Research On Fish Length Measurement Technology Based On Computer Vision

With the decline of fishery resources,fishery has changed from fishing to aquaculture.At present,fish culture is developing towards large-scale,intensive and intelligent.Therefore,the real-time monitoring and condition analysis of cultured fish has also become the core issue of realizing the intellectualization of supporting equipment for fish culture.As a non-invasive and stable observation technology,computer vision technology is being accepted by many aquaculture behavioral researchers.In intensive fish culture,the reliable estimation of fish body length is of great significance to aquaculture.At present,the measurement of fish body length is mainly carried out by aquaculture personnel through visual inspection or fishing out the fish body.However,these methods are either not accurate enough,or they may cause damage to the fish and be inefficient.Therefore,the research on non-contact automatic body length measurement of cultured fish is urgently needed to provide new ideas for intelligent aquaculture.This paper will take the cultured fish as the main research object,extract the fish body information under the actual breeding conditions by using the method of computer vision,evaluate the feasibility of obtaining the overlapping fish body length measurement line,solve the problems of poor acquisition of fish body information caused by occlusion and overlap,and develop a non-contact fish body length information acquisition method based on binocular camera.The stability and accuracy of the proposed estimation method are verified.The non-contact fish body length automatic measurement method proposed in this paper mainly includes two modules: fish body overlapping segmentation module and fish body length automatic measurement module.The main function of the fish body overlapping segmentation module is to filter out the redundant information,retain the key information of the fish body and analyze it: first,find the position of the fish body on the graph,segment the foreground of the fish body,and simplify the fish body image into a binary image;Then,the fish body contour is extracted and analyzed,and the useful information such as the overlapping points on the fish body contour and the vertices of the head and tail of the fish are extracted;Then,the head and tail information belonging to the same fish body in the overlapping contour of the fish body are found through contour segment classification and contour segment matching;Finally,these information are extracted and the complete contour of the fish body is reconstructed to complete the task of segmenting an independent fish body contour from the overlapping fish body.The automatic fish body length measurement module first extracts the fish body length measurement line from the contour of the independent fish body as the basis for measuring the real fish body length;Then,through the binocular vision system,the camera parameters calculation,distortion correction,epipolar registration and other prerequisite tasks are completed;Then,through the image matching of the body length measurement line of the same fish extracted from the left and right images of the binocular camera,the three-dimensional coordinates of the points on the body length measurement line in the real world are calculated;Finally,using the three-dimensional information of fish body length measurement line,the real fish body length data is obtained,and the task of automatic fish body length measurement is completed.