An Improved Neural Network Algorithm for Intelligent Object Recognition

With the maturity and perfection of Face Recognition, Intelligent Object Recognition has gradually become a new research hotspot in the field of Artificial Intelligence. Intelligent Object Recognition means that computers can distinguish objects by pictures or images. However, most of the existing intelligent Object Recognition algorithms are based on massive databases, which is not friendly to new researchers. And the existing algorithms have higher resolution when the components of objects are relatively static (such as plants), but the resolution of objects in pictures is greatly reduced when the object is deformed or incomplete. Aiming at these problems, we propose an improved Neural Network Algorithm: allowing objects in pictures to be incomplete; establishing a relation-table of similar objects in the database; analyzing the scene of the picture. Similar to Simulated Annealing Algorithm, when recognizing objects, the improved Neural Network Algorithm has a probability of selecting similar objects by relation-table. The result of the experiment shows that the improved Neural Network Algorithm will still have a good performance in a common data set, which is more friendly to new researchers and is worthy of further study.