Mobile phone shell appearance inspection system and inspection method

Abstract

The present invention relates to a mobile phone shell appearance detection system and detection method, and belongs to the technical field of appearance defect recognition. The system of the present invention includes a loading station and a cleaning station for the mobile phone shell, as well as a back panel detection station, a middle frame detection station, a curved surface detection station and a material sorting device for detecting the mobile phone shell. It uses a highly reliable distributed architecture, has a low probability of software failure and is easy to maintain, and takes less time for maintenance / upgrade operations; the functional components are modularized, flexibly called, and support custom workflows, and can quickly adapt to different detection scenarios such as quality inspection, out / in material inspection, so as to maximize the use of equipment resources; the method of the present invention integrates an algorithm training platform to help the existing product detection algorithm to continuously evolve and update; using an AI algorithm, it can realize compatible detection of subsequent versions of products with the same material and the same defect characteristics; built-in massive data capture nodes help to analyze product and process data. The present invention can be widely used in appearance defect recognition occasions.

Description

Technical Field

[0001] The invention relates to a mobile phone shell appearance detection system and a detection method, belonging to the technical field of appearance defect recognition. Background Art

[0002] As a product for protecting and decorating smartphones, mobile phone cases have developed rapidly in recent years, with rapid replacement and product upgrades. However, there are still some technical problems in the appearance inspection of mobile phone cases that have not been solved. Existing inspection equipment usually uses machine vision technology for defect detection. Automatic appearance inspection by machines generally adopts traditional visual algorithm methods, mainly including area analysis, color extraction, difference comparison, grayscale analysis, Blob calculation, etc. However, these methods are highly dependent on the objects being inspected, and have problems such as low detection efficiency, high false detection rate, and the need for targeted encoding for different targets. The most important function that requires the reliance on hardware modules in machine vision is image acquisition, which has the problem of low detection efficiency. Summary of the invention

[0003] In view of the above-mentioned defects in the prior art, the present invention proposes a mobile phone shell appearance detection system and detection method, which can quickly and accurately detect the quality of the mobile phone shell and improve the detection accuracy.

[0004] The mobile phone shell appearance inspection system of the present invention comprises a loading station and a cleaning station for the mobile phone shell, and a back panel inspection station, a middle frame inspection station, a curved surface inspection station and a material sorting device for inspecting the mobile phone shell, wherein:

[0005] The feeding station includes two automatic feeding components and a feeding robot arranged on one side of the automatic feeding component. The feeding robot can automatically feed materials in two ways at the same time.

[0006] The cleaning station includes a wiping device and a dust removal device. After loading, the mobile phone shell conveyed by the assembly line I contacts the surface of the mobile phone shell through the wiping device, and the dust after cleaning is removed by the dust removal device.

[0007] The back panel inspection station includes a transfer carrier I and an inspection camera I for inspecting the back panel glass of the mobile phone shell. After cleaning, the mobile phone shell conveyed by the assembly line II is conveyed to the transfer carrier I, and is moved to the inspection camera I through the transfer carrier I. The inspection camera I is set on the back of the transfer carrier I. The inspection camera I transmits the captured image to the visual inspection system, and the visual inspection system identifies the defect information such as scratches, bumps, and discoloration on the back panel;

[0008] The middle frame inspection station includes a transfer carrier II and an inspection camera II for inspecting the middle frame of the mobile phone shell. After the back panel inspection, the mobile phone shell conveyed by the assembly line III is conveyed to the transfer carrier II and moved to the inspection camera II through the transfer carrier II. The inspection camera II is set around the transfer carrier II. The inspection camera II transmits the captured image to the visual inspection system, and the visual inspection system identifies the defect information such as scratches, bumps, and discoloration on the middle frame;

[0009] The curved surface inspection station includes a transfer carrier III and an inspection camera III for inspecting the curved surface of the mobile phone shell. After the back panel inspection, the mobile phone shell is transported by the assembly line IV to the transfer carrier III and moved to the inspection camera III through the transfer carrier III. The inspection camera III is set around the transfer carrier III. The inspection camera III transmits the captured image to the visual inspection system, and the visual inspection system identifies the scratches, bumps, discoloration and other defect information on the curved surface;

[0010] The material unloading and sorting device includes two-way automatic unloading components and a unloading robot arranged on one side of the automatic unloading components. The unloading robot automatically unloads materials in two ways at the same time; according to the inspection summary results of the visual inspection system, defective products are automatically screened out.

[0011] Preferably, the loading robot comprises a lifting mechanism, a flipping mechanism and a clamping mechanism; the flipping mechanism is arranged on the lifting mechanism; the clamping mechanism is arranged on the flipping mechanism, and the flipping mechanism can drive the clamping mechanism to flip.

[0012] Preferably, the cleaning station comprises a cover body with an FFU pressurized above, and a wiping device and a dust removal device located inside the cover body;

[0013] The wiping device includes a dust-free cloth for wiping the back panel and the middle frame. The dust-free cloth is divided into two types: dry wiping cloth and wet wiping cloth;

[0014] The dust removal device includes an ultrasonic cleaning device, an FFU fan and a vacuum device. The ultrasonic cleaning device removes fine particles, dust and residual alcohol and discharges them through the bottom of the base plate. The FFU fan and the vacuum device clean the ultrasonic cleaning device separately.

[0015] Preferably, the assembly line is provided with a plurality of transfer carriers at intervals along the conveying direction, and the transfer carriers are used to transfer the mobile phone shells between adjacent inspection stations; each inspection station includes a blocking and lifting mechanism; the blocking and lifting mechanism is used to block the transfer carriers conveyed to each inspection station to the inspection station, and lift the transfer carriers by at least one height of the transfer carriers, and move them to the starting position of the assembly line via the recovery assembly line;

[0016] The cleaning station, back panel inspection station, middle frame inspection station, and curved surface inspection station of the assembly line correspond to assembly lines Ⅰ to Ⅳ respectively; the transfer carriers correspond to transfer carriers Ⅰ to Ⅳ in the cleaning station, back panel inspection station, middle frame inspection station, and curved surface inspection station respectively.

[0017] Preferably, the material unloading and sorting device further comprises a film sticking mechanism, the material feeding and transporting module moves the product to the film sticking position, and the film sticking mechanism starts to stick the film;

[0018] The unloading robot is used to unload the materials. The NG products are finally discharged from the NG flow line, and the OK products are placed in the tray for unloading.

[0019] The beneficial effects of the system of the present invention are: (1) using a highly reliable distributed architecture, the probability of software failure is low and maintenance is easy, and maintenance / upgrade operations take less time.

[0020] (2) The functional components are modularized and flexibly called to support customized workflows. They can quickly adapt to different inspection scenarios such as quality inspection and outgoing / incoming material inspection, thereby maximizing the use of equipment resources and saving costs.

[0021] The detection method of the mobile phone shell appearance detection system of the present invention comprises the following steps:

[0022] S1: The mobile phone shell is automatically loaded under the control of two feeding stations with a frequency of 2pcs per test;

[0023] S2: The loading robot delivers the mobile phone shell to the cleaning station, where it uses dry and wet wipes to wipe the back panel and middle frame respectively, and uses ultrasonic cleaning equipment to clean fine particles, dust and residual alcohol;

[0024] S3: Use the transfer carrier on the production line to inspect the back panel, middle frame and curved surface in turn at an average inspection speed of 5s / 1pcs. The detected images are transmitted to the visual inspection system. The visual inspection system uses the following algorithm:

[0025] S31: feature extraction, consisting of two 3x3 convolutional layers RELU and a 2x2 maxpooling layer;

[0026] S32: Feature sampling, which is composed of a deconvolution layer, a feature concatenation layer, and two 3x3 convolutional layers ReLU repeatedly;

[0027] S33: Statistics and summary: analyzing and processing the collected images, screening out defective images, uploading them to the server for statistics and summary, which is convenient for management;

[0028] S34: Autonomous learning, using traditional algorithms and AI hybrid detection mode to achieve autonomous learning without human intervention;

[0029] S4: The loading robot moves the mobile phone shell to the unloading and sorting device, and the NG flow line returns the NG products for secondary cleaning and retesting; the OK products start to be filmed through the film-sticking mechanism and are placed in the tray for unloading.

[0030] Preferably, the visual inspection system in step S3 includes an image processing module and an image recognition module, wherein:

[0031] The image processing module is used to filter and segment the image data to obtain the image data of the area to be inspected of the product back panel, middle frame and curved surface; and to identify the image data of the area to be inspected to determine the image data of the product target inspection area;

[0032] The image recognition module is used to classify and identify the image data of the target detection area using a platform that has been integrated with algorithm training, and obtain information on scratches, bruises, and discoloration on the surface so that they can be removed through the material sorting device.

[0033] The beneficial effects of the method of the present invention are: (1) integrating an algorithm training platform to help the existing product detection algorithm to continuously evolve and update.

[0034] (2) Using AI algorithms, compatible detection of subsequent versions of products with the same material and the same defect characteristics can be achieved.

[0035] (3) Built-in massive data capture nodes to help with product and process data analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a structural schematic diagram of the system of the present invention.

[0037] Figure 2 It is a structural diagram of the cleaning station.

[0038] Figure 3 It is a structural schematic diagram of the material sorting device.

[0039] Figure 4 It is the algorithm principle diagram of the method of the present invention.

[0040] In the figure: 1. Loading station; 2. Loading robot; 3. Cleaning station; 31. Dry wipe cloth; 32. Wet wipe cloth; 33. Ultrasonic cleaning equipment; 4. Assembly line I; 5. Back panel inspection station; 6. Assembly line II; 7. Middle frame inspection station; 8. Assembly line III; 9. Arc surface inspection station; 10. Assembly line IV; 11. Unloading and sorting device; 111. NG assembly line; 112. Handling module; 113. Film pasting mechanism; 12. Unloading robot. DETAILED DESCRIPTION

[0041] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0042] Embodiment 1:

[0043] like Figure 1 As shown, the mobile phone shell appearance inspection system of the present invention includes a loading station 1 and a cleaning station 3 for the mobile phone shell, as well as a back panel inspection station 5, a middle frame inspection station 7, a curved surface inspection station 9 and a blanking and sorting device 11 for inspecting the mobile phone shell, wherein:

[0044] The feeding station 1 includes two automatic feeding components and a feeding robot 2 arranged on one side of the automatic feeding component. The feeding robot 2 can automatically feed two paths at the same time.

[0045] Cleaning station 3, including wiping equipment and dust removal device, the mobile phone shell conveyed by assembly line I4 after loading, respectively contacts the surface of the mobile phone shell through the wiping device, and removes the dust after cleaning through the dust removal device;

[0046] The back panel inspection station 5 includes a transfer carrier Ⅰ and an inspection camera Ⅰ for inspecting the back panel glass of the mobile phone shell; the mobile phone shell transported by the assembly line Ⅱ6 after cleaning is transported to the transfer carrier Ⅰ, and is moved to the inspection camera Ⅰ through the transfer carrier Ⅰ. The inspection camera Ⅰ is set on the back of the transfer carrier Ⅰ. The inspection camera Ⅰ transmits the captured image to the visual inspection system, and the visual inspection system identifies the defect information such as scratches, bumps, and discoloration on the back panel;

[0047] The middle frame inspection station 7 includes a transfer carrier II and an inspection camera II for inspecting the middle frame of the mobile phone shell. The mobile phone shell conveyed by the assembly line III8 after the back panel inspection is conveyed to the transfer carrier II, and moved to the inspection camera II through the transfer carrier II. The inspection camera II is arranged around the transfer carrier II. The inspection camera II transmits the captured image to the visual inspection system, and the visual inspection system identifies the defect information such as scratches, bumps, and discoloration on the middle frame;

[0048] The curved surface inspection station 9 includes a transfer carrier III and an inspection camera III for inspecting the curved surface of the mobile phone shell. The mobile phone shell delivered by the assembly line IV10 after the back panel inspection is delivered to the transfer carrier III and moved to the inspection camera III through the transfer carrier III. The inspection camera III is arranged around the transfer carrier III. The inspection camera III delivers the captured image to the visual inspection system, and the visual inspection system identifies the defect information such as scratches, bumps, and discoloration on the curved surface;

[0049] The material unloading and sorting device 11 includes two-way automatic unloading components and a material unloading robot 12 arranged on one side of the automatic unloading components. The material unloading robot 12 automatically unloads materials in two ways at the same time; and automatically selects out defective products according to the inspection summary results of the visual inspection system.

[0050] Preferably, the feeding robot 2 comprises a lifting mechanism, a flipping mechanism and a clamping mechanism; the flipping mechanism is arranged on the lifting mechanism; the clamping mechanism is arranged on the flipping mechanism, and the flipping mechanism can drive the clamping mechanism to flip.

[0051] like Figure 2 As shown, the cleaning station 3 includes a cover body with FFU pressurized above, and a wiping device and a dust removal device located inside the cover body;

[0052] The wiping device includes a dust-free cloth for wiping the back plate and the middle frame. The dust-free cloth is divided into two types: a dry wiping cloth 31 and a wet wiping cloth 32;

[0053] The dust removal device includes an ultrasonic cleaning device 33, and an FFU fan and a vacuum device. The ultrasonic cleaning device 33 removes fine particles, dust and residual alcohol and discharges them through the bottom of the base plate. The FFU fan and the vacuum device clean the ultrasonic cleaning device 33 separately.

[0054] Preferably, the assembly line is provided with a plurality of transfer carriers at intervals along the conveying direction, and the transfer carriers are used to transfer the mobile phone shells between adjacent inspection stations; each inspection station includes a blocking and lifting mechanism; the blocking and lifting mechanism is used to block the transfer carriers conveyed to each inspection station to the inspection station, and lift the transfer carriers by at least one height of the transfer carriers, and move them to the starting position of the assembly line via the recovery assembly line;

[0055] The assembly lines correspond to assembly lines Ⅰ4 to Ⅳ10 in the cleaning station 3, back panel inspection station 5, middle frame inspection station 7, and arc surface inspection station 9 respectively; the transfer carriers correspond to transfer carriers Ⅰ to transfer carriers Ⅳ in the cleaning station 3, back panel inspection station 5, middle frame inspection station 7, and arc surface inspection station 9 respectively.

[0056] like Figure 3 As shown, the material sorting device 11 further includes a film pasting mechanism 113. The material feeding and transporting module 112 moves the product to the film pasting position, and the film pasting mechanism 113 starts film pasting;

[0057] The unloading robot 12 is used to unload the materials, and the NG products are finally unloaded from the NG flow line 111, and the OK products are placed in the tray for unloading.

[0058] Feeding equipment: Uses two-way automatic feeding by manipulator.

[0059] Cleaning station 3. The wiping equipment and dust removal device inside can automatically clean the dust and dirt on the phone case, which can effectively avoid the misjudgment of the algorithm due to interference such as dust and dirt.

[0060] Back panel inspection station 5. Detect various defects on the glass panel on the back of the mobile phone case, including scratches, bumps, discoloration, etc.

[0061] Middle frame inspection station 7. Detect various defects on the middle frame of the mobile phone case, including scratches, bumps, discoloration, etc.

[0062] Arc surface inspection station 9. Detect various defects on the arc surface of mobile phone cases, including scratches, bumps, discoloration, etc.

[0063] The material sorting device 11 automatically screens out defective products according to the previous test summary results.

[0064] The matching computer host is equipped with a visual inspection system, which has multiple visual inspection algorithms embedded in it. It can analyze and process the collected images, filter out defective images, and upload them to the server for statistical summary for easy management.

[0065] The mobile phone shell appearance inspection device is directly connected to the production line to perform online product inspection and automatically sort out defective products. It runs smoothly, reliably and quickly, with an average inspection speed of 5s / 1pcs, which can effectively improve production efficiency, shorten production cycle, have high inspection accuracy, low false detection, and reduce production costs.

[0066] The system of the present invention has low requirements on the operators, does not require complicated software operation and algorithm setting, is simple to operate, and is easy to use.

[0067] The system mechanism design of the present invention adopts a modular design, which is divided into a feeding module / cleaning module / detection module / film unloading module. The modified machine has strong usability and can be used for the subsequent modification of 216 products. The hardware modification cost is low and the debugging efficiency is high. The equipment adds a product return line. If the product fails the first detection, it can be returned to the cleaning station through the return line and retested after the second cleaning. There is no need for manual secondary feeding and testing, and the data report of each test can be output. An internal dust cover is added inside the cleaning equipment to isolate the cleaning environment from the workshop environment as much as possible to ensure that the cleaning environment is not polluted by secondary pollution. The product detection movement is controlled by a servo module, and the double station is used to detect 2pcs of products at a time. The stability and efficiency are high, and the CT can meet 5s / pcs. The computing power of the industrial computer is increased, the image processing time is shortened, and the unit processing time is reduced.

[0068] The light source and detection camera are loosely coupled, and can be adapted to subsequent new models with simple adjustments.

[0069] The system of the present invention uses a highly reliable distributed architecture, with a low probability of software failure and easy maintenance, and a shorter maintenance / upgrade operation time; the functional components are modularized and flexibly called, supporting customized workflows, and can quickly adapt to different inspection scenarios such as quality inspection, outgoing / incoming material inspection, etc., thereby maximizing the use of equipment resources and saving costs.

[0070] Embodiment 2:

[0071] like Figure 4 As shown, the detection method of the mobile phone shell appearance detection system of the present invention comprises the following steps:

[0072] S1: The mobile phone shell is automatically loaded at a frequency of 2pcs per test under the control of the two-way loading station 1;

[0073] S2: The loading robot 2 delivers the mobile phone casing to the cleaning station 3, and uses a dry wipe cloth 31 and a wet wipe cloth 32 to wipe the back panel and the middle frame respectively, and uses an ultrasonic cleaning device 33 to clean fine particles, dust and residual alcohol;

[0074] S3: Use the transfer carrier on the production line to inspect the back panel, middle frame and curved surface in turn at an average inspection speed of 5s / 1pcs. The detected images are transmitted to the visual inspection system. The visual inspection system uses the following algorithm:

[0075] S31: feature extraction, consisting of two 3x3 convolutional layers RELU and a 2x2 maxpooling layer;

[0076] S32: Feature sampling, which is composed of a deconvolution layer, a feature concatenation layer, and two 3x3 convolutional layers ReLU repeatedly;

[0077] S33: Statistics and summary: analyzing and processing the collected images, screening out defective images, uploading them to the server for statistics and summary, which is convenient for management;

[0078] S34: Autonomous learning, using traditional algorithms and AI hybrid detection mode to achieve autonomous learning without human intervention;

[0079] S4: The loading robot 2 takes the mobile phone shell to the unloading and sorting device 11, and the NG flow line 111 returns the NG products for secondary cleaning and retesting; the OK products start to be filmed through the film-sticking mechanism 113 and are placed in the tray for unloading.

[0080] Preferably, the visual inspection system in step S3 includes an image processing module and an image recognition module, wherein:

[0081] The image processing module is used to filter and segment the image data to obtain the image data of the area to be inspected of the product back panel, middle frame and curved surface; and to identify the image data of the area to be inspected to determine the image data of the product target inspection area;

[0082] The image recognition module is used to classify and recognize the image data of the target detection area using a platform that has been integrated with algorithm training, and obtain information on scratches, bruises, and discoloration on the surface so that they can be removed through the material sorting device 11.

[0083] The present invention adopts the traditional algorithm + AI hybrid detection mode, combining the "quantitative" ability of the traditional algorithm with the "qualitative" ability of AI to achieve higher detection accuracy; it adopts a multi-algorithm fusion architecture, which can achieve high-precision discrimination without relying on fixed defect features, and can better distinguish between similar and minor features of different defects; it supports continuous learning, and can realize autonomous learning in the later stage without human intervention.

[0084] With the development of deep learning image detection technology, especially in the automatic feature extraction and end-to-end detection, it has shown good performance, so that deep learning networks can accurately and quickly identify targets from images, and have strong robustness. Therefore, image detection technology based on deep learning has received more and more attention in the field of defect detection.

[0085] The method of the present invention integrates an algorithm training platform to help the existing product detection algorithm to continuously evolve and update; using AI algorithms, it can achieve compatibility detection of subsequent versions of products with the same material and the same defect characteristics; built-in massive data capture nodes help to analyze product and process data. The present invention can be widely used in appearance defect recognition occasions.

Claims

1. A mobile phone shell appearance detection system, characterized in that: The invention comprises a loading station (1) and a cleaning station (3) for a mobile phone shell, and a back panel inspection station (5), a middle frame inspection station (7), a curved surface inspection station (9) and a blanking and sorting device (11) for inspecting the mobile phone shell, wherein: A loading station (1) comprises two automatic loading components and a loading robot (2) arranged on one side of the automatic loading components, wherein the loading robot (2) automatically loads two components simultaneously; The cleaning station (3) includes a wiping device and a dust removal device. After loading, the mobile phone shell conveyed by the assembly line I (4) contacts the surface of the mobile phone shell through the wiping device, and the dust after cleaning is removed by the dust removal device. The back panel inspection station (5) comprises a transfer carrier I and an inspection camera I for inspecting the back panel glass of the mobile phone shell; the mobile phone shell transported by the assembly line II (6) after cleaning is transported to the transfer carrier I, and is moved to the inspection camera I through the transfer carrier I. The inspection camera I is arranged on the back of the transfer carrier I. The inspection camera I transmits the captured image to the visual inspection system, and the visual inspection system identifies the scratches, bumps and discoloration information on the back panel; The middle frame inspection station (7) includes a transfer carrier II and an inspection camera II for inspecting the middle frame of the mobile phone shell. After the back panel inspection, the mobile phone shell is transported by the assembly line III (8) to the transfer carrier II and moved to the inspection camera II through the transfer carrier II. The inspection camera II is arranged around the transfer carrier II. The inspection camera II transmits the captured image to the visual inspection system, and the visual inspection system identifies scratches, bumps, and discoloration information on the middle frame. The curved surface inspection station (9) comprises a transfer carrier III and an inspection camera III for inspecting the curved surface of a mobile phone shell. After the back panel is inspected, the mobile phone shell is transported from the assembly line IV (10) to the transfer carrier III and moved to the inspection camera III through the transfer carrier III. The inspection camera III is arranged around the transfer carrier III. The inspection camera III transmits the captured image to the visual inspection system, and the visual inspection system identifies scratches, bruises, and discoloration information on the curved surface. The material unloading and sorting device (11) comprises two-way automatic unloading components and a material unloading robot (12) arranged on one side of the automatic unloading components, wherein the material unloading robot (12) unloads materials in two ways simultaneously and automatically selects defective products according to the inspection summary results of the visual inspection system; The assembly line is provided with a plurality of transfer carriers at intervals along the conveying direction, and the transfer carriers are used to transfer the mobile phone shells between adjacent inspection stations; each inspection station includes a blocking and lifting mechanism; the blocking and lifting mechanism is used to block the transfer carriers conveyed to each inspection station to the inspection station, and lift the transfer carriers by at least one height of the transfer carriers, and move them to the starting position of the assembly line via the recovery assembly line; The assembly lines correspond to assembly lines I (4) to IV (10) in the cleaning station (3), the back panel inspection station (5), the middle frame inspection station (7), and the curved surface inspection station (9), respectively; the transfer carriers correspond to transfer carriers I to transfer carriers IV in the cleaning station (3), the back panel inspection station (5), the middle frame inspection station (7), and the curved surface inspection station (9), respectively.

2. The mobile phone shell appearance detection system according to claim 1, characterized in that: The feeding robot (2) comprises a lifting mechanism, a turning mechanism and a clamping mechanism; the turning mechanism is arranged on the lifting mechanism; the clamping mechanism is arranged on the turning mechanism, and the turning mechanism can drive the clamping mechanism to turn.

3. The mobile phone shell appearance detection system according to claim 1, characterized in that: The cleaning station (3) comprises a cover body with an FFU pressurized above, and a wiping device and a dust removal device located inside the cover body; The wiping device comprises a dust-free cloth for wiping the back plate and the middle frame, and the dust-free cloth is divided into two types: a dry wiping cloth (31) and a wet wiping cloth (32); The dust removal device comprises an ultrasonic cleaning device (33), an FFU fan and a vacuum device. The ultrasonic cleaning device (33) discharges fine particles, dust and residual alcohol through the bottom of the base plate, and the FFU fan and the vacuum device clean the ultrasonic cleaning device (33) separately.

4. The mobile phone casing appearance detection system according to claim 1, characterized in that: The material unloading and sorting device (11) further comprises a film sticking mechanism (113); the material feeding and transporting module (112) moves the product to the film sticking position, and the film sticking mechanism (113) starts to stick the film; The unloading robot (12) is used to unload the materials, and the NG products are finally detected to flow out from the unloading NG flow line (111), and the OK products are placed in the tray for unloading.

5. A detection method of a mobile phone shell appearance detection system according to any one of claims 1 to 4, characterized in that: The steps include: S1: The mobile phone shell is automatically loaded at a frequency of 2 pieces per test under the control of the two-way loading station (1); S2: The loading robot (2) delivers the mobile phone casing to the cleaning station (3), and uses a dry wipe cloth (31) and a wet wipe cloth (32) to wipe the back panel and the middle frame respectively, and uses an ultrasonic cleaning device (33) to clean fine particles, dust and residual alcohol; S3: Use the transfer carrier on the production line to inspect the back panel, middle frame and curved surface in turn at an average inspection speed of 5s / 1pcs. The detected images are transmitted to the visual inspection system. The visual inspection system uses the following algorithm: S31: feature extraction, consisting of two 3x3 convolutional layers RELU and a 2x2 maxpooling layer; S32: Feature sampling, which is composed of a deconvolution layer, a feature concatenation layer, and two 3x3 convolutional layers ReLU repeatedly; S33: Statistics and summary: Analyze and process the collected images, filter out defective images, and upload them to the server for statistics and summary to facilitate management; S34: Autonomous learning, using traditional algorithms and AI hybrid detection mode to achieve autonomous learning without human intervention; S4: The loading robot (2) conveys the mobile phone shell to the unloading and sorting device (11), and the unloading NG flow line (111) returns the NG products for secondary cleaning and retesting; the OK products start to be filmed through the film-sticking mechanism (113) and are placed in the tray for unloading.

6. The detection method of the mobile phone shell appearance detection system according to claim 5, characterized in that: The visual inspection system in step S3 includes an image processing module and an image recognition module, wherein: The image processing module is used to filter and segment the image data to obtain the image data of the area to be inspected of the product back panel, middle frame and curved surface; and to identify the image data of the area to be inspected to determine the image data of the product target inspection area; The image recognition module is used to classify and recognize the image data of the target detection area using a platform that has been integrated with algorithm training, and obtain information on scratches, bruises, and discoloration on the surface so as to remove them through the material sorting device (11).

Images