grid absorption state based on the comparison result, and perform differential optimization on the target supplementary lighting decision output process based on the determination result, wherein: When ΔRc < ΔRc0, the visual image supplementary lighting module 103 determines that the grid light absorption state is uniform and does not perform differential optimization on the output process of the target supplementary lighting decision; When ΔRc≥ΔRc0, the visual image supplementary lighting module 103 determines that the grid light absorption state is uneven, and performs differential optimization on the output process of the target supplementary lighting decision: If the target illumination decision is white light illumination mode, the white light illumination mode is processed by PWM reduction to obtain white light reduction mode, and the white light reduction mode is output as the target illumination decision. If the target illumination decision is infrared illumination mode, and image fusion processing is performed on the infrared illumination mode to obtain the infrared image fusion mode, then the infrared image fusion mode is output as the target illumination decision.

[0044] In this embodiment, the grid division refers to uniformly dividing the processed visual data into a rectangular grid of M rows and N columns on a two-dimensional plane. In this embodiment, M=8, N=8, and there are a total of 64 grids. The local absorption coefficient Rcj refers to the arithmetic mean of the gray values ​​of all pixels in the j-th grid, and the calculation formula is as follows: in Let j be the total number of pixels in the j-th grid. For the j-th grid, the first... The grayscale value of a pixel, where j represents the grid number and is a positive integer, j=1, 2, 3, ... This indicates the order of pixels within the j-th grid. The maximum light absorption value Rcmax refers to the maximum value among the local light absorption coefficients of all M×N grids, and the minimum light absorption value Rcmin refers to the minimum value among the local light absorption coefficients of all M×N grids. The grid light absorption non-uniformity refers to the difference between the maximum and minimum light absorption values, which is used to characterize the degree of light and dark difference between different grid areas on the surface of the logistics package. The value range is from 0 to 255. The larger the value, the more severe the light and dark difference on the surface.

[0045] In this embodiment, the preset non-uniformity threshold ΔRc0 refers to a preset value for judging the light absorption state of the grid. In this embodiment, ΔRc0 is obtained by statistical method of black and white coexisting package samples. The statistical method of black and white coexisting package samples includes: collecting 1000 groups of package samples with black and white areas on the surface at the same time, performing steps A01 to A03 on the processed visual data of each sample to calculate the grid light absorption non-uniformity ΔRc, obtaining the ΔRc distribution set, arranging the ΔRc distribution set in ascending order, and taking the 90th percentile as ΔRc0, resulting in ΔRc0=80.

[0046] In this embodiment, the grid light absorption state refers to the determination result of the uniformity of the light and dark distribution on the surface of the logistics package obtained by comparing the grid light absorption non-uniformity with the preset non-uniformity threshold, including uniformity and non-uniformity. The PWM reduction processing refers to the operation of reducing and adjusting the PWM duty cycle of the white LED in the sector corresponding to the high light absorption grid. The image fusion processing refers to the operation of weighted fusion enhancement of the infrared image and the visible light image in the region corresponding to the low light absorption grid.

[0047] In this embodiment, the PWM reduction process is as follows: A spatial mapping relationship between the grid and the supplementary ring LED sectors is established based on the grid division results. This spatial mapping relationship refers to matching the planar coordinate angles of each grid with the 8 sectors of the ring LED, using the center point of the processed visual data as the origin, to obtain the LED sector number corresponding to each grid. Grids with a local light absorption coefficient Rcj greater than a preset high light absorption threshold are identified as high-light-absorbing grids. The preset high light absorption threshold is the critical value for determining whether a grid belongs to a high-light-absorbing region. In this embodiment, the preset high light absorption threshold is set to 200. The LED sector number corresponding to the high-light-absorbing grid is obtained. A light reduction command is sent to the PWM driver chip of the corresponding sector via the I2C bus to reduce the PWM duty cycle of the white LED in the sector from the original duty cycle D0 to the adjusted duty cycle D1. D1 is set to D0 × (1-γ), where γ is the light reduction ratio coefficient. In this embodiment, γ is set to 0.3, i.e., a 30% reduction in duty cycle.

[0048] In this embodiment, the infrared image refers to image data acquired through an infrared LED, and the visible light image refers to image data in the conventional visible light band acquired through white LED illumination.

[0049] In this embodiment, the image fusion processing refers to the process of adding a weighted fusion enhancement operation between the infrared image and the visible light image corresponding to the low-absorption grid region to the infrared supplementary lighting mode. Specifically, this process involves acquiring a visible light image (Ivis) and an infrared image (Iir) for the same field of view, with a 5ms acquisition time interval between the two images. Grids with a local absorption coefficient Rcj less than a preset low-absorption threshold are designated as low-absorption grids. This preset low-absorption threshold is a critical value used to determine whether a grid belongs to a low-absorption region. In this embodiment, it is set... A preset low absorption threshold of 50 is set. The corresponding regions of the low absorption grids in the visible light image Ivis and the infrared image Iir are extracted to obtain the visible light sub-image Ivisroi and the infrared sub-image Iirroi of the low absorption region. The visible light sub-image and the infrared sub-image of the low absorption region are then weighted and fused to obtain the fused sub-image Ifused. Ifused = α × Iirroi + (1-α) × Ivisroi is set, where α is the infrared weight coefficient. In this embodiment, α = 0.7 is set. The positions of the corresponding low absorption grids in the visible light image Ivis are replaced according to the fused sub-image Ifused.

[0050] In this embodiment, visual data of 1000 sets of highly reflective packaged samples after processing was collected, the average gray value distribution of each grid was statistically analyzed, and the 90th percentile was taken as the preset high light absorption threshold, resulting in a preset high light absorption threshold of 200. Furthermore, by collecting 500 sets of locally overexposed scenes, the reduction in the proportion of overexposed pixels was tested when the light reduction ratio coefficients were 0.2, 0.3, 0.4, and 0.5, respectively. The minimum light reduction ratio coefficient that reduced the proportion of overexposed pixels to below 5% was selected, resulting in γ=0.3. Additionally, visual data of 1000 sets of black light-absorbing packaged samples after processing was collected, the average gray value distribution of each grid was statistically analyzed, and the 10th percentile was taken as the preset low light absorption threshold, resulting in a preset low light absorption threshold of 50.

[0051] In this embodiment, the infrared weighting coefficient α is selected using a decoding success rate optimization method. The decoding success rate optimization method includes: collecting 500 sets of black package samples, testing the barcode decoding success rate when the infrared weighting coefficient α is 0.5, 0.6, 0.7, 0.8, and 0.9 respectively, and selecting the weighting coefficient that achieves the highest decoding success rate and retains sufficient visible light details, thus obtaining α=0.7.

[0052] In this embodiment, the visual image supplementary lighting module 103 identifies surface light absorption unevenness through grid division difference comparison method and performs partition differential supplementary lighting optimization, so as to reduce the light in the white light overexposed area and enhance the infrared underexposed area in the scene of coexistence of black and white and local reflection, thereby improving the recognition accuracy of logistics packages and further improving the sorting throughput.

[0053] In this embodiment, the visual image supplementary lighting module acquires the overexposed pixel ratio Z, compares the overexposed pixel ratio Z with a preset overexposed pixel ratio Z0, determines the state of the overexposed pixel ratio based on the comparison result, and performs a fusion update on the differential optimization process based on the determination result, wherein: When Z≤Z0, the visual image supplementary lighting module determines that the overexposed pixel ratio is normal and does not perform fusion update in the differential optimization process; When Z > Z0, the visual image illumination module determines that the proportion of overexposed pixels is abnormal and performs a fusion update on the differential optimization process: If the target illumination decision is white light illumination mode, the PWM reduction process is reduced a second time; If the target illumination decision is infrared illumination mode, the image fusion process is adjusted a second time.

[0054] In this embodiment, the overexposed pixel ratio Z refers to the proportion of pixels with a grayscale value reaching the maximum quantization value of 255 in the processed visual data to the total number of pixels. The preset overexposed pixel ratio Z0 refers to a preset value for judging the state of the overexposed pixel ratio. In this embodiment, Z0 is determined by the overexposed scene quantile method. The overexposed scene quantile method includes: collecting 2000 sets of processed visual data after differential optimization, statistically analyzing the overexposed pixel ratio Z of each image to obtain an overexposed pixel ratio distribution set, arranging the overexposed pixel ratio distribution set in ascending order, and taking the 85th percentile as Z0, resulting in Z0=3%. The state of the overexposed pixel ratio refers to the image overexposed degree judgment result obtained by comparing the overexposed pixel ratio with the preset overexposed pixel ratio, including normal and abnormal.

[0055] In this embodiment, the secondary reduction process is as follows: the PWM duty cycle of the white LED in the sector is adjusted according to the secondary reduction ratio coefficient β to obtain the duty cycle D2 after the secondary adjustment. D2 is set to D1×(1-β). In this embodiment, β is set to 0.2. In this embodiment, 500 sets of differentially optimized images with residual overexposure are collected, and the reduction rate of overexposed pixel ratio and barcode decoding success rate are tested when the secondary reduction ratio coefficient β is 0.1, 0.2, 0.3, and 0.4, respectively. The minimum β value that reduces the overexposed pixel ratio to below 1% and the decoding success rate is not less than 95% is selected, and β=0.2 is obtained.

[0056] In this embodiment, the secondary adjustment process is as follows: the fusion weight of the infrared image and the visible light image of the effective overexposed area is reduced and adjusted to obtain the adjusted infrared weight coefficient αadj. αadj is set to α×(1-β), and the fused sub-image of the effective overexposed area is recalculated based on the adjusted infrared weight coefficient αadj.

[0057] In this embodiment, the visual image supplementary lighting module 103 judges the state of the overexposed pixel ratio. When there is still local overexposure after differential optimization, it performs secondary reduction and secondary adjustment according to the type of target supplementary lighting decision, so as to eliminate the damage of residual overexposure to the barcode edge information, thereby further improving the barcode recognition accuracy of complex material wrapping.

[0058] In this embodiment, as an optional implementation, the sorting time control module 104 acquires the global time balance B and sets... ,in, This refers to the barcode processing time of the previous logistics package. Using the barcode processing time Tpred as the base time, the processing time Tpred is obtained, and then the data is processed based on the barcode processing time Tpred and the base time. The global time balance B is used to downgrade the target illumination decision, where: when At that time, the target illumination decision is downgraded: If the target illumination decision is white light illumination mode, only a single frame of white light is retained for illumination; If the target illumination decision is infrared illumination mode, only a single frame of red light is retained for illumination. Otherwise, the target illumination decision is not downgraded, and the global time balance B is updated.

[0059] In this embodiment, the barcode processing time refers to the estimated visual processing time required for the current logistics package. This embodiment obtains the time by querying a preset time mapping table using the size assessment result and light absorption coefficient as a joint index. The preset time mapping table is a pre-established data retrieval table using the size assessment result and light absorption coefficient range as the joint index key, and the historical average barcode processing time as the associated content. The baseline time refers to the upper limit of the visual processing time per package required by the pre-set throughput of the sorting line. The phrase "preserving a single white light frame for supplemental lighting" means performing only single-frame white light acquisition without enabling the white light reduction mode. The phrase "preserving a single red light frame for supplemental lighting" means performing only single-frame infrared acquisition without enabling the infrared image fusion mode. The phrase "updating the global time balance B" means updating the global time balance B after the current logistics package is processed, based on... The process of recalculating the time balance involves B1 being the updated global time balance, which is then used as the global time balance.

[0060] In this embodiment, the sorting time control module 104 uses global time balance to coordinate visual processing time, so as to accumulate time surplus in simple package processing and withdraw time surplus in complex package processing, thereby improving the overall throughput of the logistics package sorting line.

[0061] The above are merely specific embodiments of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A fully automated high-speed parcel sorting device based on vision recognition, characterized in that, Includes main conveyor belt, V-shaped guide vane, flexible pressure roller, pre-triggered light curtain sensor, auxiliary area scan camera, supplementary light, main vision camera, push rod, slewing bearing, drive wheel, driven wheel, frame, and vision recognition high-speed sorting system, among which: The frame is connected to the main conveyor belt, the main conveyor belt is mounted on the frame, and the driving wheel and the driven wheel are respectively mounted at both ends of the frame and connected to the main conveyor belt. The V-shaped guide plate is fixed above the main conveyor belt, and the flexible pressure roller is connected to the V-shaped guide plate; The pre-triggered light curtain sensor is mounted on the auxiliary area scan camera, which is fixed above the main conveyor belt. The supplementary light is installed above the main conveyor belt and connected to the main vision camera; The slewing bearing is fixed to the frame; The push rod is mounted on the slewing bearing; The high-speed visual recognition sorting system is mounted on a frame and is electrically connected to the auxiliary area array camera, the main visual camera, the fill light, and the push rod.

2. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 1, characterized in that, The main conveyor belt is used to transport logistics packages; V-shaped deflectors are used to orient and center logistics packages. Flexible pressure rollers are used to smooth out logistics packages; A pre-triggered light curtain sensor is used to trigger the auxiliary area scan camera; An auxiliary area scan camera is used to acquire preview images; Supplemental lighting is used to provide supplemental lighting for logistics packages based on target lighting decisions; The main vision camera is used to acquire visual image data; Push rods are used to push packages into the corresponding sorting exit chute. Slewing bearing, used to provide slewing support for push rods; The drive pulley is used to drive the main conveyor belt; Driven pulleys are used to support and tension the main conveyor belt; The frame is used to support the main conveyor belt; The high-speed sorting system based on visual recognition is used to obtain the light absorption characteristic coefficient based on the processed visual data, output the target supplementary lighting decision, perform differential optimization on the output process of the target supplementary lighting decision, and downgrade the target supplementary lighting decision based on the global time balance.

3. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 1, characterized in that, The supplementary lighting includes white LEDs and infrared LEDs. The white LEDs and infrared LEDs are switched according to the target supplementary lighting decision. The white LEDs refer to a ring-shaped white LED light group with a color temperature between 5000K and 6500K, divided into 8 independent control sectors, each sector with 12 LED beads, and using PWM dimming. The infrared LEDs refer to a ring-shaped infrared LED light group with a peak wavelength between 850nm and 940nm, divided into 8 independent control sectors, each sector with 4 infrared LED beads, and using PWM dimming.

4. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 1, characterized in that, The visual recognition high-speed sorting system includes: The visual image acquisition module is used to acquire the preview image, evaluate the size of the preview image, obtain the size evaluation result, and acquire the visual image data based on the size evaluation result; The visual image processing module is used to process visual image data to obtain processed visual data. The visual image supplementary lighting module is used to obtain the light absorption characteristic coefficient based on the processed visual data, and output the target supplementary lighting decision based on the light absorption characteristic coefficient. It is also used to perform differential optimization on the output process of the target supplementary lighting decision based on the grid division difference comparison method. Furthermore, it is used to obtain the proportion of overexposed pixels and perform fusion update on the differential optimization process based on the proportion of overexposed pixels. The sorting time control module is used to obtain the global time balance and downgrade the target illumination decision based on the global time balance.

5. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 4, characterized in that, The visual image acquisition module acquires the preview image using an auxiliary area scan camera, performs edge detection on the preview image to obtain the barcode outline, calculates the area S of the bounding rectangle of the barcode outline, compares the area S with a preset area threshold S0, and outputs the size evaluation result based on the comparison result, wherein: When S≤S0, the visual image acquisition module outputs the small-sized barcode as the size evaluation result; When S > S0, the visual image acquisition module outputs the large-size barcode as the size evaluation result.

6. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 4, characterized in that, The visual image acquisition module acquires visual image data based on the size evaluation results, wherein: When the size level is large barcode, the acquisition resolution of the main vision camera is set to the first resolution, the ROI area is set to full frame, and the visual image data is acquired through the main vision camera. When the size category is small barcode, the acquisition resolution of the main vision camera is set to the second resolution, the ROI area is set to the barcode prediction area, and the visual image data is acquired through the main vision camera.

7. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 4, characterized in that, The visual image processing module performs image processing on the visual image data, specifically: converting the visual image data to grayscale to obtain a grayscale image; performing median filtering on the grayscale image to obtain a denoised image; performing Gamma correction on the denoised image to obtain a corrected image; performing Canny edge enhancement on the corrected image to obtain an edge-enhanced image; and outputting the edge-enhanced image as processed visual data.

8. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 4, characterized in that, The visual image supplementary lighting module obtains the average gray value Gaavg and the proportion of low gray pixels Rlow based on the processed visual data, and calculates the light absorption characteristic coefficient Rcg based on the average gray value Gaavg, the proportion of low gray pixels Rlow, the weight of average gray value w1, and the weight of low gray pixels w2, setting Rcg=Gavg×w1+Rlow×w2. The light absorption characteristic coefficient Rcg is compared with the preset light absorption threshold Rc0. Based on the comparison result, the light absorption characteristic coefficient is judged, and the target supplementary lighting decision is output based on the judgment result. When Rcg≥Rc0, the visual image supplementary lighting module determines that the light absorption characteristic coefficient is light absorption and outputs the decision to enable infrared supplementary lighting mode as the target supplementary lighting decision. When Rcg < Rc0, the visual image supplementary lighting module determines that the light absorption characteristic coefficient is not absorbed, and outputs the white light supplementary lighting mode as the target supplementary lighting decision.

9. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 4, characterized in that, The visual image supplementary lighting module performs differential optimization on the output process of target supplementary lighting decision based on the grid partitioning difference comparison method, which includes: Step A01: Divide the processed visual data into grids to obtain gridded visual data, and calculate the local absorption coefficient Rcj of each grid in the gridded visual data. Step A02: Obtain the maximum absorption value Rcmax and the minimum absorption value Rcmin based on the local absorption coefficient Rcj; Step A03: Calculate the grid absorption non-uniformity ΔRc based on the maximum absorption value Rcmax and the minimum absorption value Rcmin, where ΔRc = Rcmax - Rcmin; Step A04: Compare the grid absorption non-uniformity ΔRc with the preset non-uniformity threshold ΔRc0, determine the grid absorption state based on the comparison result, and perform differential optimization on the target supplementary lighting decision output process based on the determination result, wherein: When ΔRc < ΔRc0, the visual image supplementary lighting module determines that the grid light absorption state is uniform and does not perform differential optimization on the output process of the target supplementary lighting decision; When ΔRc≥ΔRc0, the visual image supplementary lighting module determines that the grid light absorption state is non-uniform, and performs differential optimization on the output process of the target supplementary lighting decision: If the target illumination decision is white light illumination mode, the white light illumination mode is processed by PWM reduction to obtain white light reduction mode, and the white light reduction mode is output as the target illumination decision. If the target illumination decision is infrared illumination mode, and image fusion processing is performed on the infrared illumination mode to obtain the infrared image fusion mode, and the infrared image fusion mode is output as the target illumination decision; The visual image supplementary lighting module acquires the overexposed pixel ratio Z, compares Z with a preset overexposed pixel ratio Z0, determines the state of the overexposed pixel ratio based on the comparison result, and performs a fusion update on the differential optimization process based on the determination result, wherein: When Z≤Z0, the visual image supplementary lighting module determines that the overexposed pixel ratio is normal and does not perform fusion update in the differential optimization process; When Z > Z0, the visual image illumination module determines that the proportion of overexposed pixels is abnormal and performs a fusion update on the differential optimization process: If the target illumination decision is white light illumination mode, the PWM reduction process is reduced a second time; If the target illumination decision is infrared illumination mode, the image fusion process is adjusted a second time.

10. The fully automated high-speed sorting device for logistics parcels based on visual recognition according to claim 4, characterized in that, The sorting time control module obtains the global time balance B and sets... ,in, This refers to the barcode processing time of the previous logistics package. Using the barcode processing time Tpred as the base time, the processing time Tpred is obtained, and then the data is processed based on the barcode processing time Tpred and the base time. The global time balance B is used to downgrade the target illumination decision, where: when At that time, the target illumination decision is downgraded: If the target illumination decision is white light illumination mode, only a single frame of white light is retained for illumination; If the target illumination decision is infrared illumination mode, only a single frame of red light is retained for illumination. Otherwise, the target illumination decision is not downgraded, and the global time balance B is updated.

Images