| As the ocean strategy is proposed and smart ocean becomes the new focus,the use of underwater robots instead of real people for underwater operations is an inevitable trend.During the participation in the development of the underwater master-slave high dynamic range 3D panoramic sensing system and detection equipment project,it was found that the problems of expensive underwater equipment,difficult debugging and environmental uncertainty led to the use of underwater robots in real waters for algorithm verification experiments and marine equipment education and training to become a problem to be solved.In response to the problems of expensive equipment,difficult and dangerous experiments,this paper designs a virtual simulation education system for algorithm verification and courses related to artificial intelligence and underwater machines for university students.When using the virtual simulation education system,it is found that the underwater images have low contrast,blurring and low accuracy of underwater object detection.The reason for this is that the underwater environment of the virtual simulation education system has problems with attenuation of light waves and scattering of light.These problems make the underwater object detection accuracy of the virtual simulation education system much less accurate.It can be seen that the low accuracy of underwater object detection is a problem to be solved.To address the above problems,this paper proposes an underwater image enhancement algorithm based on contrast stretching and exposure fusion of limiting coefficients,which calculates the relationship between limiting coefficients and channels to perform different contrast stretching for each channel of the image,and then uses exposure fusion to further improve the image brightness to obtain a clearer underwater image.In this paper,the proposed underwater image enhancement algorithm is experimentally compared with other methods and the algorithm is evaluated using visual and data metrics,thus proving that the proposed algorithm in this paper can effectively improve the contrast,brightness and clarity of underwater images.Secondly,to address the problem of wrong and missed detection caused by blurred underwater images,this paper proposes a dual-stream input-based SSD underwater object detection algorithm based on the SSD object detection network.The algorithm improves the detection accuracy by means of dual-stream input,intentional force mechanism and multi-scale fusion.The experimental comparison and analysis prove that the SSD underwater object detection algorithm based on dual-stream input proposed in this paper has higher accuracy.In addition,the virtual simulation teaching system is briefly introduced in this paper.Based on the course requirements of "Robotics Fundamentals" and "Programming and Problem Solving",the virtual simulation education system has been developed with relevant teaching practice functions,and relevant course experiments have been designed.The virtual simulation education system in this paper has been well received by teachers and students in practical teaching. |
