• Image Analysis with Machine Learning Algorithms 
  

Using the state-of-the-art deep learning technology to solve the problems of nondestructive detection based on hyperspectral image. The main idea of these processes could be exampled with a paper recently written as follow. A distribution model of NPK content based on convolution neural networks (CNNs) and deep neural networks (DNNs) is proposed. Firstly, the spectral noise was reduced by Savitzky-Golay filter. The initial features extraction was carried out by using a principal component analysis (PCA) to identity a number of potential characteristic wavelengths according to the weigh coefficient distribution curve of the first three principal component images under the full wavelengths. For the characteristic spectral images and the principal component images, the texture based on the gray level co-occurrence matrix were extracted from those images, and the structure information of those images was also extracted based on CNN simultaneously. The hyperspectral wavelength feature data, texture data, images structure data and the combined data were utilized to develop particle swarm optimization-support vector regression (PSO-SVR) and independent component analysis-deep neural networks (ICA-DNNs) respectively. Some main conclusions were obtained: the performance of PSO-SVR model based on characteristic spectrum is best, which the best coefficient of determination (R²) of calibration set and validation set of the entire growth state were up to 0.8983 and 0.8891, respectively. Multi-source data fusion performance of ICA-DNNs is best, and the precision of ICA-DNNs model is improved based on feature spectrum, texture feature and images structure feature, the R2 of calibration set and validation set were 0.8998 and 0.8825 respectively. The results show that the PSO-SVR model based on the feature spectrum is the best, and the determination coefficient of the calibration set and validation set model is 0.8983 and 0.8891. The model of ICA-DNNs was rained based on stochastic gradient descent (SGD) algorithm. The precision of ICA-DNNs model is improved based on characteristic spectrum, texture feature and CNN structure feature. 

What's more, He is also interested in deep learning techniques for image analysis such as image recognition, classification, and super-resolution.

• Applications of Deep Learning in Natural Language Processing 
  

His research interests focus on the applications of deep learning models in natural language processing (NLP). Specifically, using the recurrent neural networks (RNNs), convolution neural networks (CNNs), long short term memory (LSTM) models to solve the problems of text classification, sentiment analysis , and machine translation etc.

• Artificial Intelligence in Wireless Sensor Networks

To address various challenges such as data aggregation and fusion, energy aware routing, optimal deployment and localization etc. in wireless sensor networks (WSNs) using computational intelligence (CI) algorithms, especially the application of evolutionary optimization strategies in neural networks (NNs).

1. Unmanned Aerial Vehicle Detection System and Method Based on Laser and Binocular Vision (Publication No. CN105138002A, Ranking: 2/12)

2. A kind of Device and Method of Surface-to-Air Sensor Network Communication compatible with Unmanned Aerial Vehicle（Publication No. CN105828345A, Ranking:5/11)

3. Scientific and Technological Innovation Project of Water Conservancy of Guangdong Province , China. Demonstration and Research on the Key Technology of Water Saving Irrigation Based on UAV in Guangdong Province (Grant No. 2016-18), 2016-2019. Ranking: 18/20.