Vehicle inspection speech synt€♥ hesis and speech recognition system

The application of speech synthesis and speecδ <✔h recognition to vehicle inspection and ot‍↕her inspection and maintenanc☆↕"¶e industries can overc₩♦♥Ωome many shortcomings of the   $traditional method based on maintenance÷¶ list, and realize the real ‍♦✘paperless maintenance industry.

Traditional maintenan♥÷ce based on work order, there are som✔÷e inaccurate problems suc<γ ₩h as missing inspection; There is ↔♦ dirt and other conditions in the ma♣σintenance site, work order damage an♠σd stains lead to incomplete informat​↓↕•ion; After the need for π±special input, in order to prepare σ≈©for the future maintenance management and inqui‍☆ry. Specially-assigned↕λ input also increased the cost&​δ.

The maintenance system of v$↑ ehicle inspection based on sp♥‍€eech synthesis and speech recognit₩↔§βion realizes the real paperl™©ess, greatly reduces the work cost, realizes th‍ Ωe full informatization of maintenance, impr•☆>↕oves the management efficiency of maintenance★∞&, and improves the corpora←¶←&te image.

The speech synthesis and speech recognition tech®↓<nology of our company in this project€φ§ has achieved very good natural fluency↕™ and high recognition ×≠€"accuracy, which fully mee→≥≠ts the needs of practic≤≤al application.

LOGO detection, positioning and qualification de♣∑$termination system

The system is mainly divided into two moduσ≥les: detection of positioninφεπλg module and qualified determination module. ™♣¶↕Positioning module for detecγ$δ∑tion, the traditional ima≤®σ‍ge processing method is c¶™ombined with machine learning method, the ro>€bustness of using traditional image procε∞λΩessing method, obtain reliable identifica  ₹∞tion results, by machine learni <ng recognition of landing, the existing imag®•☆✘e distortion, perspe✔∞ctive transformation, target cover co↑γ"mplex situations such as object still has stronφ&≠•g ability to recognize. F★φ≠or qualified determination module, on thγ​e basis of the corporate Logo use normative do☆§$σcuments, formulate a s₩εε←eries of decision rules based on compute​'r vision technology, •×↑such as size, clarity, color, backgrδφound color Logo Logo itself, such as t €×o upload a file of the Log±→↔↕o one by one to judge, output to determin₽ σe reasons not complia α∑nce Logo. After a lot of testing ve↓εrification, achieve the desired ©‍accuracy.Use the system to achieve the↑•‌ file screening of all information, grea✔∞♣tly reduce the artificial cost expen¶"diture, promote the enterprise intern¥∏©al process standardization, improved efficiency ←™>and greatly improve t©•he image of the enterprise.

Large enterprises will manually judg>↓e the compliance of Logo€>∑  use in company documents one by one, w♣≈≠↔hich not only consumes huge labor costs÷β', but also has the unreliabili'§↓<ty of subjective judgment.™<∏ The intelligent system ♦☆developed by our company -- Logo &" detection, positioning and qualific₩£>↔ation determination sys₹₹Ωtem can effectively solve the above ≈™ problems.

The system is divided into two mod♥®×ules: detection and positioning m§γodule and qualification module. Posit∞§​ioning module for detection, th÷∞↓e traditional image processing ®♥★∑method is combined with ma•♣↕σchine learning method, the robustness of us∞✘≤'ing traditional image p¥ rocessing method, obtain r>←eliable identification®÷ results, by machine lea♠≠rning recognition of landing, the existing₩≤ image distortion, p α♥erspective transformationσ↓, target cover complex situations such as ₹✔€object still has strong abilit‌ ≈y to recognize. For the conformity determin$÷&→ation module, according to the enterpri↕✔£♦se Logo use specification do&&∏cument, a series of decision rules based on​>‍ computer vision technolog‍→γ∏y, such as size, clarity, Logo color, Logo b≠•≥ackground color, etc., judge the Logo in t∏δ'he uploaded file one by ​±one, output the decision reaso‍σ$✔ns of non-compliance Logo. Through a large numδ→ber of tests, the expected recognition÷≤γ£ accuracy is achieved.

Using this system to achieve th™✔<₩e full information of doc∞★±ument screening, greatly ↑≥♦reduce labor costs, promote the standa ↔÷$rdization of the internal process of enterp∞★Ω≤rises, improve work efficiency, great'φly enhance the image of enterφ §∞prises.

Automatic recognition system of human¥δ posture by machine vision

System for static painting, video​♦, images, such as formatΩ←≠↕ion of the target and the camera, the dat&₽a extraction, processing and ≥↕treatment, the whole process of the visual imaφ∏ge, such as data file. The system can be a veλ€Ωry good recognition s↕∏®quatting, walking, standing posture,​≠↓✔ such as accuracy above 96%.The±≠£ technology background is widely used, maδ inly in the intelligent video surveillance >, factory work condition monitor≠™✔ing, patient monitoring sy π®stems, human-computer interaction, virtual π®→≥reality, smart home, inφ≠telligent security, auxiliary training ​φathletes, content-based video ¶×retrieval and other i®σntelligent image compression tec•↑"hnology has broad application prospects

With the rise of "human-centere&↕d computing" and the emergence of ne↑≈∞w applications in life, human postu ©←re recognition and behavior und₹© ↑erstanding have gradually become the rese" ¶arch focus in the field of computer visi♣∑¶☆on.

The system can extract and procesγ∏≠s the characteristic data of the target personaδ Ω↔ge, quickly and accurately recognizeπ<≤× the pose of the target personage throu'±gh artificial intelligence deep δ π≥learning, and provide the Web API of related↕♠₩ recognition process externally for othεδΩer users to apply.

The system can extract, pr£×±ocess and discriminate data from objects formed≈‌¥↓ by still painting, video streamφ✘Ω and camera shooting images, and vis¥♦ualize images, data and other×✔ ≤ files in the whole process. At present, the sy✘¥stem can well ident↔↑ify squatting, walking, standing aβ♥♦©nd other postures, the accuracy rate is ≤• "above 96%.

The application backgrπ<γound of this technology is very ±∞♠≈wide, mainly focusing on♥α←♠ intelligent video monitoringπ✘£, factory work monit∑>oring, patient monitoring ↔¶ system, human-computer inte¶£φraction, virtual reality, smart home, intel€‌ligent security, athlete assisted training,  ♠♠ in addition, content-based video retrieval®" and intelligent imag✔φ​e compression and other tech≠↔×αnologies have broad applε₹↔ication prospects.

Information detection§☆≤ and desensitization system

Global data is chara♥↕₽cterized by explosive σ♠¶•growth and massive interaction, and ₩π★ electronic documents$₽ are widely used in enterprises and individualγ ±s because of the large amount of i₹↕nformation they can carry. More and §←♠✔more enterprises make use of←$ massive data for business anal↑δ♦×ysis and decision making. ‍>§≠However, if electr©δ₩onic documents containing sensitive inform Ωσ∞ation are leaked or made >♠ public without information filtering,×↑÷ serious consequences may result. The intel↔ φδligent system developed by our company -λ♥- information detection and b"♠≤lackening desensitization system ÷∏can effectively ensure the securit÷∑ y of electronic documents.

The system is mainly divided i₽‍®δnto two modules: detection and positi& αoning module and blackening mo∞&>dule. For the detectiφ on and positioning module, it is div​≤₽ided into text and seal handwritte✔§n signature detection. ¥÷γFor text detection, the traditio€☆↔ nal rule-based method is combined with dee"αp learning method to solve the problem of ↓★regular sensitive information detecti$₹βon of digital classes through regular matching,&≥✘✔ and a deep learning model is construcβ₹✔★ted to realize the detectionλ""β and location of named →≠←entity classes using contextα€φ information. For sealαπ and handwritten signature detection, deep learn÷φ≈ing detection is implemented t""≥o ensure the distinc•π÷tive seal detection effΩ≈β ect, while improving the handwritten signature ≤‌detection effect in the presence of occlu ≥ε∞sion or close arrangement and o'♠β≈ther complex situations. For the blackened modul↑‌"e, the detected category inf÷♣ormation and location i  ♦‌nformation is blackened, and the blackened ←♣≥♣ file picture is previewed for user× ≤‍s to view.

The system realizes the whole informa®Ωδ↑tion detection. In the background of privacyΩ₽ protection in the era o&>¥f big data, the system can make the di♣€sclosure and exchange of information more↓↕ convenient, and will be more wide•∞₽λly used in the practica≥"l application of privac♦♦★≠y protection and information exchange&☆.