33 results about "Underwater vision" patented technology

The invention discloses an image processing method in an underwater vision SLAM (Simultaneous Localization And Mapping) system. The image processing method comprises the following steps: establishing an underwater imaging model; processing influence of underwater environment factors on a camera imaging; carrying out image feature extraction and matching. According to the method, extraction on feature points of an underwater environment image in the later period is reinforced; according to a data correlation method which is newly disclosed on the basis of the improved SLAM system, the feature points are extracted in a matched manner, so that the feature points can be more rapidly and accurately extracted; real-time performance of an SIFT algorithm is improved. The method which uses a relative position factor of a binocular camera and a way point position factor as auxiliary conditions for carrying out correlation can effectively solve the problems of mismatching and poor matching efficiency in the data correlation.

The invention relates to the technical field of nuclear fuel assemblies, and discloses an underwater test platform of a nuclear fuel assembly and a test method. The underwater test platform of the nuclear fuel assembly comprises a simulation water tank and an analysis part, wherein the analysis part comprises a transfer installation module, a detection module, a drive module and an analysis module; the installation module comprises an adjustable installation platform, a reflector substrate and a reflector base; the detection module comprises a to-be-detected object and a video camera; the drive module comprises a first drive part and a second drive part; the analysis module comprises a linear displacement sensor and a data analysis terminal. According to the underwater test platform of the nuclear fuel assembly and test method, through the installation module, the detection module, the drive module and the analysis module in the analysis part, underwater test o the nuclear fuel assembly becomes possible, and a specific design basis is provided for subsequent underwater vision measurement system of the nuclear power station fuel assembly.

The invention discloses a monocular underwater vision reinforcing method based on dark channel prior. The method comprises the steps of establishing a degradation model for an atomization phenomenon and a color casting phenomenon of an underwater image, and acquiring depth-of-field information of the underwater image through calculating parallax between a bright channel and a dark channel; secondly, estimating a water body background color through the depth-of-field information; then acquiring a transmission graph of an underwater environment according to the depth-of-field information, and adjusting transmissivity in the transmission graph through an adaptive manner; and finally restoring the image, performing subsequent processing on the image through color correction, thereby eliminating residual color cast and adjusting the brightness. The monocular underwater vision reinforcing method effectively settles an underwater image reinforcing problem through an improved dark channel prior algorithm. The monocular underwater vision reinforcing method has advantages of simple model, high real-time performance, effective prevention for calculation defects of a complicated model, and better robustness to the environment. The monocular underwater vision reinforcing method can be widely used for the environments such as shallow water, clear water and water area which is rich in plankton and furthermore has wide application prospect and good economic benefit.

The invention discloses an underwater image enhancement method combining a frequency domain and a spatial domain, and belongs to the field of underwater digital image processing. The method comprisesthe following steps: reading an underwater color image, and converting the underwater color image into a grayscale image; adaptively selecting the denoising degree of a frequency domain and the contrast enhancement degree of a spatial domain; converting the grayscale image into a frequency space by using Fourier transform; de-noising the image in the frequency space; inverse Fourier transform is performed on the denoised image, and the denoised image is converted to a spatial domain; segmenting the image of the spatial domain into a plurality of sub-image blocks, and calculating a gray probability density function of each sub-image block; redistributing the probability density function; and calculating the redistributed gray level of each pixel in the sub-image blocks to obtain a finally enhanced image. Compared with the prior art, the invention has the advantages that underwater noise interference can be well removed, image detail features are enriched, consumed time is short, and theinvention is suitable for underwater vision synchronous positioning and image preprocessing before mapping.

The invention relates to an autonomous laying, recycling and charging device for an AUV (Autonomous Underwater Vehicle) under severe sea conditions. The device comprises an elliptic cylinder capturing cage, wherein the tail part of the elliptic cylinder capturing cage is opened, and an inner cavity is formed in the elliptic cylinder capturing cage; the device further comprises a capturing cage moving assembly which comprises a horizontal transverse guide rail erected at the bottom of an unmanned ship control platform and a steel cable winding and unwinding control mechanism which slides on the horizontal transverse guide rail to adjust the transverse position, and a capturing cage fixing assembly which comprises a vertical rod, a stable sleeve and a vertical rod locking clamp, the lower end of the vertical rod is fixedly connected with the elliptic cylinder capturing cage, the stable sleeve sleeves the vertical rod, the vertical rod locking clamp is fixedly connected with a steel cable winding and unwinding control mechanism; the device further comprises an AUV identification assembly which comprises an underwater lighting unit and an underwater visual sensor unit, and an AUV locking assembly which comprises a screw rod arranged in an inner cavity of the tail portion of the elliptic cylinder capturing cage and a locking buckle fixed to the screw rod. Compared with the prior art, the device has the advantages that the recoverable area is enlarged, the requirement for high-precision butt joint is lowered, and the collision risk of the bow portion during AUV recovery is lowered.

The invention provides a target re-identification method based on hypersphere embedding of densely connected convolution network, at first, DenseNet is used to extract the features of underwater deformation object in the video sequence according to the secret-level connected convolution network, so that the gradient disappear is greatly reduced, feature propagation is enhanced, feature reuse and parameter learning processes are supported , from the viewpoint of fine-grained classification, from local integration to global integration, the characteristics of underwater deformation targets are extracted by means of grouping average pooling, the more accurate feature expression ability of underwater deformation target is obtained, in order to avoid directly measuring the Euclidean distance between the coding features of underwater deformed individual targets, a complete and continuous underwater deformed individual target re-recognition model based on multi-point placement is constructedby using hypersphere loss, i.e. Angular triple loss, to focus on the inter-class difference and intra-class difference of underwater deformed individual targets and avoid directly measuring the Euclidean distance between coding features. The invention finally completes the close supervision and process tracking of the underwater deformation target individual in the close-range multi-field observation.

The invention belongs to the field of robot positioning, and particularly relates to a tight coupling initialization method for underwater vision inertial navigation pressure positioning. The method comprises the following steps of obtaining the under-scale robot poses and map feature points through a traditional monocular SLAM method, performing pre-integration on the IMU data between the adjacent images to establish a pre-integration residual error between the images; performing pre-integration on the IMU data between the adjacent images and the pressure measurement to establish a pre-integration residual error between the images and the pressure measurement, solving the initialization parameters of a system through a nonlinear optimization method, updating a map by using the initialization parameters, performing light beam adjustment optimization on the updated system, completing an initialization process, and obtaining a more accurate result. Therefore, the operation of a system is promoted. According to the method, the high-frequency IMU information is used for coupling the image information and the pressure information under different time steps, the coupling degree between the initialization parameters is enhanced, and the solving precision of an initialization algorithm is improved.

The invention discloses an underwater three-dimensional imaging device and method based on multi-view vision. The underwater three-dimensional imaging device specifically comprises an underwater image acquisition module, an image processing module and a display module. The underwater image acquisition module comprises a lead screw, a left high-definition camera, an LED lighting module and a right high-definition camera. The left high-definition camera, the LED lighting module and the right high-definition camera are installed on the horizontal lead screw, and the left high-definition camera and the right high-definition camera are symmetrically installed relative to the LED lighting module. The image processing module obtains an underwater three-dimensional image from images obtained by the left and right high-definition cameras by sequentially utilizing image preprocessing, camera calibration, stereo matching and three-dimensional reconstruction methods. The display module is used for displaying the underwater three-dimensional image. The underwater three-dimensional imaging device based on multi-view vision can effectively solve the problem of an underwater vision blind area of an underwater laser detector.

The invention relates to the technical field of image processing, in particular to an underwater visual identification method and system and a computer readable storage medium. The method specifically comprises the following steps: an image acquisition step; a preprocessing step; a noise reduction step; and a fusion step. According to the underwater visual identification method provided by the invention, an obtained optical image is preprocessed and enhanced to obtain a clearer optical image, an obtained acoustic image is subjected to noise reduction processing to obtain a clearer acoustic image, then processed optical image data and acoustic image data are fused to obtain a final composite image, and the obtained composite image can display underwater image information more accurately.

The invention discloses an underwater vision SLAM (Simultaneous Localization and Mapping) method based on constrained camera motion and distortion correction, which comprises the following steps of: creating an underwater speed optimization equation by utilizing a method of shooting adjacent frame number images and an integral camera motion path in a constrained manner, and carrying out local and integral optimization on the motion rate of an underwater camera; according to the method, the problem of bubbles caused by overlarge deviation of the movement rate of the underwater camera is solved, the quality of a shot image is ensured, and mapping of SLAM is facilitated; when the word bag is created, underwater distortion correction is performed on the shot image and the feature points thereof through a method of fusing radial distortion, tangential distortion and scattering coefficients, so that the problem of mismatching of the feature points during loopback detection can be effectively solved, underwater accumulated errors are effectively eliminated, and the real loopback rate of loopback detection is improved.

The invention belongs to the technical field of underwater robot intelligent control, and particularly relates to an intelligent control method of an underwater robot for autonomous suction and fishing of benthos. The intelligent control method is mainly used for completing detection and recognition of a target organism in a complex underwater environment, guiding the robot to operate and accurately suck a specified target. According to the intelligent control method of the underwater robot for autonomous suction and fishing of the benthos, during operation, the suction robot firstly recognizes and tracks an operation target through an underwater vision and reinforcement learning algorithm, then derives and optimizes fuzzy rules through pose feedback adjustment of the suction robot and anintelligent control system of platform movement of the robot, and guides autonomous suction and fishing operation of the benthos to be completed; and according to the intelligent control method of theunderwater robot for autonomous suction and fishing of the benthos, based on advanced achievements in the aspect of artificial intelligence research, continuous and stable tracking and autonomous suction of the target can be achieved, the intelligent control method has the advantages of being accurate in recognition, high in intelligent degree, high in fishing efficiency, low in operation cost and the like, and the intelligent control method is practically applied to underwater robot system design and has important significance in efficient autonomous suction and fishing of marine organisms.

The invention discloses a underwater visual image enhancement method and system for bionic robotic fish. The method comprises the following steps: acquiring an original target image by using a bionicrobotic fish; preprocessing the original target image by adopting a nonlinear mapping mode; and inputting the original target image into a pre-trained deep convolutional generative adversarial networkDCGAN, and outputting an enhanced initial target image.