Abstract: To address the problems of insufficient adaptability of image acquisition parameters, difficulty in simultaneously optimizing reconstruction time and peak signal-to-noise ratio, redundant input images, and low utilization of effective information in crop multi-view three-dimensional reconstruction, this study proposed an adaptive image acquisition strategy integrating multi-objective optimization and crop information proportion. The strategy was designed for wheat three-dimensional reconstruction, aiming to optimize imaging parameters and image screening at the front-end image acquisition stage, reduce redundant inputs, improve the utilization of effective crop information, and enhance reconstruction quality while controlling computational cost. Three wheat cultivars covering the entire growth cycle were used as research objects. Multi-view data were acquired using a multi-view acquisition platform under different plant height and canopy width conditions and different combinations of auxiliary camera height, camera angle, and number of reconstruction images, followed by three-dimensional reconstruction experiments. Wheat plant height, canopy width, auxiliary camera height, auxiliary camera angle, and the number of reconstruction images were used as input variables, while reconstruction time and peak signal-to-noise ratio were used as output variables. An ant colony optimization (ACO)-based support vector regression (SVR) prediction model was constructed to predict reconstruction time (Time) and peak signal-to-noise ratio (PSNR) and to characterize the nonlinear relationship among crop morphological parameters, imaging parameters, and reconstruction performance. On this basis, a bi-objective optimization model was established with the objectives of minimizing reconstruction time and maximizing PSNR. A multi-objective particle swarm optimization (MOPSO) algorithm was used to generate a non-dominated solution set, and the technique for order preference by similarity to ideal solution (TOPSIS) was combined to determine auxiliary camera height, auxiliary camera angle, and the number of reconstruction images from the candidate solutions, thereby forming a coarse strategy to achieve an initial balance between reconstruction efficiency and quality. Considering that the coarse strategy only determined macroscopic imaging parameters and image quantity but could not further evaluate the effective crop information contained in individual images, crop information proportion was further introduced as a quantitative indicator of effective image information. Combined with a Top-K mechanism, images with higher crop information proportion were selected within different views and angular intervals, thereby forming a precision strategy. This strategy further screened effective reconstruction images while maintaining the original multi-view distribution, thereby achieving progressive optimization from imaging parameter optimization to image information screening. The results showed that, compared with the uniform sampling baseline strategy, the coarse strategy reduced reconstruction time by 42.34% on average and increased PSNR by 3.20% on average. Compared with the same baseline strategy, the precision strategy reduced reconstruction time by 44.35% on average and increased PSNR by 6.49% on average, while reducing the relative errors of plant height and canopy width by 46.44% and 36.69%, respectively. Compared with the coarse strategy, the precision strategy further increased PSNR by 3.19% without increasing the number of input images or noticeably increasing computational overhead, and further improved the reconstruction accuracy of geometric phenotypic traits. These results indicated that the coarse strategy reduced redundant image inputs through joint optimization of imaging parameters and image quantity, whereas the precision strategy further improved the utilization of effective crop information through image screening. The proposed strategy can balance three-dimensional reconstruction efficiency and quality and provide a methodological reference for image acquisition in high-throughput phenotyping of complex-structured crops represented by wheat.