Vehicle inspection speech synthesis and speech recognition system

The application of speech synthesis and speech recognition to vehicle inspection and other inspection and maintenance industries can overcome many shortcomings of the traditional method based on maintenance list, and realize the real paperless maintenance industry.

Traditional maintenance based on work order, there are some inaccurate problems such as missing inspection; There is dirt and other conditions in the maintenance site, work order damage and stains lead to incomplete information; After the need for special input, in order to prepare for the future maintenance management and inquiry. Specially-assigned input also increased the cost.

The maintenance system of vehicle inspection based on speech synthesis and speech recognition realizes the real paperless, greatly reduces the work cost, realizes the full informatization of maintenance, improves the management efficiency of maintenance, and improves the corporate image.

The speech synthesis and speech recognition technology of our company in this project has achieved very good natural fluency and high recognition accuracy, which fully meets the needs of practical application.

LOGO detection, positioning and qualification determination system

The system is mainly divided into two modules: detection of positioning module and qualified determination module. Positioning module for detection, the traditional image processing method is combined with machine learning method, the robustness of using traditional image processing method, obtain reliable identification results, by machine learning recognition of landing, the existing image distortion, perspective transformation, target cover complex situations such as object still has strong ability to recognize. For qualified determination module, on the basis of the corporate Logo use normative documents, formulate a series of decision rules based on computer vision technology, such as size, clarity, color, background color Logo Logo itself, such as to upload a file of the Logo one by one to judge, output to determine reasons not compliance Logo. After a lot of testing verification, achieve the desired accuracy.Use the system to achieve the file screening of all information, greatly reduce the artificial cost expenditure, promote the enterprise internal process standardization, improved efficiency and greatly improve the image of the enterprise.

Large enterprises will manually judge the compliance of Logo use in company documents one by one, which not only consumes huge labor costs, but also has the unreliability of subjective judgment. The intelligent system developed by our company -- Logo detection, positioning and qualification determination system can effectively solve the above problems.

The system is divided into two modules: detection and positioning module and qualification module. Positioning module for detection, the traditional image processing method is combined with machine learning method, the robustness of using traditional image processing method, obtain reliable identification results, by machine learning recognition of landing, the existing image distortion, perspective transformation, target cover complex situations such as object still has strong ability to recognize. For the conformity determination module, according to the enterprise Logo use specification document, a series of decision rules based on computer vision technology, such as size, clarity, Logo color, Logo background color, etc., judge the Logo in the uploaded file one by one, output the decision reasons of non-compliance Logo. Through a large number of tests, the expected recognition accuracy is achieved.

Using this system to achieve the full information of document screening, greatly reduce labor costs, promote the standardization of the internal process of enterprises, improve work efficiency, greatly enhance the image of enterprises.

Automatic recognition system of human posture by machine vision

System for static painting, video, images, such as formation of the target and the camera, the data extraction, processing and treatment, the whole process of the visual image, such as data file. The system can be a very good recognition squatting, walking, standing posture, such as accuracy above 96%.The technology background is widely used, mainly in the intelligent video surveillance, factory work condition monitoring, patient monitoring systems, human-computer interaction, virtual reality, smart home, intelligent security, auxiliary training athletes, content-based video retrieval and other intelligent image compression technology has broad application prospects

With the rise of "human-centered computing" and the emergence of new applications in life, human posture recognition and behavior understanding have gradually become the research focus in the field of computer vision.

The system can extract and process the characteristic data of the target personage, quickly and accurately recognize the pose of the target personage through artificial intelligence deep learning, and provide the Web API of related recognition process externally for other users to apply.

The system can extract, process and discriminate data from objects formed by still painting, video stream and camera shooting images, and visualize images, data and other files in the whole process. At present, the system can well identify squatting, walking, standing and other postures, the accuracy rate is above 96%.

The application background of this technology is very wide, mainly focusing on intelligent video monitoring, factory work monitoring, patient monitoring system, human-computer interaction, virtual reality, smart home, intelligent security, athlete assisted training, in addition, content-based video retrieval and intelligent image compression and other technologies have broad application prospects.

Information detection and desensitization system

Global data is characterized by explosive growth and massive interaction, and electronic documents are widely used in enterprises and individuals because of the large amount of information they can carry. More and more enterprises make use of massive data for business analysis and decision making. However, if electronic documents containing sensitive information are leaked or made public without information filtering, serious consequences may result. The intelligent system developed by our company -- information detection and blackening desensitization system can effectively ensure the security of electronic documents.

The system is mainly divided into two modules: detection and positioning module and blackening module. For the detection and positioning module, it is divided into text and seal handwritten signature detection. For text detection, the traditional rule-based method is combined with deep learning method to solve the problem of regular sensitive information detection of digital classes through regular matching, and a deep learning model is constructed to realize the detection and location of named entity classes using context information. For seal and handwritten signature detection, deep learning detection is implemented to ensure the distinctive seal detection effect, while improving the handwritten signature detection effect in the presence of occlusion or close arrangement and other complex situations. For the blackened module, the detected category information and location information is blackened, and the blackened file picture is previewed for users to view.

The system realizes the whole information detection. In the background of privacy protection in the era of big data, the system can make the disclosure and exchange of information more convenient, and will be more widely used in the practical application of privacy protection and information exchange.