高准确率的煤矸自动识别对综放开采智能放煤至关重要。针对灰度差异较小的难辨别煤矸种类，提出了“液体介入+红外检测”的煤矸识别新思路，进行了不同混合度下液体介入煤矸识别试验，获得了不同时刻煤矸混合试样的红外图像，借助ImageJ图像处理软件定量计算了煤矸图像的混合度，深入分析了煤矸混合度、介入时间、图像处理方法对液体介入煤矸识别准确率的影响，并从红外温变速率场的角度探讨了提高液体介入煤矸识别准确率的技术路径。研究结果表明：“液体介入+红外检测”方法可有效提高难辨别煤矸识别准确率。不同的煤矸混合度条件下，液体介入后红外图像中煤样所在区域均出现显著温降现象，可作为煤矸混合度自动识别的基础。当混矸率较低时，煤样温度在煤矸红外图像变化中起控制作用；当混矸率小于20%时，煤矸识别准确率随混矸率增大而增大，液体介入后10 s内的平均准确率约85.78%；混矸率在20%～60%内识别准确率较高且较为稳定，平均准确率约为94.38%，且液体介入时间对其影响很小，高准确率区域呈现“倾斜条带状”分布特征。当混矸率大于60%后，不同处理方法的识别准确率均值随混矸率增大呈下降趋势且离散性急剧增加；煤矸图像平均红外温差减小是煤矸混合放出后期识别准确率降低的根本原因，通过选取合理的液体种类、温度、介入量等参数来增大红外温度场的变化程度，可有效提高煤矸识别准确率。

Automatic identification of coal gangue with high accuracy is vital to intelligent top coal caving mining. Ai⁃ming at the types of coal gangue that are difficult to recognize with small grayscale differences, a new methodof coal/ gangue recognition based on “liquid intervention +infrared detection” was proposed. The infrared imagesof coal gangue mixed samples with different mixing degree and different time under the condition of liquid intervention were collected. The mixing degree of coal gangue image is quantitatively calculated based on the ImageJ software,and the influence of coal gangue mixing degree,intervention time and image processing method on the recognition ac⁃curacy were analyzed. The technical path to improve the coal/ gangue recognition accuracy with liquid interventionis discussed from the perspective of infrared temperature change rate field. The results show that for the typesof coal gangue with small gray difference in visible image,the method of “liquid intervention+infrared detection” canbe used to improve the accuracy of coal/ gangue recognition. Under different coal gangue mixing degree,there is a sig⁃nificant temperature drop in the area where the coal sample is located in the infrared image after liquid intervention,which can be used as the basis for automatic recognition of coal gangue mixing degree. When the gangue mixing rate islow,the coal sample temperature plays a controlling role in the change of infrared image characteristics of coalgangue mixture; When the mixed gangue rate is less than 20%,the recognition accuracy of increases with the increas⁃ing of gangue mixed ratio, and the average accuracy within 10 s after liquid intervention is about 85.78%;When the gangue mixing rate is in the range of 20%-60%,the recognition accuracy is high and stable,and the aver⁃age accuracy is about 94.38%,and the liquid intervention time has little influence on it. The high accuracy area pres⁃ents the distribution characteristics of “inclined strip”. When the gangue mixing rate is larger than 60%,the averagerecognition accuracy of different processing methods shows a downward trend,and the discreteness increases sharply;The reduction of average infrared temperature difference of infrared image is the root reason of recognition accuracy de⁃creasing in the later stage of mixed coal gangue drawing. The change degree of infrared temperature field can be in⁃creased by selecting reasonable liquid intervention parameters such as liquid type,temperature,intervention volume,soas to improve the recognition accuracy.