Study On The Underwater Target Identification System Based On VxWorks

As a key technology to the field of the Underwater Robots, in today's image-processing field, Automatic Detection and Recognition of the underwater targets has become one research topic with meaningful and challenging. Underwater Robot Automatic Target Recognition system plays an important role for navigation, obstacle avoidance, detection of an unknown target. In order to ensure that the system running real-time and reliability, use the current most widely used embedded real-time operating system VxWorks.According to the system's real-time and reliability requirements, the hardware platform uses ShengBo's production PC 104 combined with Vx Works embedded real-time development systems,practice has proves that real-time and reliability of the system meet the design requirements.Major research of this paper is Underwater Automatic Recognition and Classification, mainly studied the three areas of the underwater target image pre-processing, underwater target feature extraction and classification of underwater target recognition. Whit analysis of the traditional image processing algorithms and recognition algorithms, comparited the advantages and disadvantages between the different approaches and adaption to the conditions, combined with the characteristics of an underwater target image, proposed a set of automatic identification of underwater objectives. This method can extract a complete representation of the characteristics of the type of underwater target, and ultimately achieve the purpose of accurate extraction of target. From the experimental results, the existence of shadow images can be accurately identified and classified, and the calculated amount is small.In conclusion, this paper did in-depth study to the underwater target recognition technology, the present automatic identification program based on geometric features and invariant moment characteristics of BP neural network, using the real pictures of underwater images for the experimental verification. Results show that the program presented in this paper obtained a good recognition effect.