Autel Robotics Co., Ltd.

Other

Tencent Technology (shenzhen) Company Limited

Xu Xiang Wang

Yu Song

Xipeng Cui

Lanping Ji

Ning Jia

A keyword extraction method is provided. The keyword extraction method is performed by at least one processor and includes: obtaining to-be-determined words of to-be-processed text; determining preceding words respectively corresponding to the to-be-determined words, where the preceding words are words appearing in the to-be-processed text and preceding the to-be-determined words; determining word sequences of the to-be-determined words according to orders in which the to-be-determined words and the preceding words respectively corresponding to the to-be-determined words appear in the to-be-processed text; inputting the word sequences of the to-be-determined words respectively into a trained cyclic neural network model; obtaining, from the trained cyclic neural network model, a probability that each of the to-be-determined words is a key word of the to-be-processed text; and determining keywords of the to-be-processed text according to the probability that each of the to-be-determined words is a keyword of the to-be-processed text and a preset threshold.

Keyword extraction method, computer equipment and storage medium

The present invention provides a target detection method and an unmanned aerial vehicle. The method includes: obtaining an image of a target, and obtaining a feature model of the target according to the image of the target; determining whether the target is lost; and if it is determined that the target is lost: obtaining a currently photographed first frame of image; obtaining an Nframe of image which is photographed after the first frame of image, N being an integer greater than 0; obtaining, according to the first frame of image and the Nframe of image, a candidate area which possibly includes the target; and determining, according to the feature model of the target and the candidate area, whether the Nframe of image includes the target. A detection efficiency is high by using the target detection method provided in the present invention.

Target detection method and unmanned aerial vehicle

The present invention provides a foldable wing which comprises a wing supporting skeleton, a sliding rail, a skin supporting rib, a skin and a wing movement unit. The wing supporting skeleton comprises a horizontal beam, a longitudinal beam, a wing front edge beam, a wing trailing edge beam, a fixture connector and a sliding block, The wing supporting skeleton is a triangular girder for maintaining planar and sectional shapes of the foldable wing, supporting the skin supporting rib and the skin, and sustaining an aerodynamic load from the skin and a load of a fuselage. After the triangular girder is subjected to a force of the wing movement unit, a shape and an area of the triangular girder are changed so as to achieve folding and unfolding of the foldable wing. A rotocraft and a glider using the foldable wing are also provided.

Foldable wing and rotocraft and glider using the same

Embodiments of the present disclosure disclose a water surface detection method, an unmanned aerial vehicle landing method and an unmanned aerial vehicle. The detection method includes: controlling the unmanned aerial vehicle to generate an airflow, where the airflow is used to generate a water surface ripple; obtaining an image of a landing area of the unmanned aerial vehicle and a flight parameter of the unmanned aerial vehicle; and processing the image, and determining, with reference to the flight parameter, whether the landing area has the water surface ripple. Damage caused by erroneous landing of the unmanned aerial vehicle on a water surface can be alleviated by determining whether the landing area has the water surface.

Growing community of inventors

Guangdong, China

Xu Xiang Wang