Chinese mitten crab is a unique aquaculture species in China. Its weight is not only an important basis for determining the feeding amount, but also an important indicator for judging its growth status and quality. Taking Chinese mitten crab as the research object, a method for estimating its weight based on multi-dimensional features and light gradient boosting machine (LightGBM) was proposed. Firstly, image segmentation were carried out on these collected crab images to obtain the carapace images. Then the geometric features of the carapace binary image was extracted as shape features (SF), extracting each channel component value of carapace images in different color spaces as color features (CF), and feature values were calculated by the calibration method. Finally, the crab weight was estimated by the LightGBM algorithm. The color feature and shape feature were extracted to form multi-dimensional features to solve the problem of low prediction accuracy caused by a single shape feature. The shape feature consisted of different carapace contour ratios, which effectively reduced the impact on the stability of the feature value caused by the random adjustment of the camera height. The proposed Chinese mitten crab weight estimation method was tested on the real dataset with the mean absolute error (MAE) of 2.751g, the root mean square error (RMSE) of 3.680g and the coefficient of determination (R2) of 0.949. Furthermore, when compared with the SF-LightGBM, SF3-LightGBM, area-OLS, MF-BPNN and MF-SVM crab weight estimation methods, the performance of each evaluation metric of the proposed method was improved. The experimental results indicated that the proposed method can accurately estimate the Chinese mitten crab weight.