US 11,853,483 B2
Image processing method and apparatus for smart pen including pressure switches, and electronic device
Qiwei Lu, Guangdong (CN); Fangyuan Chen, Hubei (CN); Yang Lu, Guangdong (CN); and Pengyu Chen, Guangdong (CN)
Assigned to SHENZHEN EAGLE SOUL INTELLIGENCE & TECHNOLOGY CO., LTD., Shenzhen (CN)
Appl. No. 17/256,211
Filed by Shenzhen Eaglesoul Education Service Co., Ltd, Guangdong (CN)
PCT Filed Aug. 24, 2020, PCT No. PCT/CN2020/110912
§ 371(c)(1), (2) Date Dec. 27, 2020,
PCT Pub. No. WO2022/016649, PCT Pub. Date Jan. 27, 2022.
Claims priority of application No. 202010703394.2 (CN), filed on Jul. 21, 2020.
Prior Publication US 2023/0140470 A1, May 4, 2023
Int. Cl. G06F 3/0354 (2013.01); G06F 3/038 (2013.01); G06F 3/042 (2006.01)
CPC G06F 3/03542 (2013.01) [G06F 3/0383 (2013.01); G06F 3/0421 (2013.01)] 19 Claims
OG exemplary drawing
 
1. An image processing method for a smart pen, the image processing method comprising:
monitoring a working state of a second pressure switch provided at a pen tip of a smart pen after a first pressure switch of the smart pen is in a closed state;
controlling, after it is detected that the second pressure switch generates a trigger signal, an infrared transceiver circuit on the smart pen to send an infrared signal to an area where the smart pen writes, and to collect, in a form of an original image, a reflected signal corresponding to the infrared signal in the writing area at a same time, wherein the reflected signal describes a writing trajectory of the smart pen in the writing area, and the original image comprises two-dimensional coordinates of the writing trajectory on a writing screen;
acquiring a lightweight network model preset in the smart pen, so as to perform feature extraction processing on the original image on a basis of the lightweight network model and a current writing speed of the smart pen to obtain a feature matrix corresponding to the original image, wherein the network model comprises a first convolution calculation channel and a second convolution calculation channel, the first convolution calculation channel and the second convolution calculation channel comprise a first convolution kernel and a second convolution kernel respectively, and a size of the first convolution kernel is smaller than that of the second convolution kernel;
performing classification processing on the feature matrix by using a fully connected layer in the lightweight network model, to obtain a trajectory classification result; and
adding current time information to the trajectory classification result to form a time-ordered trajectory vector, and sending, by means of a Bluetooth module on the smart pen, the trajectory vector to a target object with which the smart pen establishes a communication connection, so as to display a writing trajectory of the smart pen on the target object in real time.