*4.4. Objective Evaluation Indicators*

Since it is difficult to obtain the normal illumination image corresponding to the image under abnormal illumination, for the image quality after enhancement, natural image quality evaluator (NIQE) [36] was used in this paper. NIQE is a non-reference image quality index often used to measure the quality of the image, a smaller NIQE indicating a better the quality of the measured picture. In addition, we used the lightness-order-error (LOE) [37] to evaluate the contrast of the enhanced image with the original illuminated image. LOE reflects the natural retention of the image, and a smaller value indicates that the image has a better order of luminance and therefore looks more natural. Table 6 shows the objective evaluation data of the corresponding methods in Figure 1. From the data in the table, we can see that the LOE of our algorithm is lower than that of the Laplace algorithm, and it is the lowest among all algorithms, indicating that we have the best result in maintaining the naturalness of the image. Furthermore, the NIQE value of the algorithm in this paper is the lowest among all algorithms, which indicates that the method in this paper does not produce much detail, thus blurring and color distortion to the original image.



#### **5. Conclusions and Future Work**

In this paper, we propose a deep learning mineral identification method based on luminance equalization. Compared with traditional mineral identification methods, we reduce the reliance on the researcher's experience and instruments. Compared with traditional mineral identification algorithms, we reduce the influence of illumination intensity on mineral identification and greatly improve the accuracy rate. In the deep learning recognition part, we used YOLOv5 to further improve the identification accuracy. During model selection, we used the optimized YOLOv5 to further improve the identification accuracy. In the future, more features will be introduced, such as combining the density and transparency of minerals with photos to further improve the accuracy of mineral identification. However, the identification method mentioned in this paper has some limitations: when the input picture is a mineral other than fifty minerals, the closest one among fifty minerals will be given. In the future, we will collect more mineral data to address this issue.

**Author Contributions:** Conceptualization, T.L.; Data curation, J.Z. and H.L.; Funding acquisition, T.L.; Methodology, J.Z.; Supervision, T.L.; Visualization, J.Z., Q.G. and H.L.; Writing—original draft, J.Z. and Q.G.; Writing—review and editing, T.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by National Natural Science Foundation of China under Grants No. 62002332, 62072443.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing is not applicable.

**Acknowledgments:** Thanks are given to Zhi Wang, Zhujun Nie, and Zexin Wu for their help in collecting the mineral samples.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**

