CN-121986743-A - Cage fish grouping and directional feeding method and system based on stereoscopic vision

CN121986743ACN 121986743 ACN121986743 ACN 121986743ACN-121986743-A

Abstract

The invention discloses a cage fish grouping and directional feeding method and system based on stereoscopic vision. The method comprises the steps of collecting video data through a multi-mesh underwater camera system, generating three-dimensional point cloud data of fish bodies based on a binocular vision principle, extracting growth and behavior characteristics, dividing target groups by using a clustering algorithm, generating a differential feeding strategy by combining environmental parameters, driving a multi-outlet directional feeding machine, and selecting a direct-falling mode or an air-feeding mode according to water layer distribution identification to feed to a target area. The invention realizes accurate sensing and layered feeding of the fish shoal specification, solves the problem of uneven ingestion caused by specification differentiation, improves the utilization rate of feed, reduces pollution of residual bait and promotes the orderly growth of fish.

Inventors

PAN YI
CHEN CHANG
LUO YUXUAN
Zhong Yirou
YANG YUTONG
LIU JINGYI
Wu Chengtong
TAN SHULING
Cai Zetao
Zhong Zhirui

Assignees

广东海洋大学

Dates

Publication Date: 20260508
Application Date: 20260309

Claims (10)

1. The method for classifying and directionally feeding the net cage fishes based on the stereoscopic vision is characterized by comprising the following steps of: Acquiring fish shoal video data through a multi-eye underwater camera system, and processing the video data based on a binocular vision principle to generate fish three-dimensional point cloud data; Extracting body length and weight data of each fish individual from the fish three-dimensional point cloud data as growth characteristics, and dividing the fish group into at least two target groups with different growth characteristics by using a clustering algorithm in combination with behavior characteristic data; Generating a differential feeding strategy comprising feed types, feeding amounts and feeding areas according to the growth characteristics of each target group and current environmental parameters; And controlling the multi-outlet directional bait casting machine to accurately send the feed to the activity area where each target group is located according to the differential feeding strategy.
2. The stereoscopic vision-based cage fish grouping and directional feeding method of claim 1, wherein the combined behavioral characteristic data specifically comprises: And identifying the biological characteristics of the fish body by using a deep learning model, and continuously tracking each fish body in continuous video frames by using an interframe matching algorithm to obtain the swimming speed and the ingestion frequency as the behavior characteristic data.
3. The stereoscopic vision-based cage fish grouping and directional feeding method according to claim 1, wherein the classifying the fish group into at least two target groups having different growth characteristics by using a clustering algorithm comprises any one of the following means: calculating an average body length value, and dividing fish individuals with actual body length values larger than, in or smaller than a preset proportion range of the average body length value into fast, normal or slow growth groups respectively; And secondly, constructing a multidimensional feature space by taking the body length as a core feature and combining the water temperature, dissolved oxygen, salinity, PH value and flow velocity, and dynamically dividing the fish shoal into preset quantity groups or automatically determining quantity groups based on a clustering effectiveness index by utilizing a K-means clustering algorithm.
4. The stereoscopic-based cage fish grouping and directional feeding method of claim 1, wherein the generating a differentiated feeding strategy specifically comprises: invoking a nutrition demand model obtained based on fingerling basic data, grouping characteristics and seasonal rhythm training, and calculating target feeding amounts of target groups; quantitatively analyzing the matching relation between the environmental parameters and the nutritional ingredients of the feed by using an adaptation degree analysis model, outputting feed adaptation scores and matching the feed with corresponding nutrition ratios; And after the logic consistency check is carried out on the feeding strategy data, generating a standardized feeding instruction containing the water layer distribution identification.
5. The stereoscopic vision-based cage fish grouping and directional feeding method of claim 4, wherein the controlling the multi-outlet directional feeding machine accurately feeds the feed to the activity area where each target group is located according to the differential feeding strategy, specifically comprising: analyzing a water layer distribution identifier in the standardized feeding instruction; if the identification indicates that the target group is distributed at the bottom layer of the net cage, controlling the bait casting machine to execute a direct-falling type feeding mode; And if the identification indicates that the target group is distributed at the upper layer in the net cage, controlling the bait casting machine to execute the air-feeding type feeding mode.
6. The stereoscopic vision-based cage fish grouping and directional feeding method of claim 5, wherein the straight drop type feeding mode comprises adjusting a flow regulating valve of a vertical blanking channel to control a blanking rate; The air feeding mode comprises the step of adjusting the air pressure of the high-pressure fan and an air flow adjusting valve to control the air flow intensity.
7. The stereoscopic vision-based cage fish grouping and directional feeding method of claim 1, further comprising a feedback adjustment step of: periodically collecting growth data and the residual amount of feed after feeding, and comparing actual and expected growth rates; And if the deviation is lower than a preset threshold value, adjusting feeding parameters and updating the nutrition demand model.
8. Cage fish grouping and directional feeding system based on stereoscopic vision, which is characterized by comprising: The stereoscopic vision recognition module is configured to collect video data through the multi-mesh underwater camera system, generate three-dimensional point cloud data of the fish body based on a binocular vision principle and extract growth characteristics and behavior characteristic data from the three-dimensional point cloud data; The intelligent decision module is configured to divide a target group by using a clustering algorithm based on the characteristic data and generate a differential feeding strategy according to the growth characteristics and the environmental parameters; The group feeding execution module comprises a multi-outlet directional bait casting machine and is configured to send the feed to a target area according to the differentiated feeding strategy; And the central processing unit is used for scheduling data interaction and instruction execution of each module.
9. The stereoscopic vision-based cage fish grouping and directional feeding system of claim 8, wherein the intelligent decision module comprises a nutritional requirement analysis unit, a feed adaptation unit, and a command generation unit configured to calculate a target feeding amount, match a feed nutrition ratio, and generate a standardized feeding command including a water layer distribution identifier, respectively.
10. The stereoscopic vision-based cage fish grouping and directional feeding system of claim 9, wherein the grouping feeding execution module further comprises a grouping feeding controller configured to parse the water layer distribution identification, generate a first control signal to drive the bait casting machine to execute a straight-drop feeding mode when the identification indicates a bottom layer, and generate a second control signal to drive the bait casting machine to execute an air-feed feeding mode when the identification indicates an upper layer.

Description

Cage fish grouping and directional feeding method and system based on stereoscopic vision Technical Field The invention belongs to the technical field of intelligent fishery and aquaculture automation, and relates to a cage fish grouping and directional feeding method and system based on stereoscopic vision. Background With the large-scale development of the aquaculture industry, deep water cage culture has become an important mode for improving the yield of aquatic products. In the cage culture process, feeding management is a key link for determining culture cost, fish growth speed and water body environment safety. The traditional feeding mode is mostly dependent on manual experience or simple mechanical feeding with fixed time and fixed quantity, and a strategy of uniformly throwing the whole pool is often adopted. However, due to natural genetic differences, differences in feeding ability and differences in environmental suitability among fish individuals, long-term polyculture can lead to significant specification differentiation of fish populations. The conventional feeding technology has the main defects that firstly, accurate sensing means for real-time distribution state and individual specification of the shoal are lacked. The traditional method can not identify the vertical layering phenomenon of the fish shoals in the water body (such as that small-size fries are gathered on the middle upper layer and large-size adult fishes are perched on the bottom layer), so that the feeding strategy is cut at a knife. If the feeding amount is too small, the feeding capacity of the large-specification dominant population is strong, the growth of the small-specification dominant population is blocked due to the competition disadvantage, the population specification difference is further increased, the overall fish quality is reduced, and if the feeding amount is too large to meet the small-specification population demand, a large amount of feed is wasted, the residual feed is sunk and rotten, the cultivation cost is increased, the water quality is deteriorated, and the fish disease is induced. Secondly, the existing automatic bait casting equipment is single in function, mostly has only a simple flow control or horizontal rotation function, and lacks an intelligent switching mechanism for carrying out differential feeding modes aiming at different water layers. Although some researches attempt to introduce an underwater photographing technology, the method is limited to two-dimensional plane image analysis, and is difficult to accurately acquire the three-dimensional space position, the real body type size and the complex group behavior characteristics of the fish body, so that the accuracy of grouping judgment is low, and the fine directional feeding decision cannot be supported. Therefore, a method and a system capable of precisely acquiring three-dimensional characteristics of fish shoals by utilizing a stereoscopic vision technology, automatically dividing target groups with different specifications, and dynamically adjusting feeding modes and strategies according to water layers of the groups are required to be developed, so that precise, intelligent and ecological management of cage culture is realized. Disclosure of Invention In order to solve the problems in the background technology, the invention provides a cage fish grouping and directional feeding method and system based on stereoscopic vision. In order to achieve the purpose, the invention adopts the following technical scheme that the method for classifying and directionally feeding the cage fishes based on stereoscopic vision comprises the following steps: Acquiring fish shoal video data through a multi-eye underwater camera system, and processing the video data based on a binocular vision principle to generate fish three-dimensional point cloud data; Extracting body length and weight data of each fish individual from the fish three-dimensional point cloud data as growth characteristics, and dividing the fish group into at least two target groups with different growth characteristics by using a clustering algorithm in combination with behavior characteristic data; Generating a differential feeding strategy comprising feed types, feeding amounts and feeding areas according to the growth characteristics of each target group and current environmental parameters; And controlling the multi-outlet directional bait casting machine to accurately send the feed to the activity area where each target group is located according to the differential feeding strategy. Specifically, the combination behavior characteristic data specifically includes: And identifying the biological characteristics of the fish body by using a deep learning model, and continuously tracking each fish body in continuous video frames by using an interframe matching algorithm to obtain the swimming speed and the ingestion frequency as the behavior characteristic data. Specifically, th