Comprehensive introductionAll countries in the world treat road safety as the first issue. The factors of track and vehicle equipment account for 63% of railway accidents, which shows the severity of railway equipment fault and highlights the necessity of railway fault detection. Key points of innovationThis product is an automatic fault detection system supported by the theory of feature extraction and analysis and pattern recognition. Automatic intelligent detection by machine replaces manual fault identification. Forming the characteristic pattern required by fault detection and building an efficient intelligent system, the fault detection problem of key components of railway freight car body in complex environment can be solved.Technical indexMachine automatic intelligent detection replaces manual fault identification;Building an efficient intelligent system.

Comprehensive introductionAt present, aerospace and other high-end fields have higher requirement for processing level, and fine quality defects on metal shell surface are not allowed, requiring factory products for 100% qualified rate. All of these challenge the stability of the whole process of production. How to monitor products with high efficiency and high precision and realize the detection of metal shell surface defects has become an important means of production management and quality assurance in modern enterprises. This product takes small metal shell defects as the research object, and develops a set of embedded metal shell defects visual detection system. Key points of innovationCompared with the defect detection system of PC, the system has smaller volume and power consumption, and better portability and adaptability. It can be used independently or as a functional module in small industrial system, in order to realize the efficient and real-time detection of metal shell defects.Technical indexThe system has a smaller volume and power consumption, and better portability and adaptability.

1. Technological advancementDeep learning has achieved unprecedented performance in the field of Artificial Intelligence. However, due to the complexity of deep model, its deployment in practical applications with limited computing and storage resources such as mobile devices, edge devices and real-time systems becomes a difficult and painful point. In this technology, a model compression technology based on knowledge distillation is proposed for face recognition obscured by masks. The lightweight student model obtained by learning can be applied and deployed to the edge computing platform with limited resources. 2. Technology maturityThis technology is based on knowledge distillation of depth model compression, and presents a new face recognition technology for mask occlusion facing edge intelligence. Knowledge distillation is a hot topic in model compression research in recent years. It is a simple and effective model compression technology, which has been developed and used by google, Baidu and Tencent.3. Situation of intellectual property rightsThis technology is supported by Jiangsu University and has been supported by the National Natural Science Foundation.4. Expected benefits and types of enterprises to cooperate withThe technique is an edge intelligent algorithm, which can be used for face and image recognition on edge computing platform.

1. Advanced technologyDimension reduction technology is an effective way of learning high-dimensional data representation, which is widely used in feature representation of various big data analysis to explore the essential and useful structural feature information contained in data. Aiming at high-dimensional visual data, this technique proposes a new feature extraction method considering the geometric structure information of data sample points, which can be effectively applied to a variety of visual recognition scenes. 2. Technology maturityBased on the theoretical basis of graph embedding technology, this technology proposes a new feature extraction method for high dimensional visual data. Feature extraction is a classic technique used in all kinds of big data analysis and has been widely used in the practical application of high dimensional data analysis. 3. Situation of intellectual property rightsThis technology is supported by Jiangsu University and has been supported by the National Natural Science Foundation. 4. Expected benefits and types of enterprises to cooperate withThis technique is a new feature extraction algorithm, which can be used in the analysis and recognition tasks of various high dimensional data.

Comprehensive introductionStrawberry sorting is a key link in its post-production processing. At present, our country adopts artificial sorting of strawberry for the most time. Manual sorting method is mainly based on naked eye observation to judge the grade, which lacks objectivity. The size of strawberry is different, and the surface condition is complex, so simply relying on the naked eye can not reach a good sorting effect. Because the surface of strawberry fruit is very soft and fragile, manual sorting is easy to cause mechanical damage, affecting the commodity value of strawberry. Therefore, it is of great scientific significance and social economic value to study fast, accurate and efficient strawberry automatic sorting technology and equipment.  Innovation points and Technical indexBased on machine vision and robot technology, this product can do strawberry color, shape, size identification, evaluating different grades by using the image processing algorithm. It can also use the mechanical arm to achieve automatic sorting of strawberry, reducing the damage in the sorting process, improving the efficiency of strawberry sorting and sorting quality, which has important practical significance and practical value.

CCTV program I Love Invention about unmanned combine harvester with sound and subtitles in Chinese. (P.S. The program flash could not be viewed through Chrome. You may try other Internet explorers)

In order to improve the mechanized spraying effect of stunted and close planting cotton, an air-assisted cotton boom sprayer with separating row and spraying in inside and upper canopy by combining separating row and spraying inside canopy technology with air-assisted spraying technology was designed. This sprayer braced the droplet diffusion space of inside cotton canopy by row separator and improved droplet diffusion range and distribution uniformity by airflow disturbance and transport of inside and upper cotton canopy.